New hybrid system based on MMI-neural networks for the RM speech recognition task

We present a hybrid speech recognition system for speaker independent continuous speech recognition. The system combines a novel information theory based neural network (NN) paradigm and discrete Hidden Markov models (HMMs) including State-of-the-Art techniques like state clustered triphones. The novel NN type is trained by an algorithm based on principles of self-organization that achieves maximum mutual information between the generated output labels and the basic phonetic classes. The structure of the hybrid system is quite similar to a classical VQ-HMM system but the vector quantizer (VQ) is replaced by the NN. To evaluate the system we use the speaker independent part of the resource management (RM) database. We recently obtained an important improvement by introducing a novel kind of context dependent basic classes used by the acoustic processor. The average RM recognition result with a word-pair grammar is now 95,2% what is significantly better than a classical VQ-system, slightly better than a different hybrid system with a recurrent network as probability estimator, and very close to the best continuous probability density function (pdf) HMM speech recognizers.