Home

Embedding data in an audio signal, using acoustic OFDM

1. Repeat x an gt gt gt E gt Repeat Downsample1 Signal To Buffer Workspace FDATocl musict wav Select Repeat Audi A 44100 Hz 16 bi stereo Pl arms gt La S Y gt y 12 gt E Multiport Repeat Dovinsample pl 2 From audio Selector Digtal Butter au Filter Design Signal To Workspace2 cu gt power control int outi bn out Dyin out Ita bin num Din ou in an Binz adi up afam mad add of sional rom channel coding dbpsk mod i Workspace ber Signal To audio TX Eror Rate Workspaces praedam scope auon Error Rate Display To Audio Device sRet Find ejay Dl E 08 Delay En manual channel Find Delay ES 6 FDATcol ro e owi im fag om im k Our int fe cutr int he cuts nt he our nt fe la Signa To channel decoding dapsk demod remove pilot ofdm demod remove op down Workspace 1 Digtal Filter Design2 Figure 4 2 The structure of the simulation system 4 1 Simulation System 27 4 1 1 Information Source As mentioned above there are two inputs one is the useful data and another one is an audio signal The whole system is running under the frame mode After several tests finally we use 11 samples in one frame as the useful data input and the frame interval is set to be N 1024 F gt 24100 0 023 s where
2. set param Acoustic ofdm DBPSK FINAL without AUDIO S tp sw else set param Acoustic ofdm DBPSK FINAL without AUDIO S audio set handles checkbox2 enable on end function checkbox2 Callback hObject eventdata handles val get hObject value if val set param Acoustic ofdm DBPSK FINAL without AUDIO S ofdm modi powercontrol sw O sw 0 0 sat TE 3 set_param Acoustic_ofdm_DBPSK_FINAL_without_AUDIO_S tp sw 0 set handles dis enable off set handles dis string 0 else set_param Acoustic_ofdm_DBPSK_FINAL_without_AUDIO_S ofdm modl powercontrol sw 1 set_param Acoustic_ofdm_DBPSK_FINAL_without_AUDIO_S tp sw set handles dis enable inactive end function axes3_CreateFcn hObject eventdata handles axes hObject imshow ISY gif function audname Callback hObject eventdata handles function audname CreateFcn hObject eventdata handles if ispc amp amp isequal get hObject BackgroundColor get 0 defaultUicontrolBackgroundColor set hObject BackgroundColor white end function pushbutton4 Callback hObject eventdata handles filename filepath uigetfile x wav select an audio file allpath strcat filepath filename if isequal filename 0 msgbox nothing has been selected warn else set handles audname string filename set param Acoustic ofdm DBPSK FINAL w
3. B 2 Measurement System function varargout gui_Singleton 1 acoustic_measurement varargin 7 0 rl B 2 Measurement System 71 gui State struct gui_Name mfilename gui Singleton gui Singleton gui OpeningFcn Qacoustic measurement OpeningFcn gui OutputFcn Gacoustic measurement OutputFcn gui LayoutFcn Cl y gui Callback ID if nargin amp amp ischar varargin 1 gui State gui Callback str2func varargin l end if nargout varargout l nargout gui mainfcn gui State varargin else gui mainfcn gui State varargin end function acoustic measurement OpeningFcn hObject eventdata handles varargin handles output hObject set handles checkbox2 enable off load bark mat assignin base bark bark guidata hObject handles open system Acoustic ofdm DBPSK FINAL without AUDIO R open system Acoustic ofdm DBPSK FINAL without AUDIO S set param Acoustic ofdm DBPSK FINAL without AUDIO S ofdm modl powercontrol sw 0 set_param Acoustic_ofdm_DBPSK_FINAL_without_AUDIO_S tp sw 0 set param Acoustic ofdm DBPSK FINAL without AUDIO S audio sw O set param Acoustic ofdm DBPSK FINAL without AUDIO S Gain2 gain 30 set param Acoustic ofdm DBPSK FINAL without AUDIO S up Gain2 gain 10 function varargout acoustic measurement OutputFcn hObject eventdata handles varargout l handles outp
4. Block type Comb type gt Aouanbay Aouanba y 00000000000 BODOBDOVOBEOOOQ 00000000000 00000000000 00000000000 00000000000 00000000000 00000000000 00000000000 00000000000 00000000000 00000000000 00000000000 060000000000 00000000000 00000000000 00000000000 00000000000 time a 3 Figure 3 5 Pilot arrangement For most of the time we can use pseudorandom signals as the pilots A pseu dorandom signal is a kind of signal that appears to be random but it is not Pseu dorandom sequences have statistical randomness however they are generated by certain deterministic causal processes This process will restrict the length of the randomness and after that the process will produce exactly the same sequence Pseudorandom signal is very useful for lab testing or experimental verification The process to generate pseudorandom signals is easier to produce than a truly random one So we alway use pseudorandom signals to substitute the real random signals Due to the phase ambiguity problem pilots are used to do the channel estima tion in most of the OFDM systems The phase of the modulated signal will be changed after the channel We can not adjust the phase error efficiently because it is impossible to predict the phase change precisely by measuring the useful signal 18 Method itself Therefore we transmit pilots which have known phases in the channel and then measure the phase change of pilots as an refer
5. second 3 choose whether to embed the acoustic OFDM signal in the audio or not 4 choose whether to control the power of the acoustic OFDM signal 5 choose whether to listen to the audio or view the dynamic power spectrum graph when the system is ongoing 6 input the message that you want to transfer in the system 7 start the system 8 observe the received message which should be the same as the one you input in the sixth step 9 10 11 more simulation results will be displayed here 12 13 you can play or stop the original audio without running the simulation system which should be like a wave player 14 you can exit the system by clicking this button 63 User Manual of The System GUI NS E GS UN A to subcarriers 33 4 pilots wa o bandwidth 6400 8000Hz l pa symbol interval 1024 CP 600 lt Modulation Method DBPSK ka g channel coding convolutional coding interleaving Nes uw timing synchronization coarse synchronization C Gn audio select play stog g lation time a 0 iG bit error rate F embedded in audio power control audio distortion 0 mode switch gt audio Gs Message to be transferred 5 Message received Kon Fe Department of Electreal Engineering Figure A 1 The GUI for the simulation system Simulation Result transmission rate 0 bivs 3 Thesis Project Embedding
6. set hObject BackgroundColor white end
7. Amplitude L 0 0 5 1 15 2 2 5 3 3 5 Time s x10 Autocorrelation of the received signal autocorrelation peak for Bark code gt a T o gt Amplitude 0 0 5 1 15 2 2 5 3 3 5 Time s x10 Figure 4 21 The way to do the timing synchronization by using Bark code threshold with the value 0 05 to decide the position of the peak Since we only use Bark code with length 2 in one frame at the beginning of the audio signal and pad zero sequences after that so there is only one peak value which should appear at the beginning of the audio signal 4 2 3 GUI There are many similarities between the simulation system and the measurement system so the GUI of the measurement system is almost the same as the one used in the simulation system except that the receiving mechanism is a little different For the measurement system the receiver and the signal processing module will run asynchronously So the GUI in the system will have two buttons to control the data transmission and the signal processing individually Figure 4 22 illustrates the GUI for the measurement system and more details about how to operate this 4 2 Measurement System 43 system can be found in Appendix A Thesis Project Embedding data in an audio signal itani Department of Electrical Engineering mang sevwa OFM SHUAI WANG shuwa259 student liu se parameters result audio selection transmission rat
8. The frequency conversion is an important step in a real communication system It composes of up conversion and down conversion In normal conditions the digital signal processing will be handled at baseband which is not suitable for radio communication and long distance transmission Consequently for the transmitter we need to use an up converter to shift the baseband signal to high frequency band For the receiver we also need a down converter to revert the signal from high frequency band to baseband Up Conversion In the experiment we use the baseband OFDM signal to modulate the sine and cosine waves at the carrier frequency 7 2kHz These signals will be summed as Implementation 30 Unipolar to Bipolar power control PN Sequence Unipolar to Generator Bipolar Multiport Selector add pilots powercontrol ofdm implement Figure 4 6 The structure of power control for the second step 4 1 Simulation System 31 complex zeros 1 1 Matrix Concatenate1 Figure 4 7 The DC distribution in the subcarriers Multiport Selector Matrix Concatenate1 Figure 4 8 The structure of OFDM modulation 32 Implementation the up converted signal and then transmitted from the transmitter Figure 4 9 illustrates the procedure of up conversion Sine Wave Complex to Resl Imag Product Sine Wave Figure 4 9 Up conversion Down Conversion For down conversion the s
9. aud else msgbox please select an audio source error warn end function pushbutton5 Callback hObject eventdata handles global aud aud name get handles audname string if isempty aud name stop aud else msgbox no audio needs to stop error warn end function trans Callback hObject eventdata handles function trans_CreateFcn hObject eventdata handles if ispc amp amp isequal get hObject BackgroundColor get 0 defaultUicontrolBackgroundColor set hObject BackgroundColor white end function ber_Callback hObject eventdata handles function ber CreateFcn hObject eventdata handles if ispc amp amp isequal get hObject BackgroundColor get 0 defaultUicontrolBackgroundColor set hObject BackgroundColor white end function dis Callback hObject eventdata handles function dis_CreateFcn hObject eventdata handles if ispc amp amp isequal get hObject BackgroundColor get 0 defaultUicontrolBackgroundColor set hObject BackgroundColor white end function editll Callback hObject eventdata handles 70 Programming Codes function pushbutton6 Callback hObject eventdata handles filename filepath uigetfile x wav select an audio file allpath strcat filepath filename if isequal filename 0 msgbox nothing has been selected warn else set handles audname string filename set param Aco
10. 1 illustrates the system parameters Subcarriers 33 4 pilots OFDM carrier frequency 6400 8000 Hz Symbol interval 1024 samples Cyclic prefix 600 samples Modulation method DBPSK Channel coding Convolutional coding Interleaving Timing synchronization Coarse synchronization Sampling frequency 44100 Hz Data rate 896 bit s Table 4 1 The System Parameters 11 The data transmission mode in the experiment is frame based One frame con tains one OFDM symbol and the length of one frame is 1024 samples The source information has 11 samples in one frame at the beginning Since the error correc tion code rate is 1 3 so the useful data in one frame will become 33 samples after the channel coding which is also the number of subcarriers for data carrying and another 4 pilots will be distributed in these 33 subcarriers for channel estimation 12 As we designed in the experiment there are 37 subcarriers which will be used to transmit information in the bandwidth of 1600Hz and other 987 samples in the OFDM symbol will be padded with zeros So we set the OFDM symbol interval 23 24 Implementation as 1024 samples and the sampling frequency as 44100 Hz The bandwidth of each subcarrier can be calculated as follows F 441 on 43 Hz N 1024 where N is the symbol interval and F is the sampling frequency So the total bandwidth for the data transmission in this system is F 44100 Tal Nsub kan 1594 H y 2004 37 1
11. ANAL EN ee are Baia fan fr Sk BAR so 33 GN IGUANA e Geode 35 4 1 10 RECAVET 4 43 SS reale 35 4 2 Measurement SysteM 37 4 21 lA e e A Rol 37 4 2 2 Timing Synchronization o o 37 4 23 GUD a4 rai r to eo lan ii kat man gan vey EEE ta A A TR 42 5 Evaluation 45 5 1 Test Environment 45 5 25 Simulation Result 4 4 4 dry a bob ee 45 5 2 1 Power Spectrum enn 45 932 2 Bit Error Rate wo 22 2 2 ki ke ae aou ea ma man api Re Z V 48 5 2 3 Audio Distortion 49 5 3 Measurement Result 51 531 Audo Type ss at 2 AN W a ae f n n eS 52 5 3 2 Distance between Sender and Receiver 52 57373 Sound Level lt lt a ie More Pate A Gk 53 3A SD Rocke ne ae Gd See Si a A Lt te WE Ko den n dann Gan an tat B R 54 5 3 5 Audio Distortion 56 5 3 0 Limitations cs cc degree a ee Sik v r Be we eee ke dwati d v d d es 58 6 Conclusions and future work 59 Gal Conclusions e u or a a we eh Ai a A ke a de eS Res Ses 59 6 2 Future Work si ni oe matant h l SN ne AN ee ee en da AA 59 Bibliography 61 A User Manual of The System GUI 63 A 1 Simulation System o eces sba 00000000222 ee 63 A 2 Measurement System 2 2 20 02 00 ss ss ee ee ee 65 B Programming Codes 66 B 1 Simulation System 2 0000 66 B 2 Measurement System como ses
