First European Conference on Speech Communication and Technology

Paris, France
September 27-29, 1989

Speaker-Independent Word Recognition Through Telephone Networks Using Hidden Markov Models

Akihiro Imamura, Hiroshi Hamada, Ryohei Nakatsu

NTT Human Interface Laboratories, Kanagawa, Japan

In this paper, we describe an experimental telephone based system that recognizes speaker-independent isolated words. The recognition method is based on discrete HMMs. We apply the following new techniques to the conventional discrete HMM method; interpolation of observation probabilities using Fuzzy Vector Quantization, multiple model construction, model training using expanded speech end-points, and state duration control using Gaussian windows. Experiments are carried out on Japanese digits spoken by 269 speakers (238 for training, 31 for evaluation). An improvement of about 4.5% in recognition accuracy is obtained with the new techniques.

Full Paper

Bibliographic reference.  Imamura, Akihiro / Hamada, Hiroshi / Nakatsu, Ryohei (1989): "Speaker-independent word recognition through telephone networks using hidden Markov models", In EUROSPEECH-1989, 1171-1174.