Optimizing speech enhancement based on noise masked probability

An optimal approach for enhancing a speech signal degraded by uncorrelated stationary additive noise, which exploits auditory perception properties, is proposed. Based on auditory masking effects, the speech spectra estimate is performed in two cases: noisy speech spectra for noise masked and classical spectral subtraction estimate for noise unmasked. Taking into account the uncertainty of the noise presence, the enhanced speech signal spectra are obtained by a weighted sum of these two estimates, where the weights are given by the noise masked probability. The performance of the proposed speech enhancement approach has been evaluated with speech distortion and informal listening tests. Compared with Azirani's method and the MMSE-STSA estimator, results show that the speech distortion has been decreased apparently and the musical noise has been suppressed.