12. N is the symbol interval and F is the sampling frequency For the audio signal the length of the frame is set to 1024 and the sampling frequency is 44 1kHz so the frame interval is also around 0 023s In Figure 4 2 a selector module is used to select one of the audio channels to do the signal process ing if the audio source is a stereo type and the other channel will be discarded Since what we want to survey is whether the audio quality is influenced by the acoustic OFDM signals it is not so important to choose stereo audio or mono audio And the selection step will also decrease the complexity of the simulation system 4 1 2 Channel Coding Channel coding is achieved by convolutional coding and interleaving Figure 4 3 illustrates the way that these two parts work together Convolutionsl Encoder Convolutionsl interleaving Encoder Figure 4 3 The structure of channel coding The convolutional code used in this experiment has the rate of 1 3 and the constraint length of 7 In Matlab we can realize this convolutional code by using the command poly2trellis 7 171 133 157 In this command the first number 7 represents the constraint length and the last three numbers which are the default values in the system will compose the code generator with the rate of 1 3 After the coding step the length of the frame will increase from 11 to 33 samples due to the coding rate and the frame interval is still 0 023s Two types of interlea
13. The three equations above are binary addition which means that the inputs and outputs are binary and the addition calculation will follow the binary addition rules The convolutional code adopted in this experiment has a rate of 1 3 and con straint length of 7 Its specific parameter configuration will be elaborated on in the following chapters 3 2 2 Interleaving Bit errors often happen continuously in the process of signal transmission and that is because some fading valley points which last for a long time will influence the continuous bit information Moreover most of the channel coding methods can implement the error correction effectively only for signal error or errors with short length As a result in order to correct the burst errors interleaving is needed to disperse these errors 8 In this way long string bit errors can be transformed into short string bit errors which will make the forward error correction more effective Figure 3 3 shows the working principle of a pattern of interleaving ae _ by Figure 3 3 The algorithm of Interleaving by using matrix where the data is written column wise and read row wise gt DataInput Data Output The interleaving mode applied in this thesis is general block interleaving namely to interleave the rows or columns of the original bit matrix in accordance 16 Method with certain rules so that the data of the original rows or columns are switched to
14. as the one used in the simu lation system The model of the measurement system is divided into two parts which are transmitter and receiver Figure 4 15 illustrates the transmitter of the measurement system From the figure above we know that the channel used in the simulation system is substituted by an audio player block and an audio recorder block which are corresponding to a loudspeaker and a microphone in the real life channel Figure 4 16 illustrates the module of the receiver From Figure 4 15 and Figure 4 16 we know that the measurement system is similar to the simulation system except the channel So each branch of the measurement system can refer to that of the simulation system 4 2 1 Channel As mentioned above the channel used in the measurement system is composed of a loudspeaker and a microphone which together substitute the simulation channel In figure 4 17 and figure 4 18 we find that the speaker and the microphone both are connected with a laptop There is no obstacle between the loudspeaker and the microphone and the distance between them is around 1 meter which is changeable 4 2 2 Timing Synchronization As we mentioned in chapter 3 there is no timing synchronization for the simulation system However for the measurement system we need an appropriate mechanism to obtain the start position of the useful data In this experiment we utilize a coarse timing synchronization method for which the Bark code will be hidden in
15. data in an audio signal mamm Department of Electrical Engineering using acoustic OFDM SHUAI WANG shuwa259 student liu se parameters resur audio selection i a transmission rate 0 bis ale jeene audio distortion 0 E embedded in audio poner cate Message to be transferred G Message derived Figure A 2 The GUI for the measurement system A 2 Measurement System 65 A 2 Measurement System The operation steps in the measurement is similar with the one in the simulation which can be followed as Figure A 2 1 Cia en 6 choose the audio file which is used to transmit the useful data choose the simulation time with the unit second choose whether embed the acoustic OFDM signal in the audio choose whether to control the power of the acoustic OFDM signal input the message that you want to transfer in the system start to transmit the acoustic OFDM signal embedded in the audio signal through the real channel 7 8 try to receive and do the signal processing after the audio transmission observe the received message which should be the same as the one you input in the sixth step 9 10 more simulation results will be displayed here 11 12 you can play or stop the original audio without running the simulation system which should be like a wave player 13 you can exit the system by clicking this button Appendix B Programming Codes B 1 Simulation System fun
16. of pseudo random noise into the low frequency band of the audio signal and then at the receiver an autocorrelation module will be utilized to derive a peak point which will be the coarse start point of the useful symbols This principle is mostly utilized in the measurement system and in simulation system we use simulation channel to substitute the real channel Moreover for the simulation system the influences from Doppler Effect will not be considered and the environment noise will also 3 7 Timing Synchronization 21 be set to a low level Therefore we do not use coarse synchronization technology in the simulation system The specific implementation of these systems will be explained clearly in the following chapter Chapter 4 Implementation The experiment in the thesis is composed of two systems which are simulation system and measurement system For the simulation system all the functionalities such as the channel coding decoding modulation demodulation and power control are achieved in Matlab Simulink The main target for designing this system is to test the performance of an acoustic OFDM system The noise tolerance is also a factor to be considered For the measurement system the signal processing part is completed by Matlab Simulink and the simulation channel will be substituted by a real life channel which consists of a microphone and a loudspeaker or a pair of audio input output devices Table 4
17. power of the useful data which is showed in Figure 4 6 Here in Figure 4 7 we can see that there are 38 subcarriers which consists of 33 useful data subcarriers four pilot subcarriers and one DC carrier with complex zero Here we use complex zero to simulate the DC carrier In the real life system the null DC carrier can be used to eliminate the DC offset by utilizing a DC block filter since there is no information around DC 13 In the simulation system it does not matter whether we use this DC carrier or not After all the operations above the frame interval of the useful data is still 0 023s 4 15 OFDM Modulation In the experiment the OFDM modulation is implemented in two steps The first step is the pilot distribution in the data subcarriers In the second step we need to insert the DC carrier and then pad zeros in the tail of each frame to make sure that each frame consists of 1024 samples Next IFFT will be executed to finish the OFDM modulation After this kind of modulation the data can be transformed from frequency domain to time domain meanwhile the signal will be modulated to the low pass band with the bandwidth of 1600Hz The OFDM modulation increases the length of the data frame but does not affect the duration of a frame interval So the length of the frame after the OFDM modulation will be 1024 samples but the frame interval is still 0 023s Figure 4 8 illustrates the module of OFDM modulation 4 1 6 Frequency Conversion
18. the M file Simulink modules will be responsible for the signal processing of the whole system GUI is an user interface which can be used by the customers to control the system and also display the simulation results Based on the theoretical structure of the system mentioned in the previous chapters we have established a real life simulation model as showed in Figure 4 2 It is verified that this system can work steadily and robustly The whole system has two inputs one is the binary sequence which is converted from the text message on the GUI and another input is the audio resource which can also be selected from the GUI The only audio type that can be used here is wav format The system has three outputs As we can see in Figure 4 2 aul and au2 are used to calculate the audio distortion and rec is the useful binary information obtained from this simulation system In this system some useful blocks such as the bit error rate block and the spectrum view block are utilized to assist surveying the status of the system All these blocks can work without the GUI controller More details about the implementation of each subsystem will be given in the following subsections Implementation 26
19. the beginning of the audio signal 14 The hidden means that we should control the power of the Bark code to ensure that it will not affect the audio quality We use Bark code because it has the feature of high autocorrelation So it is easy to find a peak point at the start position of the useful data after an autocorrelation operation Bark code in this experiment will be combined with the audio signal directly in time domain which can be illustrated in Figure 4 15 From the figure we know that the Bark code is combined with the original audio signal in a simple way i e an addition operation The power of the Bark code will be controlled by a constant gain which can make sure that the Bark code is inaudible during the transmission of the audio signal Figure 4 20 shows Bark code in time domain and the signal after an autocor relator From this figure we know that Bark code is highly correlated and the maximal value of its autocorrelation function appears at the beginning of Bark code signal So after we combine the audio signal and Bark code we can use an Implementation 38 Signal To From Microphone Workspsce Signal From Workspace music2 wav A 44100 Hz 16 bit Multiport Downsample From audio Selector Digits Filter Design To Audio Device1 Signal From Workspace1 channel coding dbpsk mod1 Figure 4 15 The transmitter of the measurement system 39 4 2 Measurement S
20. the limitations of the experimental environment and the experiment devices the control of the bit error rate and the audio distortion is not in an ideal degree So the stability of these two systems should be improved in the future Especially for the measurement system the accuracy of deriving the start position of the useful data is low due to the use of the coarse timing synchronization and it also affects the stability of the measurement system or causes the system to crash in some unusual conditions Chapter 6 Conclusions and future work 6 1 Conclusions This thesis has mainly researched how to embed data in an audio signal using acoustic OFDM The key point of this technology is that the combination of the audio signal and the useful data will not affect the audible quality of the audio signal In the experiment we established two systems which are simulation sys tem and measurement system These two systems can fulfill the requirements of this thesis and work robustly in some proper conditions These two systems can transmit the standalone OFDM signal as well as the OFDM signal embedded in an audio signal The data rate in the system can reach almost 0 9kbit s In the experiment we also tested the systems from different aspects such as the audio type and the channel distance to analyze how these factors can affect the system performance As a result we found that the measurement system can still trans mit the acoustic OFDM signal cor
21. the power of the useful data Figure 4 5 gives an illustration of the first part of power control The audio signal is transformed from time domain into frequency domain by FFT and then the spectrum values between 6400Hz and 8000Hz will be extracted and prepared for the next step Er Frame In a Frame Conversion Selector Figure 4 5 The mechanism of power control for the first step The second step of the power control module will be integrated with the OFDM modulation block That is because the power control should be achieved after the pilot distribution in the useful data subcarriers Consequently the power control will work on 37 subcarriers in which there are 33 data subcarriers and 4 pilot 4 1 Simulation System 29 subcarriers The method to control the power of the symbols is illustrated in Figure 4 6 From Figure 4 6 we know that the power control is equivalent to the spectrum control In this step a threshold value should be set for the modulated symbols because the power of the audio signal could drop down to zero however the power of the useful data should not be zero after the power control So we use a saturation block to control the lower limit of the control information A large value of the lower limit will bring high bit error rate In the experiment we set the lower limit as le 3 W Hz to restrict the power ratio between the audio and the useful data The output of the saturation block will be used to control the
