The Research Of Acoustic Source Localization Based On Microphone Array In The Condition Of Reverberation

With the improvement of the intelligence of various devices,voice interaction has aroused more attemtion in our life.As a key technology of voice interaction,sound source localization is widely used in various fields because of its convenience and its insensition to electromagnetic wave and light intensity.In the past few years,more scholars pay more attention to its property improvement in order to make it suitable for more complex scenarios.They focus on noise resistance,positioning accuracy,real-time performance,equipment miniaturization,and hardware cost of the positioning system.This article focuses on the problem of reverberation interference in indoor sound source localization,and its specific work is as follows:First,this article classifies sound waves according to the relationship between the sound source and the microphone position,and reviews the relevant physical quantities describing sound waves,and uses the variation of these physical quantities with time and space to analyze the propagating rules of sound waves in different scenarios under the far-field transmission model.On this basis,an indoor reverberation model that reflects the reverberation is established.Second,the traditional time delay estimation(TDE)algorithms are studied,including generalized cross-correlation(GCC)TDE and minimum mean square error(LMS)adaptive TDE.And the advantages and disadvantages of each algorithm are compared.What's more,the characteristics of reverberation and direct sound,several dereverberation techniques are studied.By analyzing the received signal's distribution of each component in the cepstral domain,an improvered cepstral filtering GCC(CEP-GCC)algorithm is proposed.The new proposed algorithm not only overcomes the interference of reverberation,but also ensures the correlation between the two signals participating GCC calculation.In a high reverberation environment,when the reverberation time exceeds 0.4s,the root mean square error(RMSE)of the algorithm is about 80% of that of the time delay estimation algorithm based on homomorphic filtering before the improvement,and it has higher delay estimation accuracy.Third,we researched the relationship between structure of microphone array and time delay,besides,the elements of microphone array design are discussed.On this basis,the positioning algorithm based on the time difference of arrival(TDOA)is studied.The hyperboloid model is first established in space,and then the best solution is found to achieve positioning.The main solution methods include geometric positioning,least squares,iterative algorithm based on Taylor expansion,and maximum likelihood Estimation method.By comparing the pros and cons of each algorithm,a new positioning algorithm baesd on five-element cross array is proposed,which eliminates abnormal delay estimation points,and set the weight for azimuth angle according to the delays between different elements,sum them up to get the final estimated value of azimuth angle.Then different distance estimation algorithms are selected according to the introduced angle.The estimation range of this algorithm is not limited by the azimuth angle,and can locate half of the space on one side of the array plane.In high reverberation conditions,the average distance error of positioning is within 0.2m,which is much lower than that of positioning algorithms based on four-element cross array.Finally,a reverberation model is built in the MATLAB environment,and the strength of the reverberation is adjusted by controlling the reverberation time.The improved algorithm and the traditional algorithm are simulated experiments.The experimental results show that the improved algorithm has superior property in condition of high reverberation.