Title :
TDNN labeling for a HMM recognizer
Author :
Ma, Weiye ; Compernolle, Dirk Van
Author_Institution :
Katholieke Univ., Leuven, Heverlee, Belgium
Abstract :
A system which combines the good short-time classification properties of the time delay neural network (TDNN) with the good integration and overall recognition capabilities of hidden Markov models (HMMs) is proposed for a speaker-independent speech recognizer. The standard vector quantization is replaced by a TDNN labeler giving phonelike labels. In order to avoid hand segmentation for the training of the TDNN, a separate HMM and a Viterbi alignment derived from it are used. This gives a coarse phonetic segmentation of the training data
Keywords :
Markov processes; neural nets; speech recognition; HMM recognizer; TDNN labeler; TDNN labeling; Viterbi alignment; coarse phonetic segmentation; hidden Markov models; phoneme recognition; speaker-independent speech recognizer; time delay neural network; training data; vector quantization; Computer networks; Delay effects; Hidden Markov models; Labeling; Multi-layer neural network; Neural networks; Power system modeling; Speech recognition; Training data; Vector quantization; Viterbi algorithm;
Conference_Titel :
Acoustics, Speech, and Signal Processing, 1990. ICASSP-90., 1990 International Conference on
Conference_Location :
Albuquerque, NM
DOI :
10.1109/ICASSP.1990.115728
