ANALYSIS OF EXISTING METHODS AND MODELS OF SPEECH RECOGNITION BASED ON NEURAL NETWORKS

Fu Wenwei

The purpose of this work is to analyze existing methods and models of speech recognition based on neural networks. The features and characteristics of the most effective neural network models used for speech recognition are studied. The positive and negative sides of these models are highlighted. The models were compared and the most promising ones were identified. In conclusion, the paper notes the high prospects of using neural networks and, in particular, the convolution model of a neural network for speech recognition.

Keywords:speech, model, neural network, CNN neural network, LSTM neural network, Kohonen neural network