Research On Sound Source Location Technology Based On Microphone Array TDOA

With the advancement of science and the continuous improvement of people's quality of life,more and more high-tech products are beginning to enter daily life.Sound-related robots are also coming into people's horizons.Sound source localization technology is a branch of speech technology and a hotspot of speech research at home and abroad.At present,the microphone-based TDOA sound source localization technology has problems such as slow positioning speed and low positioning accuracy.In view of the above problems,this thesis focuses on the speed and accuracy of time delay estimation and location estimation of TDOA algorithm.The sound source localization technology of this thesis is aimed at robots used indoors,which can assist in positioning and enriching the decision-making ability of robots.This thesis introduces and analyzes the commonly used sound source localization methods.Compared with other positioning methods,the time difference of arrival(TDOA)method can meet indoor positioning requirements.Its calculation amount is smaller,complexity is lower,and feasibility is stronger.Therefore,this method is selected for sound source localization.This thesis also introduces the speech signal pre-processing before speech signal analysis,including window function framing,endpoint detection,pre-emphasis,de-noise and other techniques.The endpoint detection is related to the speed of the positioning in this thesis.In order to improve the ability to distinguish between speech signals and non-speech signals,the spectral entropy endpoint detection method is improved.Combining EMD and spectral subtraction to filter the noisy speech signal improves the signal-to-noise ratio of the speech signal.The commonly used generalized cross-correlation delay estimation method is introduced,and the generalized cross-correlation PHAT weighting is improved,and the accuracy and stability of delay estimation are improved.Using a six-element microphone hyperboloid positioning model and step estimation(determining quadrants,then determining coordinates)improves the speed of position estimation.Using the improved quasi-Newton-K nearest neighbor method to speed up the solution of nonlinear equations and reduce the influence of errors on the estimation points.Finally,using MATLAB to experiment with the improved sound source localization method for the acquired signals,the feasibility and effectiveness of the above methods are verified.The thesis has 49 pictures,6 tables and 58 references.