Speech Source 3D Localization Focusing Algorithms Based on Microphone Array

To classical wideband focusing DOA estimation algorithm, an initial estimate of the DOA's and several iterations are required, and the performance of real time is bad. A speech source 3D localization algorithm based on wideband focusing theory is presented in this paper, and it doesn't require initial estimations and iterations. The optimal criterion of this algorithm is the MMSE, i.e. the minimized mean square error of the correlation matrix of focused frequency and the others. This algorithm has the advantage of the low computational complexity for the real-time applications. Simulation results show the high performance of the 3D localization estimation is obtained when the proposed algorithm is used in planar uniform circle array.