CASPIAN JOURNAL

MANAGEMENT AND HIGH TECHNOLOGIES

RECOGNITION OF SPEECH COMMANDS USING NEURAL NETWORKS

Read

Egunov Vitaly A., Paniulaitis Stanislav V. RECOGNITION OF SPEECH COMMANDS USING NEURAL NETWORKS // Caspian journal : management and high technologies. — 2020. — №3. — pp. 53-61.

Egunov Vitaly A. - Volgograd State Technical University, vegunov@mail.ru

Paniulaitis Stanislav V. - Volgograd State Technical University, staspan34@gmail.com

The purpose of this work was to simplify the management of various automated systems, which in this case include AIDS for the disabled, the “smart home” system, etc. These systems are managed by organizing a human - machine interface using neural networks. The basis of this interface is the recognition of voice commands by a system based on a microcontroller with relatively low computational characteristics. The implemented neural network surpasses the popular FANN library [11] in terms of accuracy, learning speed, recognition speed, and required hardware resources. Research work aimed at obtaining the optimal structure of an artificial neural network designed for speech command recognition has been carried out, and appropriate algorithms and software have been developed. A technique based on neural network technologies has been developed that allows fast and efficient training of the recognition system on a microcontroller. The influence of noise on the quality of recognition is determined, and the influence of neural network topology on the quality of recognition has been determined.