A spectral subtraction noise suppression method in a framebased digital communication system is described. Each frame includesa predetermined number N of audio samples, thereby giving eachframe N degrees of freedom. The method is performed by a spectralsubtraction (150) function ~(.omega.) which is based on an estimate (140)~v(.omega.) of the power spectral density of background noise of non-speechframes and an estimate (130) ~x(.omega.) of the power spectral densityof speech frames. Each speech frame is approximated (120) by aparametric model that reduces the number of degrees of freedom toless than N. The estimate ~x(.omega.) of the power spectral density of eachspeech frame is estimated (130) from the approximative parametricmodel.
展开▼