22. 4_Callback hObject eventdata handles function edit4_CreateFcn hObject eventdata handles if ispc amp amp isequal get hObject BackgroundColor get 0 defaultUicontrolBackgroundColor set hObject BackgroundColor white end function checkbox5_Callback hObject eventdata handles function checkbox6_Callback hObject eventdata handles function edit6_Callback hObject eventdata handles function edit6_CreateFcn hObject eventdata handles if ispc amp amp isequal get hObject BackgroundColor get 0 defaultUicontrolBackgroundColor set hObject BackgroundColor white end function checkbox2_Callback hObject eventdata handles val get hObject value if val 0 set param Acoustic ofdm DBPSK FINAL AUDIO ofdm mod powercontrol sw 0 set handles dis string 0 set handles dis enable off B 1 Simulation System 69 else set param Acoustic ofdm DBPSK FINAL AUDIO ofadm mod powercontrol sw 1 set handles dis enable inactive end function edit7 Callback hObject eventdata handles function edit7_CreateFcn hObject eventdata handles if ispc amp amp isequal get hObject BackgroundColor get 0 defaultUicontrolBackgroundColor set hObject BackgroundColor white end function pushbutton4_Callback hObject eventdata handles global aud aud_name get handles audname string if zisempty aud_name play
23. 5 2 3 Audio Distortion For the simulation system another factor which should be considered in the exper iment is the audio distortion If we correctly transmit the messages at the cost of high audio distortion we will say that the experiment is failure For our designed systems the basic requirement is that the data embedment will not affect the au dio quality a lot which means that the audio distortion will be under an acceptable level We calculate the audio distortion by computing the mean square value of the difference between the original audio signal and the audio signal combined with the useful message The following equation illustrates how to calculate the audio distortion TAP 2 Daudio N Yo si n So n gt n 0 where Daudio is the audio distortion N is the time interval with the unit second si n is the original signal and so n is the audio signal combined with the useful signal This value is varying over time Figure 5 7 shows the comparison of the audio 50 Evaluation distortion between the power controlled signal and the non power controlled signal in case that the power of the standalone OFDM signal is 85dBW Hz and the average power of the acoustic OFDM signal is around 85dBW Hz i i amp no power control with power control 0 025 si koman FT ee Audio Distortion Time s Figure 5 7 The comparison of audio distortion for power controlled system and
24. 594 Hz where Nsus is the number of subcarriers Due to the existence of a cyclic prefix the final length of one data frame is 1624 samples Meanwhile DBPSK can generate one data symbol from one bit or we can say one symbol stands for one information bit So the data rate for this system is showed in the following expression F 44100 Naata A e ao 896 bits s 0 9 kb NAN 1024 7 600 t ai where Naata is the number of useful data subcarriers and Npre is the length of cyclic prefix From the result of the equation we know that the system has the ability to transmit some kind of short information in a few seconds As we mentioned above the pilots are used for the channel estimation and they need to be distributed in the useful data subcarriers Figure 4 1 shows the distribution of these four pilots As we know from the figure the bandwidth will be divided into 37 subcarriers in time Hi pilot C useful data gt frequency Figure 4 1 The pilot distribution in the subcarriers the frequency domain Four subcarriers with the indexes of 1 13 25 37 will be used to transmit the pilot information which is a pseudo random sequence In the experiment we do not need to do the phase recovery due to the usage of DBPSK modulation However the pilots are still added in to be an optional configuration in case that we use some other modulation methods such as BPSK or QPSK The experiment environment including the software and hardwa
25. Department of Electrical Engineering Link pings universitet SE 581 83 Link ping Sweden Institutionen f r systemteknik Department of Electrical Engineering Examensarbete Embedding data in an audio signal using acoustic OFDM Examensarbete utf rt i Kommunikationssystem vid Tekniska h gskolan i Link ping av Shuai Wang LiTH ISY EX 11 4518 SE Link ping 2011 Link pings universitet TEKNISKA H GSKOLAN Link pings tekniska h gskola Link pings universitet 581 83 Link ping Embedding data in an audio signal using acoustic OFDM Examensarbete utf rt i Kommunikationssystem vid Tekniska h gskolan i Link ping av Shuai Wang LiTH ISY EX 11 4518 SE Handledare Erik Axell ISY Link pings universitet Examinator Mikael Olofsson ISY Link pings universitet Link ping 12 October 2011 cs UNI Avdelning Institution Datum OS U D Division Department Date y A u m Yo Division of Communication Systems u a Department of Electrical Engineering 2011 10 12 e RS Link pings universitet PA S CE oui SA H SE 581 83 Link ping Sweden Spr k Rapporttyp ISBN Language Report category Svenska Swedish Licentiatavhandling ISRN K Engelska English amp Examensarbete LiTH ISY EX 11 4518 SE C uppsats Serietitel och serienummer ISSN D uppsats Title of series numbering ai vrig rapport URL f
26. I 11 R Prasad OFDM for Wireless Communication Systems Artech House Norwood MA 2004 H Matsuoka Y Nakashima and T Yoshimura Acoustic communication with ofdm signal embedded in audio In 29th Audio Engineering Society Conference Seoul Korea Sep 2006 B Li S Zhou M Stojanovic L Freitag and P Willett Multicarrier commu nication over underwater acoustic channels with nonuniform doppler shifts IEEE Jornal on Oceanic Engineering 33 198 209 2008 D Gruhl A Lu and W Bender Echo hiding In Information Hiding 96 pages 293 315 1996 Y Yardimic A E Cetin and R Ansari Data hiding in speech using phase coding In ESCA Eurospeech97 pages 1679 1682 Greece Sept 1997 L Boney A H Tewfik and K N Hamdy Digital watermarks for audio signals In IEEE Intl Conf on Multimedia Computing and Systems pages 473 480 March 1996 U Madhow Fundamentals of Digital Communication pages 294 310 978 0 521 87414 4 Cambridge University Press 2008 J J Kong and Keshab K Parhi Interleaved convolutional code and its viterbi decoder architecture EURASIP Journal on Applied Signal Processing 13 1328 1334 2003 J Proakis Digital Communications McGraw Hill 5 edition Aug 2000 H Matsuoka Y Nakashima and T Yoshimura Acoustic ofdm system and its extension In The Visual Computer volume 25 no 1 pages 3 12 Springer Berlin Heidelberg Oct 2009 Y Nakashima H Matsuoka and T
27. OFDM signal is 85dBW Hz There are two curves in the figure above The curve with circles represents the relationship between the bit error rate and the signal to noise ratio SNR for the standalone OFDM signal and the curve with triangles represents the relationship between the bit error rate and the signal to noise ratio SNR for the acoustic OFDM signal which is power controlled and combined with the audio signal From figure 5 6 we know that the bit error rate of the system will decrease when SNR of the system rises for the curve with triangles meanwhile the system with the standalone OFDM signal will achieve a better performance than the system with the OFDM signal embedded in an audio signal In the experiment we have tested three different messages showed below e hello world http www liu se e Embedding data in an audio signal using acoustic OFDM 5 2 Simulation Result 49 0 25 p 0 2 aki a Bit Error Rate 0 1 0 5 10 15 20 25 30 Signal to Noise Ratio dB Figure 5 6 The relationship between SNR and the bit error rate for the simulation system We found that the system with the standalone OFDM signal will correctly transmit and receive all the messages even when SNR is equal to 1dB and BER approximates with 0 05 For the system with the OFDM signal embedded in an audio signal the SNR value is around 8dB and BER is around 0 047 when the messages can be correctly and stably transmitted
28. UDIO up sl N num2str framaN set param Acoustic ofdm DBPSK FINAL AUDIO s2 N num2str frameN j sim Acoustic ofdm DBPSK FINAL AUDIO 0 str2num get handles edit7 string data_orig char bin2dec num2st r reshape rec 35 35 N 1 7 set handles receiver string data orig end msgbox please select audio or input the data that you want to send error warn return au aul au2 audiv mean au 2 set handles dis string audiv 68 Programming Codes end guidata hObject handles function pushbutton2_Callback hObject eventdata handles close function axesl_CreateFcn hObject eventdata handles axes hObject imshow LIU jpg function checkboxl_Callback hObject eventdata handles val get hObject value if val set handles checkbox2 enable off set handles checkbox2 value 0 set_param Acoustic_ofdm_DBPSK_FINAL_AUDIO audio sw 0 set_param Acoustic_ofdm_DBPSK_FINAL_AUDIO ofdm mod powercontrol sw 0 set_param Acoustic_ofdm_DBPSK_FINAL_AUDIO tp sw 0 set handles dis string 0 set handles dis enable off else set handles checkbox2 enable on set_param Acoustic_ofdm_DBPSK_FINAL_AUDIO audio sw 1 set_param Acoustic_ofdm_DBPSK_FINAL_AUDIO tp sw 1 end function axes2_CreateFcn hObject eventdata handles axes hObject imshow ISY gif function edit
29. Yoshimura Evaluation and demonstra tion of acoustic ofdm In Fortieth Asilomar Conference on Signals pages 1747 1751 Systems and Computers Oct 2006 61 62 Bibliography 12 O A Alim N Elboghdadly M M Ashour and A M Elaskary Simulation of channel estimation and equalization for wimax phy layer in simulink In Computer Engineering Systems 2007 ICCES 07 International Conference on pages 274 279 Nov 2007 13 D P Mcnamara Wireless communication unit and method for de offset estimation Technical report Wireless Technology Solutions LLC May 2011 http www freepatentsonline com y2011 0103449 html 14 J W Huang Y Wang and Y Q Shi A blind audio watermarking algorithm with self synchronization In Circuits and Systems 2002 ISCAS 2002 IEEE International Symposium on volume 3 pages 627 630 May 2002 Appendix A User Manual of The System GUI HOW TO START To run the system please open the main m file for the simulation system or the measurement system for example open acoustic simulation m for the simulation system and then click Debug gt Run acoustic simulation m or click F5 on the keyboard to run the system You can also simply type the file name in Matlab in case that the files are in the search path A Simulation System The operation steps can be followed as Figure A 1 1 choose the audio file which is used to transmit the useful data 2 choose the simulation time with the unit
30. als will be processed by FFT after the digital to analog conver sion Finally the original signals are obtained through demodulation and parallel to serial conversion Moreover the demodulator should utilize the demodulation mode corresponding to the modulator of the sender namely the demodulation mechanism for M PSK or QAM Figure 2 9 reveals the theoretical structure of the receiver for an OFDM receiver Symbol detection Re E y Y o n 9 807 O Fz Figure 2 9 The receiver for an OFDM system 4 Parallel to serial 2 2 Acoustic OFDM Based on the OFDM technology acoustic OFDM is used to transmit useful infor mation by making use of an acoustic channel This thesis mainly discusses how 2 2 Acoustic OFDM 11 to embed data in an audio signal by using acoustic OFDM without influencing the audio quality too much Therefore acoustic OFDM has two new features compared with OFDM First the power of OFDM signals should be under con trol Second some certain mechanism is required to integrate audio with OFDM signals 2 2 1 Power Control Power control is a core content in this thesis One way to embed signals into audio without impacting on the audio quality is to mimic the original audio signals by using power controlled signals The audio signals in high frequency band will be substituted by this kind of power controlled signals The whole process will be achieved in f
