Text Independent Speaker Verification Based on Mixing ICA Overcomplete Basis Functions

Text-independent MFCCs vectors classification improvement using local ICA

“…In this sense, the selected independent components are applied to a VQ model for SV purpose [8]. An ICA overcomplete representation is used to capture the basis functions of the speech signal and the acoustical basis functions of formants in order to use them as features for the text-independent SV [6]. The ICA basis functions learned locally on speech segments are used as features for SR in [7].…”

“…Independent component analysis (ICA) [4][5] is a highorder statistics method, which is used in several works for the extraction and the enhancement of speech features for SR purpose [6][7] [8]. In this paper, our goal is to propose a new classification method of MFCCs vectors for text-independent SR by combining local ICA method [9] and SVM.…”