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

A new technique for speech enhancement is proposed, which is based on psychoacoustic criteria. The technique employs the auditory masking threshold, in order to extract information for the audible noise components. Those components are then removed using adaptive non-linear spectral modification. The main advantage of such an approach is that the speech signal is not affected by processing. In addition, very little information of the features of the noise is required. 11-3590-7803-0946-4/93 $3.00 0 1993 IEEE

No abstract

No abstract