31. ations for this research orientation are stated and relevant suggestions are proposed Acknowledgments First of all I would like to thank my superviser Erik Asell for his kindness and patience to solve problems that I encountered during my thesis work I would also like to thank professor Mikael Olofsson for his guidance of the whole thesis procedure Secondly I would like to thank My lover Yingying Huang for her support in Sweden during my thesis work My parents for their moral support during my master study period Last but not least I would like to thank all my friends in Sweden who are accompanying with me during the past two years SHUAI WANG Link ping Sweden 2011 vil Abbreviations ASCII AWGN American Standard Code of Information Interchange Additional White Gaussian Noise Cyclic Prefix Differential Binary Phase Shift Keying Direct Current Frequency Division Multiplexing Forward Error Correction Fast Fourier Transform Guard Interval Graphical User Interface Inter Carrier Interference Inverse Fast Fourier Transform Inter Symbol Interference Long Term Evolution M Phase Shift Keying Orthogonal Frequency Division Multiplexing Quadrature Amplitude Modulation Quadrature Phase Shift Keying Signal to Distortion Ratio Signal to Noise Ratio Universal Resource Locator White Gaussian Noise Wireless Fidelity World Interoperability for Microwave Access Contents 1 Introduction 1 1 Problem Formula
32. auses ICI CI of SC2 on SC1 j I I Deere AAA A A IAOEE I I I Guard Interval i FFT Integral Time N I I gt Figure 2 4 The ICI problem for adjacent subcarriers In order to solve this problem a concept called cyclic prefix CP 1 should be introduced which is showed as Figure 2 5 If we replicate the data with the same length as GI from the end of the use ful data frame and substitute the original guard interval then the OFDM signals 8 Background of acoustic OFDM Time Figure 2 5 The distribution of cyclic prefix in OFDM symbols formed in this way can still keep the orthogonality between the neighboring sub carriers in the course of demodulation Accordingly inter channel interference can be avoided Figure 2 6 explains how CP ensures the orthogonality between the neighboring subcarriers of OFDM signals Subcarrier 1 Gl FFTintegraltime I Delayed subcarrier 2 of previous OFDM Figure 2 6 The effect of cyclic prefix for persistence of the orthogonality of subcarriers High Capacity for Fading Resistance Due to the existence of multiple carriers some of the subcarriers might be seriously influenced by the frequency selective fading and no data can be correctly transmit ted on these subcarriers However the subcarriers that are affected by the fading will still work steadily and robustly So the data transmission on these subcarriers will not be influenced too
33. cle in between these devices so this room can be considered as a sealed environment 52 Evaluation DODDA Figure 5 9 The scenario of the measurement system 5 3 1 Audio Type For decreasing the complexity of the experiment and enhance the operability of our system the only audio format we used in the measurement system is wav The wav format is a kind of high quality lossless audio format which is developed by Microsoft Corporation The tone quality of the audio in this format is similar with CD quality So this kind of audio signal will be appropriate to be used in our measurement system which can achieve a precise measurement on the audio distortion 5 3 2 Distance between Sender and Receiver One of the tests in our experiment concerns the transmission capability of the measurement system with varying channel distance Here we use assume that the power of the standalone OFDM signal is 85dBW Hz and the average power of the power controlled OFDM signal is around 85dBW Hz Figure 5 10 illustrates the testing results for the audio type of piano The horizontal axis in the figure represents the distance between the loud speaker and the microphone with the unit of meter and the vertical axis repre sents the bit error rate The curve with circles denotes the transmission of the standalone OFDM signal and the curve with triangles denotes the OFDM sig nal embedded in an audio signal From the figure we know that with the same di
34. ct eventdata handles function receiver_CreateFcn hObject eventdata handles if ispc amp amp isequal get hObject BackgroundColor get 0 defaultUicontrolBackgroundColor set hObject BackgroundColor white end function pushbutton3_Callback hObject eventdata handles a 1 sim starpoint tran evalin base tran N evalin base N val point max star 0 5e5 1 end tran2 tran point 0 5e5 end assignin base tran2 tran2 sim Acoustic ofdm DBPSK FINAL without AUDIO R data out char bin2dec num2str reshape rec 35 35 N 1 7 assignin base rec rec 35 35 N 1 set handles receiver string data out function edit3 Callback hObject eventdata handles function edit3_CreateFcn hObject eventdata handles if ispc amp amp isequal get hObject BackgroundColor get 0 defaultUicontrolBackgroundColor set hObject BackgroundColor white end function pushbutton3_ButtonDownFcn hObject eventdata handles function checkboxl_Callback hObject eventdata handles val get hObject value B 2 Measurement System 73 if val set param Acoustic ofdm DBPSK FINAL without AUDIO S audio set handles checkbox2 enable off set handles dis enable off set handles dis string 0 set handles checkbox2 value 0 set param Acoustic ofdm DBPSK FINAL without AUDIO S ofdm modl powercontrol sw 0
35. ction varargout acoustic ofdm varargin gui Singleton 1 gui State struct gui Name mfilename gui Singleton gui Singleton gui OpeningFcn acoustic_ofdm_OpeningFcn gui OutputFcn acoustic_ofdm_OutputFcn gui LayoutFcn EJ gui_Callback ID if nargin amp amp ischar varargin 1 gui State gui Callback str2func varargin l end if nargout varargout l nargout gui mainfcn gui State varargin else gui mainfcn gui State varargin end function acoustic_ofdm_OpeningFcn hObject eventdata handles varargin open system Acoustic ofdm DBPSK FINAL AUDIO set param Acoustic ofdm DBPSK FINAL AUDIO audio sw O set param Acoustic ofdm DBPSK FINAL AUDIO ofam mod powercontrol sw 0 set_param Acoustic_ofdm_DBPSK_FINAL_AUDIO ofdm mod sat LowerLimit le 3 set param Acoustic ofdm DBPSK FINAL AUDIO auon sw 1l set param Acoustic ofdm DBPSK FINAL AUDIO scope OpenScopeAtSimStart 0 set handles checkbox7 string mode switch gt audio guidata hObject handles set handles checkbox1 value 0 set handles checkbox2 enable off set param Acoustic ofdm DBPSK FINAL AUDIO tp sw O uicontrol handles sender function varargout acoustic_ofdm_OutputFcn hObject eventdata handles 66 B 1 Simulation System 67 varargout l handles output function sender Callback hObject eventdata handles function send
36. e 0 bitis play simulation time 5 audio distortion 0 E embedded in audio power control Message to be transferred Message derived Figure 4 22 The GUI for the measurement system Chapter 5 Evaluation 5 1 Test Environment The analysis of the experimental results will be divided into two steps due to that we have established two systems namely simulation system and measurement system First of all we need to know that the whole experiment will be processed in an enclosed environment and in these two systems the Doppler Effect will be ignored For the simulation system we will survey the influences on the SNR of the system and the audio distortion which are caused by the transmitted OFDM signal Since we only use one modulation method the system data rate will be fixed in the experiment which has been described in chapter 4 For the measurement system we mainly consider the factors of the environment noise and the distance between the loudspeaker and the microphone These ele ments will affect the bit error rate of the system and the audio distortion directly 5 2 Simulation Result 5 2 1 Power Spectrum To explain the working principle of acoustic OFDM we have carried out a series of testing to observe the signal spectrum Figure 5 1 illustrates the power spectrum of the standalone OFDM signal As we can see the bandwidth of the OFDM signal is around 1 6 kHz There is no power control on it
37. ement the sound level is measured in dBSPL or dBspr for which SPL stands for sound pressure level Mostly we can just use dB to indicate the sound level Sound pressure level is a logarithmic measure of a sound with certain sound pressure to a reference value It is measured in dB compared with a reference level Commonly we use 20RMS as the zero reference Here dBSPL or dBspr is not a SI unit but an approach to measure the sound pressure level In this figure the curve with circles represents the relation between the sound level and the bit error rate for the standalone OFDM signal and the curve with triangles represents this relationship for the OFDM signal embedded in an audio signal As we know from figure 5 11 the bit error rate of the whole system will have an attenuation when the sound pressure enhances continuously We also test different kinds of audio such as speech piano guitar and song in the measurement system and figure 5 12 shows the relationship between the sound level and the bit error rate of the system for these audio types Here we still set the average power of the power controlled OFDM signal as around 85dBW Hz From the figure we 54 Evaluation S standalone OFDM signal j ___ 4 OFDM signal embedded in audio Bit Error Rate 3 5 u E Ga A 75 80 Sound Level dBSPL Figure 5 11 The transmission capability of the measurement system know that the system can achieve the best performa
38. ence for the phase adjustment of useful signals Channel estimation can correct the phase ambiguity problem decrease the bit error rate and enhance the system stability 3 5 Power Control To understand the functionality of power control we can refer to Figure 3 6 which compares the difference when using power control or not power power without power control Audiosignal Eliminated part frequency Low band i Highband i i power power with power control frequency frequency Figure 3 6 The power spectrum of the system with and without power control From the figure above we know that if we combine the audio signal and the OFDM symbols without power control the spectrum of the mixed signal in the high frequency band will maintain at a constant value This situation will bring a lot of noise to the audio signal However if we use power control the power of the high frequency band will vary continuously and mimic the power of the original audio to diminish the noise and mitigate the audio distortion In particular the power control procedure is performed in the frequency domain Once the audio is converted to the frequency domain the amplitude values of the corresponding passband will be extracted to control the power of the modulated symbols 10 The schematic diagram is shown in Figure 3 7 According to this diagram the power control takes place in the frequency domain after the modulation The powe
39. er_CreateFcn hObject eventdata handles if ispc amp amp isequal get hObject BackgroundColor end function receiver_Callback hObject eventdata handles function receiver_CreateFcn hObject eventdata handles if ispc amp amp isequal get hObject BackgroundColor end function pushbuttonl_Callback hObject eventdata handles data_input get handles sender string audio_input get handles audname string set handles receiver string set handles ber string 0 set handles trans string 0 set handles dis string 0 vall get handles checkboxl value val2 get handles checkbox2 value if zisempty data_input amp amp isempty audio_input else end trans rate round 33x 1 1624 44100 assignin base ber ber ber s ber l set handles ber string ber s end set handles trans string trans rate if vall 1 amp amp val2 get 0 defaultUicontrolBackgroundColor set hObject BackgroundColor white get 0 defaultUicontrolBackgroundColor set hObject BackgroundColor white data bin num bingen data input N length data bin num sim time 1024 44100 N 11 frameN ceil N 11 if sim time gt str2num get handles edit7 string msgbox strcat simulation time needs gt num2str sim_time error warn else assignin base data_bin_num data_bin_num set param Acoustic ofdm DBPSK FINAL A
40. h utilization of the frequency band strong noise immunity and strong capability to resist frequency selective fading Moreover the theories of acoustic OFDM technology will be described in detail and the power control technique and the combination manner between audios and OFDM signals will be especially elaborated on Chapter 3 describes the systematic theory structure established in the experiments including channel coding modulation demodulation power control compatibility design of audio and OFDM signals and the manner of timing synchronization Chapter 4 elaborates on the simulation system and implementation method based on the design theories Besides the difference between the two systems and possible problems are also analyzed in this chapter Chapter 5 first describes the testing environment of the experiment and then analyzes the testing results of the two experiment systems For the measurement experiment we will mainly focus on the factors that could affect the experimental result Chapter 6 concludes the research results of this thesis and it also makes related evaluation of the experiments mentioned in chapter five In addition some problems which still exist and demand prompt solutions are described Finally the research direction of the problem that I have dealt with in this thesis will be proposed in this chapter Chapter 2 Background of acoustic OFDM 2 1 OFDM OFDM is the abbreviation of Orthogonal Frequenc
