Power level difference as a criterion for speech enhancement

This paper deals with the problem of speech enhancement in near field condition, when two microphones are available. Proposed technique relies on the difference in power of received signals at the two microphones. This difference is employed to estimate the clean speech signal power. The method has the capability of dealing with non-stationary noise, a drawback of many noise reduction techniques. Superiority of the presented method over some of prominent methods in this field is demonstrated by conducting both subjective and objective quality tests.