Speaker independent isolated speech recognition for Arabic language using hybrid HMM-MLP-FCM system

Summary form only given. We compare speaker independent isolated word recognition performance obtained with standard hidden Markov models (HMM) and hybrid approaches using a multilayer perceptrons (MLP) to estimate the HMM emission probabilities. This latter approach has recently been shown particularly effective on a large vocabulary, speaker independent, speech recognition task. As a consequence, the main goal is to compare the performance, which can be achieved by the different approaches for both task dependent and independent training. Our hybrid HMM/MLP system use the fuzzy c-means (FCM) algorithm to segment the acoustic vectors.