A Pre-processing Method to Modify Irregular Pitch Variations for Quality Enhancement of Synthesised Speech

In low bit rate speech coders, pitch and voicing level estimation play an important role in quality of the synthesised speech. Although pitch usually evolves smoothly, sometimes it has irregular variations and as a result the estimated pitch and the voicing level differ from the real ones. This affects the performance of the speech coder. We propose to use a new modification as a preprocessor. This methodology modifies the residual speech signal such that the pitch period evolves more smoothly without distorting perceptual speech quality. Thus, the pitch and the voicing level can be determined correctly. Experimental results show that combination of the proposed method with 2.4 Kb/s MELP coder provides better quality.