Abstract:

A voice recognition speech enhancement method for dual microphones is proposed .This method, making use of two voice signal spatial correlation and temporal correlation, can reduce noise by spatialtemporal prediction .When the input SNR is higher than 0dB, remarkable improved SNR can be obtained. This method is based  on two microphones, with a simple structure; the reduction to a point noise is remarkable; the irrelevance of noise between two microphones is not required; and the weak correlative noise on the target direction can be reduced.

Key words: voice recognition;speech enhancement;twochannel;spatialtemporal correlation