SVM-based Voice Activity Detection for Distributed Specch Recognition System

Voice Activity Detection (VAD) algorithms based on machine learning techniques have shown competitive results in the area of automatic speech recognition. This paper describes a new approach of VAD based on Support Vector Machines (SVM) for Distributed Speech Recognition (DSR) system. In the proposed scheme, the speech and the non-speech frames are detected from the compressed Mel Frequency Cepstral Coefficients (MFCCs), at the back-end (e.g. server) side, with the aim of improving the VAD performance and reducing the compression bit-rate from the front-end side (e.g. client). By using the trained SVM with polynomial kernel, the SVM-based VAD show an encouraging detection results. The classification task conducted from the Aurora-2 speech database, using different noise conditions, illustrates comparable VAD performance, with respect to ETSI Advanced Front-End (ETSI-AFE) VAD algorithm. Keywordsvoice acivity detection; support vector machines; DSR system; mel frequency cepstral coefficients