Speech Communication - Special issue on speech processing in adverse conditions
Discrete Time Processing of Speech Signals
Discrete Time Processing of Speech Signals
The voice segment type determination using the autocorrelation compared to cepstral method
WSEAS Transactions on Signal Processing
Hi-index | 0.00 |
When automatic speech recognition systems trained with clean signals are tested with noisy signals, deterioration in their performance have been observed. Continuous multiresolution entropy have shown to be robust to additive noise in applications to different physiological signals and, in particular, in some speech signal contexts. In this paper we present its extension to different divergences and we propose them as new dimensions at the pre-processing stage of a speech recognizer. Methods proposed here are tested with speech signals corrupted with babble and white noise. Their performance are compared with the classical mel cepstra parametrization. Results suggest that continuous multiresolution entropy related measures provide valuable information that could be considered as an extra component in a pre-processing stage.