Research On Sound Source Localization Based On Microphone Array

In the new era where various technologies emerge in endlessly,the development of multimedia technology can be said to be advancing by leaps and bounds.Various intelligent devices enter the homes of ordinary people,enriching and facilitating people's lives.Speech signal processing technology is an important part of multimedia technology and has received extensive research attention.Sound source localization is a popular research object in the field of speech signal processing.The sound source localization technology based on the microphone array has a wide range of applications in the fields of smart speakers and video conferencing.This thesis mainly studies the positioning theory based on the time difference of arrival.The specific research contents are as follows:First,the research background of sound source localization technology and the current research status at home and abroad are reviewed.Then,the factors that affect the effect of sound source localization and the preprocessing part of the sound source localization algorithm are briefly introduced.Then,in view of the problem that the generalized cross-correlation algorithm is not effective in the presence of noise and reverberation,the cepstral pre-filtering algorithm only improves its anti-reverberation effect,and the performance is still poor in the low SNR environment,the improvement is made.By introducing second cross-correlation,a second cross-correlation algorithm based on cepstral pre-filtering is proposed.At the same time,the phase transformation weighting function is improved by using second cross-correlation power spectrum instead of cross-correlation power spectrum of two input signals and introducing weighting factor,it makes the algorithm more robust to noise and reverberation.Simulation results show that: the improved second cross-correlation algorithm based on cepstral pre-filtering is superior to the delay estimation algorithm based on cepstral pre-filtering and the phase transformation weighted generalized cross-correlation algorithm in anti-reverberation and anti-noise ability.In the case of low SNR and strong reverberation,more accurate time delay estimation can still be obtained.Finally,the phase transformation weighted steered response power algorithm has the disadvantages of poor positioning effect when the noise and reverberation increase and heavy computation,a hybrid localization algorithm based on improved steered response power algorithm is proposed.By introducing the aforementioned improved second cross-correlation algorithm based on cepstral pre-filtering,the obtained secondary cross-correlation function is substituted for the phase transform weighted generalized cross-correlation function to improve the anti-reverberation and anti-noise ability of the positioning algorithm.However,the positioning algorithm has the defect of a large amount of calculation and a slow running speed.Thus,the five-element cross-shaped array is first used for geometric positioning.A rough localization result is obtained.Then,the improved steered response power algorithm is used to search twice based on the location result.Through two-step positioning,the positioning speed is improved.Simulation results show that: the hybrid localization algorithm is superior to the geometric localization algorithm and phase transformation weighted steered response power algorithm in terms of anti-reverberation and anti-noise ability.And the positioning speed is also improved.