ITRW on Non-Linear Speech Processing (NOLISP 05) Barcelona, Spain April 19-22, 2005

ew Sub-band Processing Framework using Non-linear Predictive Models for Speech Feature Extraction

Mohamed Chetouani (1), Amir Hussain (2), Bruno Gas (1), Jean-Luc Zarader (1)

(1) Laboratoire des Instruments et Systemes d'lle-De-France, Université Paris VI, Paris, France
2 Dept. of Computing Science and Mathematics, University of Stirling, Scotland, UK

Speech feature extraction methods are commonly based on time and frequency processing approaches. In this paper, we propose a new framework based on sub-band processing and non-linear prediction. The key idea is to pre-process the speech signal by a filter bank. From the resulting signals, non-linear predictors are computed. The feature extraction method consists in the association of different Neural Predictive Coding (NPC) models. We apply this new framework to phoneme classification. The experiments carried out with the NTIMIT database show an improvement of the classification rates in comparison to the full-band approach. The new method gives also better performances than the traditional ones (LPC, MFCC and PLP).

Full Paper

Bibliographic reference.  Chetouani, Mohamed / Hussain, Amir / Gas, Bruno / Zarader, Jean-Luc (2005): "ew sub-band processing framework using non-linear predictive models for speech feature extraction", In NOLISP-2005, 269-274.