Abstract: The speaker tracking method of the distributed unscented Kalman filter (DUKF) needs accurate values of the state transition noise covariance matrix and the measurement noise covariance matrix. Otherwise, the tracking performance will degenerate if their values deviate from the actual values. In this paper, the Sage-Husa adaptive strategy is adopted to iteratively estimate the local state transition noise covariance matrix and measurement noise covariance matrix after updating the DUKF measurement. With the iteration of DUKF measurements, the estimates of the state transition noise covariance matrix and the measurement noise covariance matrix gradually approach to their actual values, thus the speaker tracking accuracy is improved. Experimental results show that the speaker tracking method with the proposed distributed adaptive UKF (DAUKF) achieves better tracking performance than with the conventional DUKF even under poor noise and reverberation conditions, stronger robustness under node damage conditions, and can obtain more accurate speaker position information.