雲间烟火Efficient Acoustic Front-End Processing for Tamil Speech Recognition using Modified GFCC Features
to propose efficient front-end processing techniques for Tamil Speech Recognition by implementing Five Pass Pre-processing and modified GFCC features.
The paper is organized as follows. Section 2 presents the related works on GFCC technique. Section 3 broadly discusses the proposed Pre-processing and feature extraction techniques introduced for Tamil ASR. The experimental results of the existing and proposed techniques are briefly presented in section 4. The findings and discussions are given in section 5. Finally, conclusion than conventional features as the frequency bands spacing is similar to the auditory ERB scale. In order to test the robustness of wavelet based features, various noise conditions are involved using NOISEX-92 database. The experimental outcome proved that the WERBC is better when compared to the WMFCC especially in case of noisy condition.
