A Sign Language Recognition Method Relying On Convolutional Recurrent Neural Network

Abstract

As a communication method widely used among hearing-impaired people, sign language is a natural language that transmits information through the three-dimensional space of the "manual-visual" channel, and solves the communication problem of deaf people through human-computer interaction. The problem is a qualitative leap, and human-computer interaction effectively solves the communication problem between deaf people and ordinary people. Therefore, this paper relies on the convolutional recurrent neural network (CRNN) to study the sign language recognition (SLR) method. This paper first describes the two concepts of SLR and image processing, and then builds the CNN-LSTM model through the CRNN structure, network training and parameter settings, and finally analyzes the CNN-LSTM model. Model analysis shows that the CNN-LSTM model has high SLR accuracy and good recognition performance.