Sinusoidal+All-Pole Modification Based Spectral Smoothing for Concatenative Speech Synthesis

The conventional LPC spectral smoothing algorithm causes an evident degradation in the speech quality when the smoothing amount is large. To improve speech quality of the smoothed speech, we proposed a new spectral smoothing algorithm. The source LPC spectral envelopes are first interpolated to generate the smoothed target spectra. Then the sinusoidal + all-pole modification is performed on the source speech to get the spectra of the modified speech which will coincide with the target spectra. Experimental results show that this method can get smooth spectral envelope even if the speech boundaries have large spectral distance. Experimental results prove that this algorithm is effective on avoiding degradation in quality of smoothed speech.