Natural gradient works efficiently in learning
Neural Computation
Hi-index | 0.00 |
In this paper we propose a scheme for online convolutive blind acoustic source separation in frequency domain. A restriction term of DOA (Direction of Arrival) for sources is added to the cost function of mutual information. Minimizing the modified mutual information in each frequency bin, the source order can be adjusted on line. So the problem of permutation ambiguity can be settled on line, and it does not need the location information of receivers like classical DOA method. Simulation results show that the proposed algorithm has better performance than other existing methods.