NOISE-ROBUST PROCESSING ALGORITHM OF VOICE COMMANDS FOR VOICE CONTROL SYSTEMS 

Alimuradov Alan Kazanferovich, candidate of technical sciences, director of student research and production business incubator, researcher of research department, Penza State University (40 Krasnaya street, Penza, Russia), alansapfir@yandex.ru

004.934

Background. A daily rhythm of the environment requires a voice control system (VCS) to function in conditions of intense interference. Because of noise, a difference between the registered voice commands and templates (obtained during the system training) increases, and the efficiency of the VCS is rapidly deteriorating. For this reason, to improve the VCS noise immunity is urgent.
Materials and methods. To implement a noise-robust processing, a modification of classical Empirical Mode Decomposition (EMD) – the Complementary Ensemble EMD (CEEMD), is proposed. To conduct the research, a software package MATLAB for solving technical problems was used.
Results. A brief overview of the current noise-robust processing algorithms is presented. A noise-robust processing algorithm of voice commands for voice control systems is developed on the basis of the CEEMD method. The investigations of the proposed algorithm are carried out, the results of which have confirmed noise immunity improvement on average by 3,1 % in a wide range of noise (from 20 to –5 dB). 
Conclusions. In accordance with the research results, the proposed noise-robust processing algorithm of voice commands is recommended for practical application in VCS functioning in conditions of intense interference.

speech signal processing, voice control system, the Complementary Ensemble Empirical Mode Decomposition, signal-to-noise ratio, intense noise