APPLICATION OF SPEECH CONVERSION TO ALARYNGEAL SPEECH ENHANCEMENT

Two existing speech conversion algorithms were modified and used to en hance alaryngeal speech, The modifications were aimed at reducing spec tral distortion (bandwidth increase) in a vector-quantization (VQ) bas ed system and the spectral discontinuity in a linear multivariate regr ession (LMR) based system. Spectral distortion was compensated for by formant enhancement using chirp z-transform and cepstral weighting, Sp ectral discontinuity was alleviated using overlapping clusters during the construction of conversion mapping function, The modified VQ and L MR algorithms were used to enhance alaryngeal speech, Results of perce ptual evaluation indicated that listeners generally preferred to liste n to the alaryngeal speech samples enhanced by the modified conversion s over original samples.