Phonetic vocoding with speaker adaptation

Abstract

This paper describes a phonetic vocoding scheme which relies on speaker adaptation to capture important speaker characteristics. These are typically lost in phonetic vocoders which transmit only information about the phones which are recognized, together with some prosodic information. In our scheme, however, additional speaker characteristics are transmitted in vowel regions (average values of LSP coefficients for each phone). This additional information yielded potentially good speaker recognizability results, in informal listening tests, while still achieving a rather low average bit rate, suitable for many transmission and storage applications. This work extends our previous phonetic vocoding scheme described in [5]. The vocoder is now fully quantized and the number of transmitted parameters had been significantly reduced.

Improving speaker recognisability in phonetic vocoders

“…Our adaptation strategy is slightly different [7], being based on the modification of LSP coefficients, instead of directly modifying formant frequencies and bandwidths. As discussed in [2], the closer two consecutive LSP coefficients are together, the narrower the bandwidth of the corresponding pole of the vocal tract filter.…”

“…Relatively to our previous work [7], new codebook generation strategies have been derived, which we will discuss in section 2. Automatic speech segmentation is described in section 3.…”