SPEECH ENHANCEMENT BASED ON AUDIBLE NOISE SUPPRESSION

A novel speech enhancement technique is presented based on the definit ion of the psychoacoustically derived quantity of audible noise spectr um and its subsequent suppression using optimal nonlinear filtering of the short-time spectral amplitude (STSA) envelope, The filter operate s with sparse spectral estimates obtained from the STSA, and, when the se parameters are accurately known, significant intelligibility gains, up to 40%, result in the processed speech signal, These parameters ca n be also estimated from noisy data, resulting into smaller but signif icant intelligibility gains.