A phonetic Vocoder for Very-Low-Rate Speech Coding

Abstract

In this paper, we describe a phonetic vocoder based on the concatenation of syllable-units which represent speech waves by extremely low rate (100 bits/s) using a speech recognition tequnique. We take syllables into consideration as the unit of recognition/synthesis, because a syllable contains the coarticulation effect between a consonant and a vowel. Speech waves are transformed into a sequence of frames, each of which consists of LPC cepstrum, PARCOR coefficients, pitch and power. After the O(n) DP matching with 500 reference patterns, the input speech is transformed into a sequence of Japanese syllables. The information of recognized syllable contains the category of syllables, duration, power and pitch, and is represented by 16 bits. Using this vocoder, speech can be represented by only 100 bits/sec and the intelligibility of phrase for an unlimited task is about 60%. If the number of references is enlarged, say, 1600 patterns, the intelligibility becomes of more than 70%. In this case, the coding rate is about 112 bits/sec.