Vowel Recognition with Neural Networks
ORI KOST, A. Cohen · 2005
A comparison between two Neural Networks (NN) is presented. The Multi-Layer-Perceptron (MLP) which is a deterministic NN and the Boltzmann machine (BM) - a stochastic NN, both are suited for classification tasks. The networks performance is compared using both synthesized gaussian data and real speech. The results show that the BM converges in less learning cycles than the MLP. The MLP's excess error is, however, smaller than the BM's. Both networks perform worse Man the gaussian maximum log likelihood classifier on steady state vowel utterances' classification.