Distinctive Phonetic Features (DPFs)-based isolated word recognition using Multilayer neural networks

Abstract

This paper describes an isolated word recognition method based on distinctive phonetic features (DPFs). The method comprises two multilayer neural networks (MLNs). The first MLN, MLNLF-DPF, maps local features (LFs) of an input speech signal into discrete DPFs and the second MLN, MLNDyn, restricts dynamics of outputted DPFs by the MLNLF-DPF. In the experiments on To hokudai Isolated Spoken-Word Database in clean acoustic environment, the proposed recognizer was found to provide a higher word correct rate (WCR) as well as word accuracy (WA) with fewer mixture components in hidden Markov models (HMMs) in comparison with the method proposed by T. Fukuda, et al. [6].