Spectral subtraction iterated with weighting factors

Spectral subtraction (SS) is a simple and effective method for speech enhancement. In this paper, the performance of SS is improved under severe noise conditions. Two methods are presented in both of which SS is utilized iteratively with the aid of an adjustable weighting factor. Two types of iteration; long-time iteration and short-time iteration, are investigated. Experimental results show that the two methods provide a performance improvement relative to the standard SS method.