噪声相关情况下说话人跟踪方法

Speaker Tracking Method Research with Correlative Noises

利用滤波方法进行说话人跟踪时,过程与观测噪声相互独立的假设容易产生误差偏移,导致跟踪精度降低.针对这一问题,以高斯噪声为背景,在重要性权重条件最小方差意义下推导了噪声相关情况下的滤波步骤,并将该方法应用到说话人跟踪问题中.仿真实验表明,该方法较好地改善了噪声相关情况下的非线性跟踪问题,有效地提升了说话人跟踪方法的适应性和抗干扰能力.

In order to address the speaker tracking problem in the strong noise cases, a tracking method of microphone array based on particle filter with correlative noises is proposed in this paper. This method is based on the Gaussian correlative noise background. The optimal proposal distribution function of particles is deduced under the condition of minimum variance significance and is applied to the speaker tracking problem. The proposal distribution function is updated with the observation information of the speech signal arrival time difference, and provides the optimal weights for noise correlative cases. The simulation experimental results show that the proposed method improves the precision of nonlinear filtering in noise related cases, and promotes the adaptability and anti- interference ability of the speaker tracking system effectively.