Research On Sound Source Localization Technology Based On Generalized Cross-power Spectrum Phase And Device Realization

With the further development of the speech communication technology,in terms of receiving and processing information,we have higher and higher requirements for voice,the speech recognition,speech enhancement and sound source localization have become one of the research hotspots,and sound source localization is the premise and foundation for the speech recognition and speech enhancement.In this thesis,In order to improve sound source localization effects in indoor environments with complex noise and serious reverberation,it makes the research work of the following aspects:(1)The research background and purpose of the sound source localization technology and the domestic and foreign research status are introduced.(2)Some basic concepts and formulas about the sound theory are studied,and the far field model、near field model and the signal model of microphone array are discussed.At the same time,the selected microphone has been tested and calibrated.The microphone array structure is studied to select a right structure,and several acoustic localization algorithms are compared.(3)The formation and characteristic of reverberation are studied,and the definition,property and computing method of cepstrum are analyzed.Then,a lot of parameters of the filter in the cepstral domain are simulated to determine the best value.(4)The time delay estimation algorithm based on cross-power spectrum phase is studied and improved,and then the generalized cross-power spectrum phase is proposed.The digital filter is designed by using the window function method,and a variety of weighted functions in matlab are simulated.This thesis also puts forward to reduce the peak search range、eliminating zero pseudo peak and parabolic interpolation,and derives a formula for the geometric location algorithm.(5)The sound source localization device based on TMS320VC5509A is designed,which can achieve four-way sound signal acquisition.Then,an efficient analog signal processing circuit is designed,and voice acquisition and SDRAM are extended.At the same time,the user interface based on VB6.0 to display sound source location is programmed.(6)The algorithm is tested by doing a lot of experiments,the angle error and distance error of the improved algorithm are ±10°and ±15cm,respectively,the success rate is 80%.At last,the factors that affect the results of the experiment are analyzed.