ITRW on Non-Linear Speech Processing (NOLISP 07) Paris, France May 22-25, 2007

A Hybrid Genetic-Neural Front-End Extension for Robust Speech Recognition over Telephone Lines

Sid-Ahmed Selouani (1), Habib Hamam (2), Douglas O’Shaughnessy (3)

(1)Université de Moncton, Campus de Shippagan, (2)Université de Moncton, Campus de Moncton, (3)INRS-Énergie-Matériaux-Télécommunications, Canada

This paper presents a hybrid technique combining the Karhonen-Loève Transform (KLT), the Multilayer Perceptron (MLP) and Genetic Algorithms (GAs) to obtain less-variant Mel-frequency parameters. The advantages of such an approach are that the robustness can be reached without modifying the recognition system, and that neither assumption nor estimation of the noise are required. To evaluate the effectiveness of the proposed approach, an extensive set of continuous speech recognition experiments are carried out by using the NTIMIT telephone speech database. The results show that the proposed approach outperforms the baseline and conventional systems.

Full Paper

Bibliographic reference.  Selouani, Sid-Ahmed / Hamam, Habib / O’Shaughnessy, Douglas (2007): "A hybrid genetic-neural front-end extension for robust speech recognition over telephone lines", In NOLISP-2007, 59-62.