Enhanced recognition of Arabic numerals based on the combination of TESPAR and MFCC encoders
Abdelmajid Lamkadam, Mohammed Karim · 2024
This paper presents a summary of the work concerning the achievement of enhanced recognition of Arabic numerals, based on TESPAR technology, on the MFCC extractor and we reinforced the extraction by the combination of these two methods. The resulting combination increased the number of acoustic parameters, so to discriminate and reduce the representation space we used the LDA method, and to optimize the acoustic vectors we applied the DEF technique. In order to improve the robustness and performance of our recognition system, it is essential to have corpora recorded under conditions as close as possible to those of the target applications, and also to reduce discrepancies between the reference and test corpora, they should be recorded under the same conditions as far as possible.This work (combination) showed an observable improvement in the rate of recognition of Arabic numerals.