Abstract
The article is devoted to the features of speech commands recognition using an artificial neural network. The existing methods of speech recognition are considered. The algorithm of operation of the speech recognition system is presented. The stages of speech signal processing are described, indicating widely used methods. It is shown that the use of neural network analysis increases the efficiency of speech signal processing. As a type of neural network architecture, we used a recurrent bidirectional deep learning neural network Bidirectional Long Short-Term Memory and a method for obtaining mel-frequency cepstral coefficients. The choice of this neural network is due to the fact that the network uses both a priori and a posteriori information. To train the neural network in the study, a ready-made open database of speech commands Google Speech Commands Dataset was used. An experimental study of speech command recognition using BiLSTM artificial neural network was carried out using the Matlab software package and the AudioLabeler additional application.