41. he same audible quality as the original audio signal The length of the audio frame after the up sampling will be 1024 x 203 207872 After the up sampling a down sampling module with factor 128 will be utilized to decrease the frame length to 207872 128 1624 Figure 4 12 illustrates the procedure of audio re sampling After these operations we know that the frame lengths of the audio signal and the useful data are the same meanwhile after the re sampling the frame interval of the audio signal is unchanged which can be verified by the following expression Niotal X Pup 1624 x 128 0 023 F X nason 44100 x 203 i where N otai is the length of an OFDM symbol nup is the up sampling factor Fs is the sampling frequency and Ndown the down sampling factor 4 1 8 Channel For the simulation system we use a simulation channel which consists of addi tive white Gaussian noise AWGN and some delays Figure 4 13 illustrates the structure of the channel In this figure there are three channels with different delays which are then combined and go through an AWGN channel module Channel A is the channel without delay and it can be used to mimic the real signal which transmits directly to the receiver without the influence of obstacles Channel B is the channel with 57 samples delay and the gain of the channel is set to 0 15 Channel C is the channel with 500 samples delay and the gain of the channel is set to 0 1 For channel B Imp
42. ies of the channel but also enhance the communication 13 14 Method information hel lo ASCII in decimal type 104 101 108 108 111 m ASCII in binary type 01101000 01100101 01101100 01101100 01101111 O ASCII transformation ASCII transformation Store and processing Store and processing Y Y Figure 3 1 Information Transformation reliability Therefore the information will have some redundancy after the channel coding The channel coding method utilized in this experiment is the combination of a convolutional code and interleaving 3 2 1 Convolutional Code A convolutional code is a kind of forward error correction code FEC 7 which has an excellent capability of correcting random errors Figure 3 2 is an example of a convolutional code with a rate of 1 3 and constraint length of 2 y2 y3 Figure 3 2 An example of a convolutional encoder with rate 1 3 and constraint length 2 From the figure above we know that this system has three registers one input and 3 2 Channel Coding 15 three outputs The system will generate three output values from each input data and then shift the state values of the registers to the right side corresponding to the input value The iteration will keep working until all the information bits are encoded The mathematical expression for this system is e yi t u t u t 1 u t 2 e yo t ult 1 u t 2 e y3 t u t u t 2
43. ignal at high frequency band will be multiplied by sine and cosine waves at the carrier frequency 7 2kHz and this will create signals centered on 2 x 7 2kHz 14 4kHz so a low pass filter will be used to eliminate these components Figure 4 10 shows the procedure of down conversion Real Imag to Complex Digital Filter Design Sine Wave1 Figure 4 10 Down conversion 4 1 Simulation System 33 4 1 7 The Signal Compatibility We know that the cyclic prefix is used to avoid the inter symbol interference and the inter carrier interference In this system the last 600 symbols of the data frame with length of 1024 will be used as the cyclic prefix of the original data frame In Matlab we can use a selector block to achieve this step and after that the length of the new data frame becomes 1624 and the frame interval is still 0 023s For the status above the audio signal should also have the same frame length and frame interval to be compatible with the OFDM data frame In this experiment we adopt the method of re sampling which is illustrated in Figure 4 11 Repest Downsample Digital Figure 4 11 The way to re sample the audio signal The original audio signal should be handled first by a low pass filter with the pass band of 5512 5Hz and then each sample of the audio signal should be repeated 203 times to achieve the up sampling step Here we repeat each sample to make sure that after the down sampling we can still get t
44. ignals must be the same However it is not easy to meet this demand The sampling rates of the useful signals are different before and after OFDM mod ulation Meanwhile the power control of the audio data takes place before the modulation and its combination with useful signals takes place after the modula tion Considering this fact we need to re sample the audio signals to ensure the same sampling rate corresponding to the useful signals Figure 3 8 illustrates the way how audio signals are combined with the OFDM symbols ZIN Output Mapping Power control OFDM modulation Figure 3 8 The combination mechanism for audio and OFDM signals audio data As shown in figure 3 8 re sampling consists of two steps up sampling with larger factor and down sampling with smaller factor Since the factor for the up sampling is greater than the factor for the down sampling the audio quality will not be affected from this procedure Finally the overall sampling rate will be in accordance with the modulated useful signals More details about the factor configurations will be described in the following chapters 3 7 Timing Synchronization The role of timing synchronization is to obtain the starting position of the useful signals at the receiver and then demodulate the OFDM signals and recover the transmitted information This thesis mainly adopts coarse synchronization to obtain the start position of the useful signal We add some kind
45. ing and make people react to them However in the world of computer we can not simply transmit a word or a number in their original type The computer can only accept binary information which means that we need to find an approach to make the real life information compatible with the computer Or we can not do any digital signal processing such as the channel coding and modulation in this thesis experiment In computers each letter number or sign has an unique number to represent it and this number could be decimal or binary This is the so called ASCII which stands for American Standard Code of Information Interchange If we want to handle some information in a computer we can just transform the real life information into ASCII with their binary type and then the computer will identify this kind of information Figure 3 1 shows an example of how the information transformation works In my thesis experiment the source information is some kind of text message or URL so for the transmitter 1 need to transform them into their binary type first by using ASCII and then process the binary sequences in the system Also for the receiver we need to revert the information by using ASCII from binary sequences into the type that people can understand 3 2 Channel Coding Channel coding is aiming to achieve the role of error correction by adding in some new supervisory code elements according to certain rules It cannot only match with the statistical propert
46. io signal and the OFDM signal This will bring a lot of noise to the original audio signal and it is unacceptable in our real life Figure 5 5 illustrates the power spectrum of the combination between the audio signal and the power controlled OFDM signal From this figure we can see that the power spectrum of OFDM signal varies over time to mimic the power spectrum of the audio signal As a result the noise caused by the OFDM signal will be weakened or eliminated This is the primary 48 Evaluation 50 60 70 80 Magnitude squared dB 90 100 110 30 20 10 0 10 20 30 Frame 137 Frequency kHz Figure 5 5 The power spectrum of the combined signals with power control feature of acoustic OFDM For figure 5 5 we assume that the average power of the power controlled OFDM signal is around 85dBW Hz Due to that the environment noise for the measurement system is controlled under an acceptable level so for the simulation system we also set a low level environment noise to make sure that the system can work stably and obtain a good enough experiment result We have also tested the system at different noise levels to analyze how the noise can influence the system performance in the following sections 5 2 2 Bit Error Rate In the simulation system one of the factors that affect the system performance is the environment noise Figure 5 6 illustrates the experiment result when the power of the standalone
47. ithout AUDIO S From audio inputFilename allpath original audio f wavread allpath global aud aud audioplayer original audio f end e l i T4 Programming Codes function pushbutton5 Callback hObject eventdata handles global aud aud name get handles audname string if isempty aud name play aud else msgbox please select an audio source error warn end function pushbutton6 Callback hObject eventdata handles global aud aud name get handles audname string if isempty aud name stop aud else msgbox no audio needs to stop error warn end function axesl_CreateFcn hObject eventdata handles axes hObject imshow ISY gifE function trans_Callback hObject eventdata handles function trans_CreateFcn hObject eventdata handles if ispc amp amp isequal get hObject BackgroundColor get 0 defaultUicontrolBackgroundColor set hObject BackgroundColor white end function dis_Callback hObject eventdata handles function dis CreateFcn hObject eventdata handles if ispc 66 isequal get hObject BackgroundColor get 0 defaultUicontrolBackgroundColor set hObject BackgroundColor white end function edit7 Callback hObject eventdata handles function edit7_CreateFcn hObject eventdata handles if ispc amp amp isequal get hObject BackgroundColor get 0 defaultUicontrolBackgroundColor
48. lementation emplitede original sampling wa after up sampling ann nn ERA A na 7 ANN FR ate atann A O M HDD DU EE WO GU EE AA otan eave antann nonm DAR aaa tn ANA TO MN RR DN TC TC TRUST PURI RR KE PE AE KA KE MR A ARR AE KE KE EE GO EE FO EFO KE ME pa eee TC TR TP CEU aa nt atann LOC RR TUL t n mmplilnde after down sampling Ti te cei Figure 4 12 The processing of audio re sampling 4 1 Simulation System 35 O un AWGN n B Out1 57 Add z AWGN Gain Delay Channel h 500 E Gan Delay Figure 4 13 The channel structure of the simulation system and C we set smaller gains to mimic the situation that the reflected signals will be weakened when transmitting in the real channel After combining these three channels an additional white Gaussian noise model will be used to simulate the existing noise in the real life and the signal to noise ratio in the model will be set to 15dB 4 1 9 GUI The GUI of this system is built by the GUI toolkit in Matlab and it has the following functionalities e audio selection e transmission of standalone OFDM signals e transmission of acoustic OFDM signals without power control e transmission of acoustic OFDM signals with power control e switch between audible mode and spectrum view mode e play and stop the audio only e changeable simulation time e data rate bit error rate and audio deviation illustra
49. lity can be mitigated by using power control of OFDM signals 2 The idea behind acoustic OFDM is showed in Figure 1 1 Before acoustic OFDM several approaches have been proposed to derive useful data from the audio signals such as echo hiding 4 phase coding 5 and spread spectrum 6 How ever these methods can only achieve a very low data rate Thanks to this technology some short information such as a URL or media information advertising can all be ef fectively transmitted to the terminal end like a mobile phone through the manner of audio such as music 1 1 Problem Formulation In common acoustic OFDM technology the sender could directly broadcast the OFDM signals through an audio generator such as a loud speaker This kind of 1 2 Introduction Power spectrum Power spectrum Audio signal Eliminated part OFDM signal frequency frequency Low band High band High band Power spectrum Power controlled OFDM frequency Figure 1 1 The algorithm of acoustic OFDM sound is usually screaming noise similar to white Gaussian noise which would im pact people s normal life once used in the real world However if the OFDM signal is embedded in an audio signal and transmitted in the manner of power control no noise will be produced and the audio quality will not be affected much More importantly the useful information could be transmitted effectively in this way There are mainly three issues discussed in this pape
50. much Figure 2 7 illustrates what the frequency selective fading is and how it affects the channel Although the power of some subcarriers is restricted due to the influence of frequency selective fading the other subcarriers can still work with normal power This can achieve a low bit error rate for the data transmission 2 1 OFDM 9 Multipath Fading fading Vp VAN ai Figure 2 7 The effect of frequency selective fading on multiple carriers frequency 2 1 3 System Model A fully formed OFDM system should consists of the parallel serial conversion modulation demodulation and time frequency conversion Figure 2 8 illustrates the theoretical structure of the transmitter for an OFDM system Constellation mapping Eh Jag Bud ae F sa s n I Serial to parallel Figure 2 8 The transmitter for an OFDM system As is showed in the figure after one serial to parallel conversion the signals will be conveyed through several subchannels By using for example M PSK or QAM modulation the information bits will be mapped into a constellation diagram This will show the information bits in the complex plane which is also a conversion from time domain to frequency domain Then the modulated symbols are processed by IFFT to form the OFDM symbols Moreover the role played by IFFT here is not only to convert signals from frequency domain to time domain but also to achieve the orthog
