Fast multi-channel colored noise reduction algorithm using a novel statistical data fusion estimation

This paper proposes a fast multi-channel colored noise reduction algorithm based on a statistical data fusion estimation. Conventional multi-channel algorithms for colored noise reduction have a difficulty in noise covariance matrix estimation and take more computational time. The proposed algorithm has a much fast speed without estimation problem of the noise covariance matrix. Moreover, the proposed algorithm has a good performance of unbiased estimation when the noisy speech signals come from a same speech source. Simulation results confirm that the proposed algorithm possesses colored noise reduction performance with a fast speed.