Research On Time Delay Estimation Algorithm Based On Speech Sound Source Localization In Reverberation Environment

In recent years,with the rapid development of artificial intelligence technology,the localization of speech sound source has been paid more and more attention in the fields of human-computer interaction,speech enhancement and speech recognition,in which the sound source localization algorithm using the arrival time difference is simple in principle and small in calculation,and hence is widely used in real-time localization of sound sources.The algorithm of time delay estimation is a key technology in the sound source localization algorithm based on the arrival time difference.The accuracy of the time delay estimation directly affects the accuracy of the final sound source localization.Especially in indoor environments,the presence of noise and reverberation will be detrimental to sound source localization.Therefore,in the presence of reverberation and noise,it is significant to study theoretically and practically the algorithm of time delay estimation,and to improve the accuracy of delay estimation.The contents of this paper are as follows:Firstly,the paper briefly introduces the basic pre-processing theory of speech sound source,and presents the principle of sound source localization algorithm based on arrival time difference in detail.Secondly,the paper analyzes the model of indoor sound source signal,the method of dereverberation based on inverse filter and spectrum subtraction,and introduces the theories of several commonly used time delay estimation algorithms.Since indoor reverberation has different effects on different speech frequency components,the same technology of signal processing for different frequency components can not effectively improve the accuracy of the time delay estimation algorithm.This paper focuses on the utilization of time delay estimation algorithms based on frequency division and Wigner-Ville distribution.This method firstly divides the two segments of signals received from the sensor into different sub-bands signals with Gammatone filter,and then employs the inverse-filter and spectrum subtraction to achieve the purpose of dereverberation for each sub-band signal.Finally,the correlation function of the different sub-band signals based on the Wigner-Ville distribution is calculated.After the summation,the correlation values of the original two signal segments are obtained and the discrete time corresponding to the maximum value is the estimated value of time delay.Simulation experiments show that the proposed algorithm can effectively reduce the influence of indoor reverberation and improve the accuracy of time delay estimation.Finally,this paper establishes a simple two-dimensional sound source localization system to exam and analyze the improved time delay estimation algorithm.The experimental results show the same trend of performance as the simulation experiment,that is,the improved time delay estimation algorithm proposed in this paper is superior to the other time delay estimation algorithms introduced in this paper.