Enhanced ASR by acoustic feature filtering

Wellekens, Christian J
ICSLP 1998, 5th International Conference on Spoken Language Processing, 30 November-4 December 1998, Sydney, Australia

Several recent results demonstrate improvement of recognition scores if some FIR filtering is applied on the trajectories of feature vectors. This paper presents a new approach where the characteristics of filters are trained together with the HMM parameters resulting in improvements of the recognition in first tests.Reestimation formulas for the cut-off frequencies of ideal LP-filters are derived as well for the impulse response coefficients of a general FIR LP-filter.


Type:
Conference
City:
Sydney
Date:
1998-11-30
Department:
Digital Security
Eurecom Ref:
181
Copyright:
© ISCA. Personal use of this material is permitted. The definitive version of this paper was published in ICSLP 1998, 5th International Conference on Spoken Language Processing, 30 November-4 December 1998, Sydney, Australia and is available at :

PERMALINK : https://www.eurecom.fr/publication/181