一种基于粒子滤波的鲁棒声源跟踪算法

为了提高噪声和混响环境中声源跟踪的精度,提出一种基于粒子滤波的鲁棒声源跟踪算法。在基于麦克风阵列的粒子滤波声源跟踪算法框架下,该方法分别采用常规可控波束形成和相位变换加权的可控响应功率两种声源定位函数来构造似然函数,并且分别用这两种似然函数评价粒子权重,再将各自的粒子权重归一化并对两种粒子权重做加权平均得到新的粒子权重。仿真结果表明,在高信噪比或弱混响条件下,该方法的跟踪性能与传统方法接近;在信噪比低于10dB,混响时间大于300ms条件下,该算法的跟踪误差比传统算法减小15%~20%。文中提出的声源跟踪算法结合了两种定位函数的优点,在低信噪比,较强混响环境下有好的鲁棒性。

Robust acoustic source tracking algorithm based on particle filtering

A robust particle filter for tracking acoustic source is presented to improve tracking accuracy in noisy and reverberant environment.A general framework of particle filtering for acoustic source tracking based on microphone array has been available.Under the framework,two localization functions,viz.the conventional steered beamformer and the steered response power-phase transform ave used to form their corresponding likelihood functions.The particle weight is measured by the two likelihood functions respectively.Then every particle has two kinds of weight which will be normalized respectively.At last,the new particle weight will be produced by computing the weighted average of the two kinds of weight.The simulation results show that the tracking accuracy of both the proposed algorithm and the traditional algorithm is similar in the condition of high signal-to-noise ratio or mild reverberation.When the signal-to-noise ratio is lower than 10dB,and the reverberation time is more than 300ms,the tracking error of the proposed algorithm decreases by 15%-20%,compared to the traditional algorithm.The method integrates the merits of the two localization functions,thus has good robustness in low signal-to-noise ratio and moderate reverberation environment.