SVM classifiers for ASR: A discussion about parameterization

José Miguel García-Cabellos, Carmen Peláez-Moreno, Ascensión Gallardo-Antolín, Fernando Pérez-Cruz, Fernando Díaz-de-María

Abstract: Automatic Speech Recognition (ASR) is essentially a problem of pattern classification, however, the time dimension of the speech signal has prevented to pose ASR as a simple static classification problem. Support Vector Machine (SVM) classifiers could provide an appropriate solution, since they are very well adapted to high-dimension classification problems. Nevertheless, the use of SVMs for ASR is by no means straightforward, because SVM classifiers require a fixed-dimension input. In this paper we propose and compare three alternatives for adapting the parameterization to the fixed-input dimension required by SVMs. We show that SVM classifiers outperforms the conventional HMM-based ASR system, when the speech signal is parameterised at properly selected instants.