European and American Audio-Visual Speech Recognition, Using SVM in Portuguese Language
Adriano de Andrade Bresolin, Diamantino Freitas, Adrião Duarte Dória Neto, Pablo Javier Alsina · 2008
This paper proposes an audio-visual speech recognition system using SVM (support vector machine) in European and American Portuguese language. The main objective in this work is to find a model that can be used in both languages. Furthermore, two new methods to extract the mouth region (ROI-Region of interest) and lip contour are presented. Two audio and four video features are used in the experiments. These features are combined in pairs, totalizing eight tests in the speaker dependent-case. Experiments were performed at various SNRs (0-40dB) with additive white Gaussian noise. The results showed that the proposed method can be used in both languages without any adaption.