基于分层采样粒子滤波的麦克风阵列说话人跟踪方法

Microphone Array Speaker Tracking Using Layered Sampling Particle Filtering

针对噪声与混响环境下的说话人跟踪问题,本文提出了一种基于粒子滤波的麦克风阵列声源定位与跟踪方法.该方法在粒子滤波框架下,将无混响影响的语音建立信号作为观测信息,通过计算麦克风阵列波束形成器的输出能量来构建似然函数,同时考虑语音信号不同频率成分在声源定位中的作用,利用分层采样方法提高粒子的采样效率.实验结果表明,本文方法提高了说话人声源跟踪系统的抗噪声与抗混响能力.

A new sound source localization and tracking method based on microphone array is presented for speaker tracking problem in noisy and reverberant environments.In the framework of particle filtering,the echo-free onset signal is extracted accordhag to the echo-avoidance model of the precedence effect, and likelihood function is constructed by the output power of beamformer. Considering the effects of different frequency components on localization and tracking, layered sampling approach is used to improve the computational efficiency of particle filter. The results of simulation data and real data show that the proposed method can reliably track the speaker even in moderately reverberant rooms.