Automation of Empirical Mode Decomposition to Increase Efficiency of Speech Signal Processing

The article considers an urgent problem of a subband decomposition of speech signals into non-overlapping frequency bands. A brief overview of known decomposition methods based on the Fourier transform and wavelet transform is presented. Their advantages and disadvantages are considered. The prospects for application of a new method of empirical mode decomposition are revealed. A detailed mathematical description of various decomposition methods is presented, and the relevance of automation of improved complete ensemble empirical mode decomposition with adaptive noise to increase the efficiency of speech signal processing is emphasized. A method for automated control of the improved decomposition for measuring pitch frequency, and a brief mathematical description are presented. The investigations of the method aimed at determining optimum operational settings of the improved decomposition for increasing the frequency-selective properties are carried out. In accordance with the results of the research, it is concluded that the method proposed by the authors successfully solves the problem of automating the process of the improved decomposition to increase the accuracy of the pitch frequency measurement.