Study On The Algorithm For Sound Source Localization Based On Microphone Array

Array signal processing is an important direction in the areas of signal processing,which developed rapidly in nearly forty years,its application involves communication,earthquake,radar,and many other fields.Coherent signal DOA(direction of arrival)estimation is a hot topic of array signal processing,contraposition the DOA estimation of coherent signals,scholars put forward many algorithms.All the algorithms can be departed into two types:decreasing dimension algorithms and nondecreasing dimension algorithms.Spatial smoothing technique is a common technique of decorrelation.The technique have good performance of estimating coherent sources with little complexity and easier to be accomplished.However,the effective aperture of array will be reduced with the decomposition process of coherent signal,in addition,the estimation performance is deteriorated under the circumstance of lower SNR(signal-noise-ratio).In the thesis,the author proposed some improved algorithm after analyzing the spatial smoothing method and the questions of existing in them.The main research work is as follows.1.Based on the uniform linear array,the traditional algorithm of DOA estimation,such as MUSIC,the ESPRIT algorithm and algorithm of feature space MUSIC,are discussed.The conventional algorithm's limitations is analyzed,its performance will declined rapidly under the circumstance of coherent signals,at the same time,the computer simulation results are given.2.The basic principle of spatial smoothing algorithm are given,several efficient spatial smoothing methods,such as forward spatial smoothing algorithm,backward spatial smoothing algorithm,forward backward spatial smoothing method and improved spatial smoothing algorithm,are introduced.Among them,the forward-backward spatial smoothing and forward spatial smoothing algorithm only use the autocorrelation of outputs of the individual subarrays,but ignoring the correlation information array,which lose part of the matrix of information.Improved backward spatial smoothing algorithm overcomes this shortcoming,which can take full advantage of the autocorrelation and cross-correlation information,minimizing the damage of the effective aperture of the array,thus the ability of decoherence is improved.Since these algorithms will sacrifice effective array of array antenna,thus reducing the number of sources that can be identified.3.Contraposition the limitation of traditional spatial smoothing techniques,a modified MUSIC algorithm(MMUSIC)is discussed.The algorithm will not lose the effective aperture of the array,in addition,it can reduce the correlation coefficient between the source.However,the MMUSIC algorithm has a obvious drawback,that's,algorithm performance decreased rapidly when the angle of a coherent signal source is closely.Therefore,combined with improved bidirectional spatial smoothing algorithm with the feature space MUSIC method,a MUSIC feature space method based on spatial smoothing technique is proposed,namely the ISS-ESMUSIC method.The method can estimate the angle of adjacent coherent signals effectively,which improves the DOA estimation performance of algorithm.The simulation results show that the improved method not only estimate the independent signals,but also estimate the angle of adjacent coherent signals with small SNR effectively.