Research of Frequency-Selective Properties of Empirical Mode Decomposition Methods for Speech Signals´ Pitch Frequency Estimation

The problem of efficiency increasing of speech signals´ analysis in pitch frequency (PF) estimation algorithms at the pre-processing stage is considered. A brief review of Empirical Mode Decomposition (EMD) methods is presented. The Complementary Ensemble Empirical Mode Decomposition (CEEMD) is offered as a new method of adaptive decomposition of speech signals. The research results of frequency-selective properties of the CEEMD and its application in PF estimation algorithm at the preprocessing stage are provided.