M.M. Saidi, O. Pietquin, and R. André-Obrecht (France)
EMD, IMF, MFCC, ANN, Nasalized phones, Vowels.
This work deals with the application of the Empirical Mode Decomposition (EMD) with the goal of showing it is capa bilities and limitations when applied to nasalized or oral vowel phones. The method used in this study consists in three classical stages: signal preprocessing, feature extrac tion and decision. Firstly, the speech signal is decomposed using the EMD method to extract the three first Intrinsic Mode Functions (IMF). Then, Mel-Frequency Cepstral Coefficients (MFCC) are extracted from these IMFs or a (par tial) sum of it. Finally, Artificial Neural Network (ANN) is used to distinguish nasal vowels from oral vowels in French (French database Bref80). Besides the fact that this study resulted in a significant improvement in the level of dis crimination, when we use our method compared to the stan dard application of MFCC to the original signal. It has also allowed us to know which IMFs allows to better characterize the nasal vowels from the oral vowels.
Important Links:
Go Back
