Date Added: Apr 2011
There are many situations where speech is affected by different kind of acoustic noise. The authors propose an improved spectral subtraction method for the reduction of colored acoustic noise added to the speech. Their implementation uses a multi-band spectral over subtraction method to reduce the colored noise. They use a non-linear Bark scale distribution to estimate the over-subtraction factor. The noise power spectral density is estimated, using a time-recursive algorithm, by tracking the minimum of the noisy speech spectrum in each frequency band. Simulations show a better quality in terms of Itakura Saito distance and perceptual evaluation of quality for the enhanced speech.