Speech recognition system including speech section detecting section

Abstract

A trained vector generation sectiongenerates beforehand a trained vector v of unvoiced sounds. An LPC Cepstrum analysis sectiongenerates a feature vector A of a voice within the non-voice period, an inner product operation sectioncalculates an inner product value VA between the feature vector A and the trained vector V, and a threshold generation sectiongenerates a threshold θv on the basis of the inner product value VA. Also, the LFC Cepstrum analysis sectiongenerates a prediction residual power ε of the signal within the non-voice period, and the threshold generation sectiongenerates a threshold THD on the basis of the prediction residual power ε. If the voice is actually uttered, the LPC Cepstrum analysis sectiongenerates the feature vector A and the prediction residual power ε, the inner product operation sectioncalculates an inner product value VA between the feature vector A of input signal Saf and the trained vector V, and a threshold determination sectioncompares the inner product value VA with the threshold θv and determines the voice section if θv≦VA. Also, a threshold determination sectioncompares the prediction residual power ε of input signal Saf with the threshold THD and determines the voice section if THD≦ε. The voice section is finally defined if θv≦VA or THD≦ε, and the input signal Svc for voice recognition is extracted.