Speech recognition in a noisy car environment based on LP of the one-sided autocorrelation sequence and robust similarity measuring techniques

The performance of the existing speech recognition systems degrades rapidly in the presence of background noise. A novel representation of the speech signal, which is based on Linear Prediction of the One-Sided Autocorrelation sequence (OSALPC), has shown to be attractive for noisy speech recognition because of both its high recognition performance with respect to the conventional LPC in severe conditions of additive white noise and its computational simplicity. The aim of this work is twofold: (1) to show that OSALPC also achieves a good performance in a case of real noisy speech (in a car environment), and (2) to explore its combination with several robust similarity measuring techniques, showing that its performance improves by using cepstral liftering, dynamic features and multilabeling. R&urn6 Les performances des systkmes actuels de reconnaissance de parole se dCgradent rapidement en prksence de bruit. Une nouvelle reprksentation du signal de parole, basee sur la p&diction lintaire de sCquence d'autocorr6lation unilat&ale (One-Sided Autocorrelation Linear Prediction: OSALPC), s'est av&6e &tre inGressante pour la reconnaissance de la parole bruitie, B la fois pour ses bonnes performances (par rapport au codage LPC conventionnel) dans des conditions diffciles de bruit blanc additif et pour sa simplicid de calcul. Le but du travail pr&ent6 dans cet article est double: (1) il s'agit de montrer que OSALPC foumit Cgalement de bonnes performances pour de la parole bruitie en contexte r6el d'usage (en voiture), et (2) d'explorer sa combinaison avec diverses techniques robustes de mesure de similar@ en montrant que ses performances s'amCliorent en utilisant une ponderation cepstrale, des indices dynamiques et l'itiquetage multiple.