Speech Recognition Using The Empirical Mode Decomposition Method.
M S Medvedev · Zenodo (CERN European Organization for Nuclear Research) · 2018
In this paper the using of empirical mode decomposition for creation of the Russian phoneme models in a system of speech-to-text conversion is considered. The proposed method is compared with the Fourier transform and wavelet transform. The experimental evaluation has shown that this method has advantages in the task of speech features formation (within neural network approach).