In this paper we report on progress made at LIMSI in speaker-independent large vocabulary speech dictation using newspaper-based speech corpora in English and French. The recognizer makes use of continuous density HMMs with Gaussian mixtures for acoustic modeling and n-gram statistics estimated on newspaper texts for language modeling. Acoustic modeling uses cepstrum-based features, context-dependent phone models (intra and interword), phone duration models, and sex-dependent models. For English the ARPA Wall Street Journal-based CSR corpus is used and for French the BREF corpus containing recordings of texts from the French newspaper Le Monde is used. Experiments were carried out with both these corpora at the phone level and at the word level with vocabularies containing up to 20,000 words. Word recognition experiments are also described for the ARPA RM task which has been widely used to evaluate and compare systems.

In this paper, we present a large vocabulary, speaker independent speech recognition system (KT-STOCK) and describe its performance over the telephone network. KT-STOCK is a stock information retrieval system with which we can obtain the current price of a stock by saying a stock name among 710 stock names listed on the Korea stock exchange. The system is an HMM(hidden Markov model)-based isolated speech recognizer which uses phoneme-like unit as a basic unit. Four digital signal processors are used for real time. And we also implement echo cancellation function for recognizing speech spoken over the voice announcement. Currently, we have achieved the recognition rate of 78.4% in the real environment. 1. INTRODUCTION One of major applications in the area of speech recognition technology is to recognize speech over the telephone network. Recently, many progress has been made in this field. Examples of such applications are services that generate new revenues. People will obtain inform...