Research On Sound Source Localization Based On Microphone Array

Microphone array-based sound source localization technology refers to the use of multiple microphones at different location points to collect the sound signal and then use algorithms to process the collected sound signal,thereby obtaining information such as the azimuth,elevation and distance of the sound source point relative to the microphone reference point.The effect of the microphone array on the reception of the sound source signal is directly related to its structure.The different structures of the microphone array will directly lead to its different array flow shape,and thus the use of the sound source localization algorithm on the sound source positioning,its positioning effect due to the different structure of the microphone array will also have certain differences.In the actual environment and applications,most of the sound sources are coherent sources.For coherent sources of the sound source,localization is a research hotspot in array signal processing.Most decoherence algorithms are for the processing of uniform linear array structures,for nonlinear array structures are still relatively little research.This thesis has done research related to the above two sound source localization problems,the main research content is as follows.(1)The basic theory of array signal processing is studied.A mathematical model of the acoustic signal received by the microphone array is established.Three microphone array flow forms are derived,namely,uniform linear array flow-form and uniform circular logarithmic spiral array flow-form.(2)The three microphone array structures were studied for sound source localization using the MUSIC algorithm to compare the performance of the MUSIC algorithm on different array structures.After the simulation study,it is concluded that the uniform line array can only accurately locate the signal azimuth angle,the uniform circular display and the logarithmic spiral array can identify the source azimuth and pitch angle when the source is localized,and the logarithmic spiral array has fewer side flaps in the imaging map than the uniform circular array.The spectral peaks are more smoothly imaged.(3)For the research of coherent signal sources,three decoherence algorithms based on a uniform linear array structure are first investigated: the forward-space smoothing algorithm,the backwards-space smoothing algorithm and the modified MUSIC algorithm.Simulation studies show that the modified MUSIC algorithm outperforms the other two decoherence algorithms.To address the problem that the traditional decoherence algorithm cannot be used for logarithmic spiral array,this paper proposes a VA-MMUSIC algorithm that virtualizes the covariance matrix of the data received by a logarithmic spiral array into the covariance matrix of a linear array and then uses the idea of the modified MUSIC algorithm to localize the source of the coherent signal.Simulation studies show that this algorithm can achieve the sound source localization of coherent signal sources by logarithmic spiral arrays.However,the algorithm requires a spectral peak search,and the algorithm is time-consuming.The VAMESPRIT algorithm is proposed,which is significantly less time-consuming but has a slightly larger error than the VA-MMUSIC algorithm.(4)The experimental hardware platform was built,and the algorithm program was written in MATLAB.When a logarithmic spiral array was used to localize a single source,the experimental results have shown that the logarithmic spiral array was able to accurately locate the single source with fewer side peaks.Two logarithmic spiral array-based decoherence algorithms are also verified.Experimental results shown that the VA-MMUSIC algorithm has less error than the VA-MESPRIT algorithm,but the VA-MMUSIC algorithm was computationally time-consuming.