A Method to Detect Human Psycho-Emotical Disorder Based on the Empirical Mode Decomposition and Formant Analysis of Speech Signals

Abstract

A method for detecting psycho-emotional disorders, which is applicable to operators of safety-critical control systems, is proposed. The nature of the method is the decomposition of speech signals using the improved complete ensemble empirical mode decomposition with adaptive noise and formant analysis, namely, frequency and bandwidth measurement of the formant regions. A detailed mathematical description of decomposition varieties is presented, and a block diagram for the proposed method is given. The research of the method using the formed base of phonetically balanced speech signals registered in healthy people, and people under psycho-emotional disorders, both male and female, at the age of 18 to 79, is carried out. According to the results of the study, the developed method provides a high detection accuracy of psycho-emotional disorders: the first-kind error is 8.16%, and the second kind error is 5.66%.