针对混响和空间噪声,本文提出一种基于三麦克风特征值分解的时延估计算法。该算法将混响和空间噪声下的时延估计构造成一个双入三出(DITO)系统的信道盲辨识问题,通过对房间冲激响应进行盲辨识来去除混响的影响,通过双初值初始化来分离声源与空间噪声的时延信息,从而提高了实际环境下时延估计算法的性能。仿真实验结果表明,在混响和空间噪声环境下,本文算法能有效地进行时延估计。
This paper proposes an eigenvalue decomposition algorithm for robust time-delay estimation (TDE) based on triple mi- crophone in situations where reverberation and spatial noise are present. This algorithm regards time delay estimation with spatial noise and reverberation as a blind channel identification issue of a double-input triple-output system, and reduce reverberation by estimating the room impulse response, separate the time delay information of speech source from spatial noise by using double intialization, the spatial noise. Simulation results show that the proposed algorithm performs well when both spatial noise and reverberation exist.