51. n can be em bedded in audio and then transmitted so that the receiver can obtain the required information through certain demodulation mechanisms without severely affecting the audio quality This thesis mainly discusses how to embed and transmit information in audio by making use of acoustic OFDM Based on the theoretical systematic structure it also designs a simulation system and a measurement system respectively In these two systems channel coding manners of modulation and demodulation timing synchronization and parameters of the functional components are configured in the most reasonable way in order to achieve relatively strong stability and robustness of the system Moreover power control and the compatibility between audio and OFDM signals are also explained and analyzed in this thesis Based on the experimental results the author analyzes the performance of the system and the factors that affect the performance of the system such as the type of audio distance between transmitter and receiver audio output level and so on According to this analysis it is proved that the simulation system can work steadily in any audio of wav format and transmit information correctly However due to the hardware limitations of the receiver and sender devices the measurement system is unstable to a certain degree Finally this thesis draws conclusions of the research results and points out unsolved problems in the experiments Eventually some expect
52. nce for speech But for other types of audio the system will become unstable and inefficient This is because for the music there are a lot of high frequency components which need to be used to control the power of the OFDM signal However for the speech most of the sound components are located in low frequency band and the power spectrum in high frequency band approximates zero So to some degree the OFDM signal embedded in the speech has no power control but decreases its power spectrum under the threshold of audibility 5 3 4 SDR Another interesting experiment for the measurement system is about the relation ship between the signal bit error rate and SDR which stands for signal to distortion ratio We know that when we increase the volume of the loud speaker or move the speaker and microphone closer the power of the signal for both the audio and the OFDM parts will be increased It also means that the OFDM signal power is proportional to the totally received power When we increase the power of the OFDM signal the audio distortion will also increase and the bit error rate of the signal will have some relative changes Figure 5 13 illustrates this relationship for the measurement system without power control and figure 5 14 is for power controlled measurement system From these two figures we know that speech has higher SDR and achieves lower bit error rate than other types of audio when the powers of the OFDM signal are equal no matter it is
53. non power controlled system In the figure above the curve with circles represents the audio distortion of the combined signals without power control and the curve with triangles represents the audio distortion of the combined signals with power control From the figure we know that the audio distortion of the power controlled signal is lower than the one without power control and the average gain during 7 seconds is around 0 0032 The audio distortion of the former one is out of the range of human s hearing This is also a reason why we can here little noise from the audio signal which is combined with the power controlled OFDM signal For the previous experiment the power of the standalone OFDM signal is set to 85dBW Hz which is a constant value But we also need to know how the signal power affects the audio quality Figure 5 8 illustrates the relationship between the audio distortion and the power of the acoustic OFDM signal for different kinds of audio In this figure we assume that the audio distortion for all types of audio is an average value when the audio is played 5 seconds From this figure we apparently find that the speech can achieve lower audio distortion than any other kind of audio when the power of the acoustic OFDM sig nal rises That is because for speech most of its frequency components concentrate at the low frequency band which is normally less than 5kHz The power controlled signal is arranged in the high frequency band wi
54. ompatibility between audio and OFDM signals are also explained and analyzed in this thesis Based on the experimental results the author analyzes the performance of the system and the factors that affect the performance of the system such as the type of audio distance between transmitter and receiver audio output level and so on According to this analysis it is proved that the simulation system can work steadily in any audio of wav format and transmit information correctly However due to the hardware limitations of the receiver and sender devices the measurement system is unstable to a certain degree Finally this thesis draws conclusions of the research results and points out unsolved problems in the experiments Eventually some expectations for this research orientation are stated and relevant suggestions are proposed Nyckelord Keywords acoustic acoustic OFDM power control embed Abstract The OFDM technology has been extensively used in many radio communication technologies For example OFDM is the core technology applied in WiFi WiMAX and LTE Its main advantages include high bandwidth utilization strong noise im munity and the capability to resist frequency selective fading However OFDM technology is not only applied in the field of radio communication but has also been developed greatly in acoustic communication namely the so called acoustic OFDM Thanks to the acoustic OFDM technology the informatio
55. on sequently in order to promote the channel utilization rate and anti interference ca pability digital signals will be combined with some kind of carrier signals through a certain modulation approach which can also facilitate Digital to Analog Con version For acoustic OFDM the main modulation approach is M PSK In this experiment we use the approach of differential binary phase shift keying DBPSK DBPSK transmits data in the way of shifting the carrier phases namely using the phase shift to modulate carriers and finally send out binary information For ex ample if 0 and 1 respectively represent adding 0 and 180 to the current phase of the carrier then the output waveform of DBPSK can be shown as Figure 3 4 in case that the input signals are 1001 and the phase reference is 0 One advantage of utilizing DBPSK is that the bit error rate caused by the phase ambiguity can be offset 9 3 4 Pilot In telecommunications a pilot is a kind of signal which is transmitted with the useful signal for supervisory control or equalization purposes There are two common distributions for pilots which are block type and comb type For block type the pilots will be distributed in different time interval and for comb type the 3 4 Pilot 17 time 0 T 2T 3T an ST Figure 3 4 The timing diagram for DBPSK pilots will be deployed at different frequencies or subcarriers Figure 3 5 illustrates the pilots arrangement
56. onality among subcarriers Complex signals that are derived from IFFT will be divided into real and imaginary components before going through a digital to analog conversion Then these analog signals will be used to modulate cosine and sine waves at certain frequency Finally the data of the real part and imaginary part will be combined and sent out from the radiation device Referring to Figure 2 8 the mathematical expression of the low pass OFDM signals 10 Background of acoustic OFDM can be written as 2 1 s t X X e 7 T 0 txT 0 gt Il where X is the modulated symbol N is the number of subcarriers T is the period of an OFDM symbol Here the bandwidth of each subcarrier should equal to 1 T which can guarantee the orthogonality between every two subcarriers This property can be indicated as the following expression sin T T 1 erty es2mkat T q er k ka t T di 0 de 0 In real systems problems such as inter carrier interference or inter symbol in terference should all be taken into consideration so cyclic prefix should be added in to resist these influences Under the condition of utilizing cyclic prefix the expression of a low pass OFDM signal can be written as the following s t Din Kre rkt T _T lt t lt T where T is the length of the guard interval in which the cyclic prefix is transmitted At the receiver all the signals will be converted down to baseband and then the time domain sign
57. other locations In this way the possible bit errors can be fully dispersed thus greatly improving the capability of the error correction 3 2 3 Viterbi Decoding From the description in the previous subsection we know that the convolutional encoder can be considered as a finite state machine It has 2 states where n is the number of registers in the system so we can also represent the output of the system on a Trellis diagram Trellis diagram is used to illustrate how the states of the coding system change in time It does not only show the instantaneous transitions but also gives the most probable system outputs and the state transitions After handled by the convolutional encoder the coded signals can be decoded by using the Viterbi algorithm Viterbi algorithm is a kind of decoder that utilizes the maximum likelihood decoding method Based on the received information the Viterbi decoder can find a route which has the most likely information sequence corresponding to the original transmitted information on the Trellis diagram This route is also called Viterbi path on which a cluster of the information bits is exactly the one we should obtain through the decoding process 7 8 3 3 Modulation Method The digital signal has a very poor capability of anti interference and anti noise when it propagates in the channel What s more under normal conditions the in formation which can be transmitted in the channel should be analog signals C
58. power controlled or non power controlled Piano guitar and song have the similar feature that the bit error rate will decrease sharply when the SDR increases 55 5 3 Measurement Result Figure 5 12 The comparison of the bit error rate for different type of audio for power controlled OFDM signal Y i 1 H 1 i A 1 i i l 1 A i 1 i an v 1 i i 1 i ae i i i 1 The relationship between BER and SDR for different types of au dio without power control Figure 5 13 56 Evaluation n an p Bit Error Rate Signal to Distortion Ratio dB Figure 5 14 The relationship between BER and SDR for different types of audio with power control The difference between figure 5 13 and figure 5 14 is that for the system without power control the bit error rate is much lower than the power controlled system when the power of the non power controlled OFDM signal is similar with the average power of the power controlled OFDM signal 5 3 5 Audio Distortion For the measurement system we also need to calculate the audio distortion for different types of audio We choose different types of audio files to test the audio distortion over time and assume the average power of the power controlled OFDM signal is around 85dBW Hz Figure 5 15 shows the audio distortion for different types of audio From this figure we know that the audio of speech can achieve the lowest disto
59. r One is how to embed OFDM information into an audio signal and to transmit information without affecting the audio quality too much Second how to control the power of OFDM information and realize the compatibility between audio and OFDM signals Third how to make use of MATLAB simulation results to study the factors that affect the stabil ity and robustness of the acoustic OFDM system as well as to which degree those different factors affect the audio quality and the performance of the communication system 1 2 Thesis Objective This thesis project aims to realize the following points 1 Complete detailed study and analysis of acoustic OFDM systems 2 Design the acoustic OFDM simulation system and measurement system 3 Study the factors such as audio type audio output level and distance between microphone and loudspeaker which affect the robustness of the system by analyzing the operating results of the designed system 1 3 Thesis Outline Chapter 1 gives an overview of the application of OFDM technology in the field of acoustics and how to make use of acoustic OFDM to transmit information In addition this chapter also proposes the overall research direction of this thesis 1 3 Thesis Outline 3 and the required experiments based on software Finally the objective of this thesis is pointed out clearly Chapter 2 makes a further introduction of the theories of OFDM technology and how the OFDM technology realizes relatively hig
