基于正则化粒子滤波的说话人跟踪方法

Speaker tracking method based on regularized particle filtering

针对噪声与混响环境下的说话人跟踪问题,提出一种基于正则化粒子滤波(RPF)的麦克风阵列声源定位与跟踪方法.该方法在正则化粒子滤波框架下,采用适应性较强的布朗运动模型,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.实验结果表明,本文方法优于标准粒子滤波,有效提高了说话人声源跟踪系统的抗噪声与抗混响能力,即使在低信噪比(SNR=-5dB)情况下,也能有效跟踪.

Aimed at the problem of speaker tracking in noisy and reverberant environments,a speaker＇s sound source positioning and tracking method was presented based on microphone array with regularized particle filtering（RPF）.In the framework of regularized particle filtering,the stronger adaptability of Brown motion model was used to construct a likelihood function by means of calculating beamforming output energy of the microphone array.The results of simulation data showed that this method was better than the standard particle filtering,and the capability of antinoise and anti-reverberation of the speaker tracking system of sounds source was improved effectively.Tracking would be effective even in the case of low signal-noise ratio（SNR=-5 dB）.