Измерение, мониторинг, управление, контроль (Sep 2023)
SOFTWARE AND HARDWARE IMPLEMENTATION OF A SYSTEM TO DETECT AND CLASSIFY HUMAN EMOTIONAL STATES FROM SPEECH SIGNALS
Abstract
Background. To improve the accuracy of detection and classification of emotional states in the areas of human activity associated with the risk of man-made and biogenic accidents has become an urgent task. The aim of the study is to improve the efficiency of a system to detect and classify human emotional states from speech signals. The object of the research is the structure and software and hardware implementation of the system for detecting and classifying emotional states. The subject of the study is the means and techniques for software and hardware implementation. Materials and methods. The client-server methods for hardware and software implementation of the system based on the original software modules for speech signal processing and classifying emotional disorders, as well as several hardware implementations of the server device have been used. Results. The article presents hardware and software implementation of a system to detect and classify human emotional states from speech signals based on the client-server architecture. The novelty of the proposed concept is the creation of server clusters interconnected by a high-speed channel. This approach to the organization of the client-server architecture of the system allows increasing productivity and the speed of data processing, ensuring stability and timely delivery of results in real time. Results and conclusions. The research results have evidenced the accuracy of the developed system for the classification of emotional states being 94.7–95.6 %, and that for the classification of emotions being 93.1–95.6 %. The accuracy of the developed system is on average 3.15 % and 2.35 % higher for the classification of emotional states and emotions, respectively, as compared with related products on the speech technology market. In the future, it is planned to conduct further research on the performance of hardware and software implementation of the system to detect and classify human emotional states from speech signals.
Keywords