COMPARISON OF A NEW HYBRID CONNECTIONIST-SCHMM APPROACH WITH OTHER HYBRID APPROACHES FOR SPEECH RECO - Acoustics, Speech, and Signal Processing, 1995. ICASSP-95., 1995 International Conference on

This paper compares a newly proposed hybrid connectionist-SCHMM approach [5] with other hybrid a p proaches. In the new approach a multilayer perceptron (MLP) replaces the conventional codebooks of semicontinuous HMMs. The MLP is therefore trained on s w d k d basic elements (phones and phone parts) in such a way that the outputs of the network estimate the a posteriori probabilities of these elements, given a context of input vectors. These a posteriori estimates are! converted into scaled liieliioods, which are then used as observation probabilities in the framework of <:lassical SCHMMs. The remaining parameters of the SCHMMs are trained with the well-known Baum-Welch algorithm using the estimated likeliioods of the MLP. This approach compared fkvorably with other recently proposed hybrid systems and classical approaches on an isolated German digit recognition task over telephone lines. It exhibited the highest recognition rate of all systems, followed by an approach using LVQ3 optimization of the codebook.