ENHANCEMENT OF SPEECH SIGNAL SEGMENTATION USING TEAGER ENERGY OPERATOR

Abstract

Background. Speech signal segmentation is detection of the boundaries of the beginning and the end of sections of voiced and unvoiced speech, and pauses. Accurate detection of the boundaries both improves the quality of speech signal segmentation, and reduces the number of computational operations. The aim of the work is to improve the efficiency of segmentation based on the energy analysis of speech signals using the Teager energy operator. Materials and methods. The second-order differential Teager energy operator, which makes it possible to estimate the energy characteristics of a signal, was used in this work. The Teager operator is simple, efficient, and highly susceptible to changes in signal amplitude and frequency.The software implementation of the method was performed in ©MATLAB (MathWorks) mathematical modeling environment. Results. An improved method for speech signal segmentation, providing an increase in the efficiency of detecting voiced and unvoiced areas, and pauses, has been developed. The nature of the method is the energy analysis of speech signal fragments using the Teager energy operator; analysis of zerocrossing rate and short-term energy of the energy characteristic function. Research to assess the efficiency and noise robustness of the improved method over the known segmentation methods, was carried out. Conclusions. In accordance with the obtained research results, it was revealed that due to the good susceptibility of the Teager energy operator to sharp changes in signal amplitude and frequency, the improved method provides an increase in the segmentation efficiency by 2.97 % and 2.49 % for the 1st and 2nd kind errors, respectively.

Keywords

• speech processing
• speech segmentation
• voiced and unvoiced speech
• pauses
• teager energy operator