Articulation based admissible wavelet packet feature based on human cochlear frequency response for TIMIT speech

(in lingua inglese)

The present study focuses on noisy English speech recognition with HMM. Experiments were conducted in a sequential step. A set of wavelet features derived by using the filters similar like ERB filter was presented. Articulation effects of speech signal were considered while extracting the features. Experiments were carried out in sequential steps to see the performance of WERBC. Comparative study with baseline systems is also presented to show the robustness of the articulation based WERBC feature.