Spoken Digits Recognition Using Wavelet Transform and Power Spectrum Density Estimation
Abd-Errahim Maazouzi, Nabil Aqili, Ahmed Hammouch, Mourad Raji · International Review on Computers and Software (IRECOS) · 2016
This work aims to develop an automatic recognition system for isolated spoken words based on power spectrum density estimation and similarity measurements. The pre-processing step prepares the signal for the ulterior phases. For features extraction, we apply the discrete wavelet transform to the speech signal and then the algorithm of Welch is used to estimate the power spectrum density. At the stage of matching, we determine the similarity between power spectra using discrete to continuous algorithm. The experiments give considerable recognition rates with the parameters used.