60. r elektronisk version http www comnsys isy liu se http www ep liu se Titel Title Embedding data in an audio signal using acoustic OFDM F rfattare Shuai Wang Author Sammanfattning Abstract The OFDM technology has been extensively used in many radio communication echnologies For example OFDM is the core technology applied in WiFi WiMAX and LTE Its main advantages include high bandwidth utilization strong noise im munity and the capability to resist frequency selective fading However OFDM echnology is not only applied in the field of radio communication but has also been developed greatly in acoustic communication namely the so called acoustic OFDM Thanks to the acoustic OFDM technology the information can be em bedded in audio and then transmitted so that the receiver can obtain the required information through certain demodulation mechanisms without severely affecting he audio quality This thesis mainly discusses how to embed and transmit information in audio by making use of acoustic OFDM Based on the theoretical systematic structure it also designs a simulation system and a measurement system respectively In these wo systems channel coding manners of modulation and demodulation timing synchronization and parameters of the functional components are configured in the most reasonable way in order to achieve relatively strong stability and robustness of the system Moreover power control and the c
61. r in the corresponding frequency band of the audio signal will be derived to control the power of useful signals Then the audio signal in the low frequency band is combined with the useful symbols which are OFDM modulated before they are emitted One problem of power control is that the power of the useful signals will not be zero while the audio power at some frequencies will be zero Hence we need to set a threshold value for the audio control module and this will prevent the power of the useful signals from becoming zero and avoid the occurrence of higher bit error rate 19 3 5 Power Control mais an Low pass filter Re sample time domain Transform to Derive values in scope 6400 8000 HZ frequency domain FFT Amplitude OFDM modulation Modulation DBPSK Useful data control Power control Figure 3 7 The mechanism of Power Control 20 Method 3 6 Compatibility between audio and useful data The low frequency band of the audio signal needs to be combined with the power controlled OFDM symbols before being transmitted through the air Therefore the two signals can be mixed together only if they share some common character istics Since the frame mode is used for the signal transmission in the experiment the audio signal and the useful symbols should be the same in terms of the frame length and the frame interval It also means that the sampling frequencies of these two s
62. re are showed in table 4 2 below 4 1 Simulation System 25 Software MATLAB Version 7 11 0 584 R2010b Laptop ACER Aspire 5738ZG Loudspeaker LOGITECH S 220 2 1 Speaker Microphone Cosonic MK 208 Frequency range 100 16KHz Sensitivity 58dB Table 4 2 The Configuration of Software and Hardware What needs to be declared is that the sampling frequency of the audio in put output hardware driver should be configured correctly in the experiment The sampling frequency used in Matlab is 44100 Hz in this experiment So the sam pling frequency of the audio input output hardware driver could be set as 192 kHz Due to this configuration no data will be lost during the transmission through the hardware such as the microphone and loudspeaker Another configuration is that for the 2 1 loudspeaker the bass value should be set to zero and the functionality of the loudspeaker will be only amplifying the sound level of the audio In this experiment we only test the mono audio signals and the two loudspeakers that we used in the testing will transmit the same audio signals synchronously which can enhance the sound pressure and increase the stability of the system 4 1 Simulation System The simulation system consists of three components namely M file Simulink module and GUI M file with the suffix m consists of the programming code It is the centrum of this system and all other components should be controlled by
63. rectly when the channel distance is up to 1 2 meters So the whole experiment could be considered as efficient and successful 6 2 Future Work More work should be done in the future for the research on this topic As I said in the previous chapters we simplified several configurations of the system in order to decrease the complexity of the system and enhance its operability It means that we omitted or discarded some functionalities or steps which are not so important in the system So for the future work the following work can be done to improve the performance of the system 1 Support different formats of the audio signal such as mp3 acc etc 2 Support more types of files to be transmitted in the system such as PDF image etc 3 Apply more than one modulation methods such as BPSK or QPSK for which channel estimation will be required as well 59 60 Conclusions and future work 4 Utilize a more precise timing synchronization method to derive the useful data at the receiver 5 Create a more beautiful and easy to use GUI for people who are not studying the relevant area 6 Improve the structure of the system to provide better functionalities with high stability I believe that the system will work better after these improvements I also trust that more potential research methods on this topic will be created and introduced and this technology will be utilized in our daily life in the future Bibliography 1 2
64. requency domain In this way the audio quality can still persist in a good situation and the useful data will also be transmitted correctly 2 2 2 Compatibility Compatibility means that the audio signals are compatible with the modulated useful information On the one hand the sampling rate of the useful signals will increase after the OFDM modulation On the other hand the audio will not only be used to control the power of the useful signals but also integrated with the OFDM signals Therefore we need to consider how to ensure the same sampling rates of the two signals during integration or a series of problems will arise As to this issue we use the method of re sampling Since the sampling rate of useful information will increase after the OFDM modulation we will up sample the audio signals to increase more sampling points in a unit time This method will realize the compatibility of these two kinds of signals during their integration and it will not bring any influence on the quality of the original audio The specific implementation method will be elaborated on in the following chapters Chapter 3 Method An acoustic OFDM system is composed of several subsystems each of which plays its special role This chapter will explain the theoretical structure and implemen tation methods of these subsystems in the experiment 3 1 Information Transformation Information is some kind of meaningful signs that can be understood by human be
65. rtion compared with other types of audio Piano guitar and song have the similar audio distortion For the measurement system we still need to inspect the relationship between the audio distortion and the power of useful signal Figure 5 16 illustrates this relationship for different kinds of audio Here we still assume that the audio distortion is an average value by playing different kinds of audio for 5 seconds and the average power of the power controlled OFDM signal is around 85dBW Hz From this figure we can get the same conclusion as the simulation system that the speech will achieve much lower audio distortion than other kinds of audio when the power of OFDM signal increases But we can also find that all the audio distortions in the measurement system is much larger than that in the simulation 57 5 3 Measurement Result Figure 5 15 The audio distortion for different kinds of audio for power controlled OFDM signal Figure 5 16 The relationship between the audio distortion and the power of the power controlled OFDM signal for measurement system 58 Evaluation system The main reason for this situation is that the real life system will be restrained by the hardware and the environment Therefore it can not reach the ideal level as in the simulation system 5 3 6 Limitations For the whole experiment these two systems can fulfill the requirements and achieve a good performance at most of the time However due to
66. ses oso 200000 70 Chapter 1 Introduction In the development of today s communication technology the widely used tech niques of information exchange such as WiFi WiMAX and LTE are achieved through radio communication technologies These techniques share one common feature by using Orthogonal Frequency Division Multiplexing OFDM As a core technology in many communication standards OFDM has been extensively ap plied mainly due to its high bandwidth utilization rate strong noise immunity and the capability to resist frequency selective fading 1 Also OFDM technology cannot only be applied in radio communication technology but in acoustic signal transmission namely the so called acoustic OFDM 2 Making good use of the advantages of the OFDM technology acoustic OFDM can modulate the useful in formation which is then transmitted in air or water with the help of sound sending devices such as loudspeakers In this way the receiver such as a microphone could obtain such useful information through some kind of demodulation mechanisms once they received the sound Acoustic OFDM is mostly applied in underwater information exchange 3 for example in the short distance information exchange between different hulls However another more updated application of acoustic OFDM is embedding data in different audio to transmit the information The high frequency band of the audio is partly replaced by OFDM signals and the impact on the audio qua
67. so the value of its spectrum is nearly constant At the central frequency there is attenuation due to the existence of a DC subcarrier Figure 5 2 and Figure 5 3 illustrates the power spectrum of the OFDM signals which are shifted from low frequency band to high frequency band with the central frequency of 7 2 kHz This is still a standalone OFDM signal which has no power control In the experiment one of the functionalities for our system is transmitting this kind of signal without the audio signal 45 46 Evaluation Figure 5 1 The power spectrum of the OFDM signal in low pass band Figure 5 2 The power spectrum of the OFDM signal in high pass band 1 5 2 Simulation Result 47 65 70 75 80 85 90 Magnitude squared dB 95 100 105 110 5 55 6 6 5 7 7 5 8 8 5 9 95 Frame 137 Frequency kHz Figure 5 3 The power spectrum of the OFDM signal in high pass band 2 Figure 5 4 is the combination of the audio signal with type of piano and the OFDM signal without power control 50 7 eudio signal 60 7 rd 2a n 7 OFDM symbols x El y 5 5 8 80 d 3 5 90 100 110 30 20 10 0 10 20 30 Frame 137 Frequency kHz Figure 5 4 The power spectrum of the combined signals without power control So as we can see in the figure it is just a simple addition operation between the aud
68. stance the curve with triangles has a higher bit error rate than the curve with circles which is in accordance with the previous simulation result For the acoustic OFDM signal the bit error rate of the measurement system can reach 0 1 when the distance is 1 7 meters In this condition a lot of error bits occur and the stability of the system is dropping down rapidly For this experiment we can conclude that the standalone OFDM signal can be transmitted correctly in the measurement system when the channel distance is less than 2 5 meters and the acoustic OFDM signal can be transmitted correctly in this system when the distance is less than 5 3 Measurement Result 53 0 6 0 5 0 4 Bit Error Rate 0 3 0 2 0 1 eLo 0 5 1 15 2 25 3 Channel Distance meter Figure 5 10 The transmission capability of the measurement system with varied chan nel distance 1 2 meters 5 3 3 Sound Level Another factor we need to consider in the measurement system is the sound level which can influence the system performance For this testing we need to control the volume of the loudspeaker to observe the bit error rate of the system We also assume that the channel distance is 0 5 meter the power of the standalone OFDM signal is 85dBW Hz and the average power of power controlled OFDM signal is around 85dBW Hz Figure 5 11 illustrates the relationship of the bit error rate of the measurement system and the sound level For the measur
