TEXT-INDEPENDENT Mfccs VECTORS CLASSIFICATION IMPROVEMENT USING LOCAL ICA
Abdenebi Rouigueb, Salim Chitroub, Ahmed Bouridane

Abstract:
In this paper, we propose a new classification scheme of MFCCs vectors in the context of speaker identification. The solution is built around the binary SVM classification between each speaker class and the background model class over the underlying spaces of the local independent components analysis using clustering. Experiments have been conducted on a sample of the MOBIO corpus.