ISCA Archive ECST 1987
ISCA Archive ECST 1987

Speaker adaptive phoneme recognition in continuous speech based on vector quantization

Katsuhiko Shirai, Kazunori Mano, Koji Sano

To realize a large vocabulary continuous speech recognition, phoneme level recognition and speaker adaptation techniques are fundamental. And also, it is necessary - to develop an effective procedure of selecting possible word candidates from the large word dictionary. This paper proposes a new method applicable for continuous speech recognition which is based on a vector quantization (VQ) technique.