Speaker Tracking Based on Distributed Particle Filter in Distributed Microphone Networks
提出一种基于分布式粒子滤波的说话人跟踪方法，利用麦克风网络各节点的广义互相关函数构建观测，通过改进的粒子滤波实现鲁棒跟踪，仅需局部通信且抗节点失效。
A speaker tracking method based on a distributed particle filter (DPF) for distributed microphone networks is proposed in this paper. First, the generalized cross-correlation (GCC) function is estimated at each node. To cope with the spurious effects due to the noise or reverberation, multiple delays related to the largest local peaks of the GCC constitute the local observation. Next, based on an optimal fusion rule, a modified DPF is presented, and a modified multiple-hypothesis model is also developed as its likelihood function by incorporating the information of the GCC. Finally, the modified DPF is used to track a moving speaker with a distributed microphone network. The proposed method requires only local communication among neighboring nodes, and is robust against nodes failure. Simulation and real-world experimental results demonstrate the validity of the proposed method.