Distributed iterated extended Kalman filter for speaker tracking in microphone array networks

In this paper, a distributed iterated extended Kalman filter (DIEKF) is proposed to estimate time-varying speaker's positions in microphone array networks. In the proposed method, the time difference of arrival (TDOA) of the speech signals received by each pair of microphones is estimated by the generalized cross-correlation (GCC) method. Meanwhile, each node in the networks checks whether the TDOA is true or not and ensures that the outliers caused by room reverberations are removed. Then the distributed IEKF is used to perform speaker tracking. By performing several iterations, the proposed speaker tracking method can improve the tracking performance successfully. Especially, it can capture the speaker's moving trajectory faster when the initial target state is far from optimal value. Therefore, the proposed method can obtain a smoothed trajectory of the speaker's motion accurately and robustly in noisy and reverberant environments. Simulation results reveal the tracking performance of the proposed method.