Research And Realization Of Sound Source Localization Algorithms Based On Dsp

Sound source localization based on microphone array can be used for a videoconference system to guide its camera to pick up the faces of the speakers automatically. In addition to larger computation of current sound source localization, it is difficult to complete in embedded systems. In order to achieve the method of source localization in TMS320LF2407 DSP, we mainly do the research work as following.The paper firstly makes a primary study of three kinds of existing sound source localization methods based on microphone array, which is maximum output power controllable beam forming location method, high-resolution spectral estimation location method and sound wave arrival time difference (TDOA) location method. Based on the analysis of their disadvantage, the method of spatial spectrum estimation algorithm for MUSIC is adopted in this paper. This method has a higher positioning accuracy, although the amount of computation is more the first methods.Then we discussed the traditional signal from narrowband signal model and the microphone array signals of near field model. Analysis of the classical MUSIC and we realize the sound source localization of narrowband microphone signal. According to compress-mapping that can reduce computational complexity and easy to realize the principle of engineering, we improve the MUSIC arithmetic in near-field wideband model. A power iterations method based on contractive mapping is proposed that can run in DSP. The results of the simulation show that the algorithm is feasible.In TMS320LF2407 development board as the platform, completed a simple sound source localization experiment system. This system uses improved MUSIC algorithm. We get the data covariance matrix decomposition by making use power iterations method based on contractive mapping .And then getting noise sub-space matrix ,searching spectral peak, realizing the distance and the azimuth positioning at last.