Speaker Number Estimation Based On Microphone Array

Microphone array can simultaneously obtain information of sound source in time, frequency and spatial domain. Compared with the single microphone, advantages such as high signal gain, application flexibility and anti-noise ability are achieved. Currently, microphone array is widely used to solve sound source localization and tracking, speech enhancement, blind source separation and many other problems, and has been applied in communications, computer, multimedia, digital entertainment and other fields. Precisely known number of sources is fundamental to many array signal processing algorithms to achieve high performance. Source number estimation is an important research topic in microphone array signal processing. This thesis studies speaker number estimation using microphone array and the major work is as follows:(1) Information criterion theory method and Gerschgorin disk estimation method used for narrow band source number estimation are reviewed;(2) Information theoretic criteria method and Gerschgorin disk estimation method are introduced into microphone array to detect source number based on the signal produced by the ISM wideband source method, and speaker number estimation methods using microphone array are presented;(3) Several focusing matrix construction methods of CSM wideband source method are analized, and covariance matrix transformation construction method suitable for source number estimation is combined with source number estimation methods to detect speaker number.(4) The principle of jackknifing method and its application in array signal processing are studied. Jackknifing method is introduced into the presented speaker number estimation method, and simulation results reveal its validity.