Research On Sound Source Localization Algorithm Based On Binaural Signals In Reverberant Environment

Binaural sound source localization is involved in human-computer interaction,video conferencing and voice enhancement and other important technologies.In the engineering application scenario,especially for closed environment,the presence of reverberation causes sound source localization performance degradation.Therefore,it is of great theoretical significance and engineering value to research on the technology of binaural sound source localization in reverberation environment.Firstly,the theoretical bases of binaural sound source localization algorithm are introduced.The principles of binaural sound source localization are studied deeply,thus,the binaural signals model is analyzed and the overall framework of the system is designed.Secondly,the thesis studies the relevant principles of the common time delay estimation algorithms.For the problem of the positioning error caused by the difference of the reverberation to the different frequencies according to the same processing of the sound source signal,generalized cross correlation time delay estimation algorithm based on frequency division is proposed.The binaural signal is divided into sub-band signals by Gammatone filter.Each sub-band signal is subjected to cepstrum prefiltering de-reverberation and then transformed into time domain.The summation operation of each sub-band’s cross correlation value obtained by cross correlation operation on each sub-band of left and right signals is carried on,and the maximum cross correlation value corresponds to the interaural time difference.Experiments show that,the algorithm effectively reduces the positioning errors caused by reverberation in multi-frequency sound source signals,which makes de-reverberation processing adequate for various frequency component of the speech signals.Then,the azimuth estimation method is researched deeply.For the problem that the signal will lose the phase information in the cepstrum domain,resulting in interaural time difference information loss,thereby,a sound source localization algorithm based on a cepstral binaural room impulse response binaural cross-correlation(CEP-BRIR-CC)is proposed.The algorithm is based on subtracting the reverberation component from the BRIR,and the estimated time domain impulse response is derived from the cepstral BRIR inverse transformation.Then by performing cross-correlation operation with the database head related impulse response,the maximum cross-correlation value corresponds to the position corresponding to the estimated location of the sound source.By comparing the similarity of the two pulses,the algorithm avoids the calculation of the interaural time difference.Experiments show that the algorithm effectively helps improving the positioning accuracy.Finally,binaural sound source localization system is constructed,experimentally tested and analyzed.Experimental results indicate that the CEP-BRIR-CC algorithm has better positioning accuracy than other sound source localization algorithms based on time delays estimation.Experimental results are in agreement with the theoretical calculation trend.When the reverberation time is 0.3s and 0.5s,the algorithm’s accurate estimation rates are 89.9% and 87.7%,respectively.It can effectively solve the influence of reverberation on sound source localization in the engineering environment.