Blind separation algorithm for microphone array based on recursive neural network

The blind signal separation technology is found to be an effective way of audio enhancement for the microphone array. Due to the size restriction of the microphone array, each microphone in the array is close to one another, therefore the vectors of the mix-matrix are highly similar to each other. This will case the limitation of convergence slowly and increase the error of outputs. In this paper, we give out a blind signal separation algorithm based on recursive neural network, and prove it to be useful in the situation mentioned above.