Context-Dependent Multiple Distribution Phonetic Modeling with MLPs
Michael Cohen, Horacio Franco, Nelson H. Morgan, David E. Rumelhart, Victor Abrash · Neural Information Processing Systems · 1992
A number of hybrid multilayer perceptron (MLP)/hidden Markov model (HMM) speech recognition systems have been developed in recent years (Morgan and Bourlard, 1990). In this paper, we present a new MLP architecture and training algorithm which allows the modeling of context-dependent phonetic classes in a hybrid MLP/HMM framework. The new training procedure smooths MLPs trained at different degrees of context dependence in order to obtain a robust estimate of the context-dependent probabilities. Tests with the DARPA Resource Management database have shown substantial advantages of the context-dependent MLPs over earlier context-independent MLPs, and have shown substantial advantages of this hybrid approach over a pure HMM approach.