WERNICKE: A Neural Network Based, Speaker Independent, Large Vocabulary, Continuous Speech Recognition System

Dr Tony Robinson

This project is run in collaboration with Lernout and Hauspie Speech Products (Belgium), Instituto de Engenharia de Sistemas e Computadores (Portugal) and the International Computer Science Institute (CA, USA). The project is investigating the use of feedforward and recurrent multi-layer perceptrons as probability estimators for a Markov model recogniser. The hybrid recognition system has been applied to large vocabulary, speaker independent, continuous speech, recognition and competes in the ARPA continuous speech recognition evaluations. Sponsor: EC Esprit Programme.

For more details see the Abbot home page