花沢利行,川端豪,鹿野清宏
HMM音韻認識におけるモデル継続時間長の制御法
Abstract:Two kinds of duration control for HMM (Hidden Markov Model) phoneme
recognition are proposed: phoneme-duration control for an HMM phone model
and a state-duration control for an HMM state. The phoneme-duration control is
carried out by combining an HMM output probability with a phoneme duration
penalty. The phoneme duration penalty is calculated using a phoneme duration
histogram obtained from training samples. Phoneme duration control is effective
in discriminating phonemes with different durations such as /n/ and /N/. State-
duration control is realized as a state duration penalty calculated from an HMM
state duration distribution of training samples. State- duration control is
effective in discriminating phonemes with different event structures such as /s/
and /ts/. Recognition experiments are carried out using Japanese phonemes
extracted from an isolated word database uttered by one male speaker. The
phoneme recognition rate is improved from 84.8% to 89.8% using these
duration control methods.