MelFrequency Discrete Wavelet Coefficients (MFDWC) for Isolated word Hindi Speech Recognition

MelFrequency Cepstrum Coefficient (MFCC) have become the default set of features for speech recognition right from the time they were discovered. However, with more research it was found that MFCC perform rather poorly when speech input is noisy. At the same time wavelets have gained a lot of importance recently in the field of signal processing. A natural question that was asked was "Can wavelet coefficients be used to overcome the drawbacks of MFCC?" Research in this area has only begun recently with different kinds of wavelet coefficients being tried. I explore one such set of features called MelFrequency Discrete Wavelet Coefficients (MFDWC) for the purpose of isolated word speech recognition.



Name :
EmailID :
Contact No :
Comment :