69. th OFDM signal sa Delayed th OFDM signal 4 Delayed th OFDM signal Figure 2 2 ISI for OFDM signals without guard interval Considering the fact that ISI will occur frequently without a protection mech anism we will set a guard interval GI 1 between adjacent OFDM signals This 2 1 OFDM 7 guard interval should be longer than the time period of the channel pulse response to ensure that the inter symbol interference can be eliminated Figure 2 3 shows the GI distribution in OFDM signals FDM Symbo ae Fan date date time T Symbol o ee che ooo t Ty T guard interval data interval Figure 2 3 The distribution of guard intervals in OFDM symbols For the simple case a guard interval will be filled with zero values namely silent guard interval which means that no information is conveyed in it The inter symbol interference problem can be truly resisted in this way However due to the influence of the latency caused by multipath propagation there will be several different versions for each signal at the receiver Consequently when the same OFDM signals with different latencies are received the phase difference between adjacent subcarriers will not maintain an integer cycle This will cause the loss of orthogonality among these subcarriers and finally cause the inter carrier interference in the process of OFDM demodulation at the receiver Figure 2 4 demonstrates the reason that c
70. th the central frequency 7 2kHz So the power of the power controlled OFDM signal will be almost the same at different subcarriers and the value of the power is very low which will lead to a 5 3 Measurement Result 51 Audio distortion Power of acoustic OFDM signal dBW Hz Figure 5 8 The relationship between the audio distortion and the power of the acoustic OFDM signal for simulation system lower audio distortion For other types of audio their frequency components will distribute both at low frequency band and high frequency band Consequently the power of the power controlled acoustic OFDM signal will vary continuously with the audio signal Since the acoustic OFDM signal can not mimic the original audio signal precisely from time to time so the audio distortion of piano guitar and song will rise to a higher level than that of speech 5 3 Measurement Result The testing of the measurement system is a very important step for our experi ment In this step we can explore how our theoretical designing can be used in a real life system and what is the difference between the simulation system and the measurement system Figure 5 9 illustrates the scenario of our measurement system In this room there are several experimental devices L represents a laptop which controls the whole system S represents a loudspeaker which will be used to play sounds M represents a microphone which will be used to collect sounds There is no obsta
71. tion 2 22 2 Como nn 1 2 Thesis Objective ma ia g 2 eae een a E a a A 153 Thesis Outlines a won won SRA na Rae EE hae ae A a 2 Background of acoustic OFDM AT HOR DMS oc dit 2 Ban ene are ge Beh a 21 1 Orthogonality ma 2 2a 2 u aoe eee Baha 2 1 2 Advantages of OFDM o o soo 2 1 3 System Model o nn 2 27 Acoustic OFDM sa a a e aa e a E EE A vi s t a A 2 2 1 Rower Control 205 td a sen ana ee 222 Gompatibility edit Ss a an a i ai li ware 3 Method 3 1 Information Transformation 3 2 Channel Coding 2 2 oc fr ge an a dak ay a kap ke may S eat 3 2 1 Convolutional Code ooo 3 2 2 1 Anterleaving x sen a pant son peny ae sye kc oe n aha GOS 3 2 3 Viterbi Decoding on nn 3 3 Modulation Method 2 2 on nn 34 PUOVA 4228 64550 8532 A a EEG de ng 3 5 Power Control tu Me a ke ee eS SERRE es 3 6 Compatibility between audio and useful data 3 7 Timing Synchronization e se ss oss 4 Implementation 4 1 Simulation Systems e see ai ab each 4 1 1 Information SOUrce e en onon 4 1 2 Channel Coding s a a wo cetacea dan dp va sapara f A f f we 4 13 DBPSK Modulation 2 2 2 2 2 nommen 4 1 4 Power Control man nei ee wu AA 4 1 5 OFDM Modulation 2 22cm 4 1 6 Frequency Conversion 2 2 22mm 4 1 7 The Signal Compatibility xi xii Contents AAS Channel fic
72. tion e changeable data source The GUI module used in this experiment is shown in Figure 4 14 and more details about how to operate this system can be found in Appendix A 4 1 10 Receiver The mechanism of the receiver is similar to the transmitter due to that each block in the receiver is a conversion of the corresponding block in the transmitter The only difference in the receiver is that the channel decoding is achieved by Viterbi decoding 36 Implementation subcarriers 33 4 pilots bandwidth 6400 8000Hz symbol interval 1024 CP 600 Modulation Method DBPSK channel coding convolutional coding interleaving timing synchronization coarse synchronization parameters ________________________ Simulation Result transmission rate 0 bit s simulation time 5 ena Zr bit error rate 0 embedded in audio power control audio distortion 0 mode switch gt audio Message to be transferred Message received konen Department of Electrical Engineering Figure 4 14 The GUI for the simulation system 4 2 Measurement System 37 4 2 Measurement System The measurement system consists of four components which are M file Simulink model audio input output devices and GUI The Simulink model is responsible for the signal processing the audio input output devices substitute the simulation channel and the GUI has the same functionalities
73. ume that there are two signals x t and y t which are orthogonal and then they will meet the following equation in the time domain T 5 0 v0 0 0 where denotes complex conjugate operator T is the time period for one OFDM symbol In frequency domain orthogonality means that the value of each subcarrier is exactly zero at the central frequency of its neighboring subcarriers which will not result in inter carrier interference ICT 2 1 2 Advantages of OFDM High Utilization Ratio of Spectrum Due to its advantages the OFDM technology has been widely applied in the field of radio communications According to the previous introduction it is known that the overlaps of adjacent subcarriers can raise the spectrum utilization ratio In the following we will describe two other merits of OFDM High Anti Interference Capability An OFDM system adopts multiple carriers mode to transmit data For each carrier the number of bits can be adjusted according to the signal to noise ratio SNR of each carrier i e for the carriers with high SNR more data bits will be assigned So the anti noise capability of OFDM system is enhanced Due to the reflection phenomenon the transmitted signal is affected by multipath propagation and the receiver will obtain several versions of the useful data with different latency Therefore inter symbol interference can easily occur if there is no protection for this case Figure 2 2 depicts this situation l
74. ustic ofdm DBPSK FINAL AUDIO From audio inputFilename allpath original_audio f wavread allpath global aud aud audioplayer original_audio f end function audname Callback hObject eventdata handles function audname_CreateFcn hObject eventdata handles if ispc amp amp isequal get hObject BackgroundColor get 0 defaultUicontrolBackgroundColor set hObject BackgroundColor white end 11_Callback hObject handles function eventdata function radiobutton2 Callback hObject eventdata handles function 12_Callback hObject eventdata handles function pcl_SelectionChangeFcn hObject eventdata handles function slider3_Callback hObject eventdata handles function slider3_CreateFcn hObject eventdata handles if isequal get hObject BackgroundColor get 0 defaultUicontrolBackgroundColor set hObject BackgroundColor 9 9 9 end function checkbox7_Callback hObject val get hObject value if val 0 eventdata handles set_param Acoustic_ofdm_DBPSK_FINAL AUDIO auon sw 1 set_param Acoustic_ofdm_DBPSK_FINAL AUDIO scope OpenScopeAtSimStart set hObject string mode switch else set param Acoustic ofdm DBPSK FINAL audio AUDIO auon sw 0 set_param Acoustic_ofdm_DBPSK_FINAL AUDIO scope OpenScopeAtSimStart set hObject string mode switch gt end spectrum view
75. ut function pushbuttonl_Callback hObject eventdata handles data input get handles sender string audio input get handles audname string set handles receiver string vall get handles checkboxl value val2 get handles checkbox2 value if zisempty data_input amp amp isempty audio_input data bingen data_input N length data assignin base N N assignin base data data frameN ceil N 11 set param Acoustic ofdm DBPSK FINAL without AUDIO S up sl N num2stn set param Acoustic ofdm DBPSK FINAL without AUDIO S s2 N num2str fy sim Acoustic ofdm DBPSK FINAL without AUDIO S 0 str2num get handles edit7 string set hObject enable on assignin base tran tran trans_rate round 33 1 1624 44100 set handles trans string trans rate if vall au aul au2 frameN ameN 72 Programming Codes audiv mean au 2 set handles dis string audiv end else msgbox please select audio or input the data that you want to send error warn return end function sender_Callback hObject eventdata handles function sender_CreateFcn hObject eventdata handles if ispc amp amp isequal get hObject BackgroundColor get 0 defaultUicontrolBackgroundColor set hObject BackgroundColor white end function pushbutton2_Callback hObject eventdata handles close function receiver_Callback hObje
76. ving have been used in the experiment namely matrix interleaving and general block interleaving which can be illustrated in Figure 4 4 We use two interleavers to enhance the error correction capability of the channel coding and improve the stability of the system The matrix interleaving method will read in data row wise 28 Implementation and read out data column wise For the block interleaving all the rows of the matrix will be rearranged and each row will be located at a different position The frame length and frame interval will be unchanged after the interleaving General Block Interleaver Matrix Interleaver Matrix General Block interleaver Interleaver Figure 4 4 The structure of interleaving 4 1 3 DBPSK Modulation As we described in the previous subsection after the channel coding the useful data in one frame is composed of 33 samples In the system one sample represents one bit information During the DBPSK modulation these 33 information bits will be converted into 33 complex symbols so the rate is 1 bit symbol What is worth noting is that the signals are transformed from time domain into frequency domain after the DBPSK modulation The frame length and the frame interval are also unchanged after the modulation 4 1 4 Power Control The power control module is divided into two parts the first part is used to extract the spectrum information from the audio signal and the second part is used to control
77. y Division Multiplexing It is a special case of Frequency Division Multiplexing FDM for which the carriers are orthogonal OFDM uses a lot of orthogonal subcarriers to carry and transmit data The data will be divided into several parallel channels or streams and each subcarrier will occupy one channel For each subcarrier the data will be modulated with a conventional modulation scheme such as BPSK or PSK with high orders at a low symbol rate Before introducing OFDM we need to clarify the difference between FDM and OFDM in terms of frequency spectrum distribution Figure 2 1 illustrates the spectrum distributions of FDM and OFDM respectively FDM OFDM Saved Spectrum s frequency Figure 2 1 The spectrum utilization for FDM and OFDM From Figure 2 1 we know that for the FDM modulation method the signals will be modulated and sent out in different subcarriers at the same time In addition there is a protection interval in the frequency domain between every two subcarriers to resist inter carrier interference ICI 1 Therefore the utilization ratio of the spectrum is relatively low For OFDM every two subcarriers will overlap with each other which can save a lot of spectrum to transmit more data 6 Background of acoustic OFDM 2 1 1 Orthogonality Orthogonality is a key point for the OFDM technology In every OFDM signal period the neighboring subcarriers are orthogonal 1 To understand the concept of orthogonality we ass
78. ystem Signal To channel decoding dapsk demod 1 Workspace Digital Filter Design remove pilot Signal From Workspace ofdm demod1 Figure 4 16 The receiver of the measurement system 40 Implementation Figure 4 17 The loudspeaker used in the measurement system Figure 4 18 The microphone used in the measurement system 4 2 Measurement System 41 Figure 4 19 The whole measurement system Bark code in time domain Autocorrelation of Bark code 0 10 0 2 8 3 04 3 6 E 3 06 Za 0 8 2 A 0 n n n A n 312 314 316 318 32 312 314 316 Time offset 0 ma Time ms Time offset 0 mo Time ms Figure 4 20 Bark code and its autocorrelation 318 42 Implementation autocorrelator to derive the starting position of Bark code Since we also need to combine the audio signal with the OFDM signal synchronously so the starting point of Bark code can also be consider as the starting position of the OFDM signal This is the way how the timing synchronization works In this experiment we use barker code generator model in simulink to generate a Bark code with code length 2 at the transmitter and at the receiver we determine the starting point of the OFDM signal by using an autocorrelator model A peak value will appear at the beginning of the autocorrelated signal Figure 4 21 can illustrate this procedure To determine whether it is the peak of Bark code we utilize a Received signal in time domain

Embedding data in an audio signal, using acoustic OFDM

Contents

Download Pdf Manuals

Related Search

Related Contents