Consolidation of Robust Speaker and Speech Recognition for Intelligent Doorway Application

Consolidation of Robust Speaker and Speech Recognition for Intelligent Doorway Application

In this paper, integration of speaker identification and speech recognition for intelligent doorway application has been proposed. Two target speakers will be identified through an one-word speech utterance. Moreover, this utterance will be recognized to be a pre-defined speech command. The speaker identification in the proposed framework is based on support vector machine (SVM). The“one-versus-one”approach is applied in this paper to classify test point input utterance according to the number of votes. As for the speech recognition, we use confusion matrix to develop an efficient phonetic set for a command-based multi-lingual system, the confusion matrix calculates acoustic similarities between every two phonemes. The proposed framework has been realized in the intelligent doorway application and will be applied to many other daily life computer speech applications.