Based On Microphone Array Sound Source Localization Study

Microphone arrays are widely used in audio/video conferences, speech recognitions, and speech enhancements etc. Speech localization technology based on microphone arrays is one of the array signal processing key technologies. It is a new hotspot in acoustic signal processing field. Focus on this technology, the following work has been done:The basic theories of main kinds of speech localization technologies have been summarized. A realization mode of room impulse response in room reverberation environment is proposed, using ray-tracing method. Acoustic signal pre-process is proposed, because of the particularity of the acoustic signal and the room reverberation environment, voice activity detection etc. In actual use, acoustic signal is either in near field or in far field, so we research the speech localization technology in the two conditions respectively. In near field, we use near-field adaptive beamforming and speech localization based on higher order statistics. The presented robust near-field adaptive beamformer is developed by combining a new formulation of the point-constrained robust beamformer and a self-calibration technique. 3-D ESPRIT algorithm is used to estimate the speech location, and compared with the traditional near-filed source localization method. In far field, we researched the subspace method. It is consists of two parts, source direction technical and source locating technical. The speech localization technology using orthocenter localization algorithm has been proposed. A subspace method 2-D MUSIC algorithm is applied to estimate the plane angle and the polar angle. With the bearing results of sub-arrays, main-array confirms a point in the triangle enclosed by location lines as the target position.