A Continuous Speech Recognition System Embedding MLP into HMM

We are developing a phoneme based. In [Bourlard & Wellekens]. it was shown that MLPs were approximating Maximum a Posteriori (MAP) probabilities and could thus be embedded as an emission probability estimator in HMMs. It is shown here that word recognition performance for a simple discrete density HMM system appears to be somewhat better when MLP methods are used to estimate the emission probabilities.