Building System to Recognize Deaf Language Using BlazePalm and LSTM Network Model

This paper is carried out to serve the learning and research process on the trends of modern technology and its application to real life. The paper addresses two problems. Firstly, the system will detect the hands in the frame and extract their features. Secondly, it recognizes the meaning of gestures from hand cues. We select the BlazePalm model, a highly renowned Google model for detecting human posture and gesture for the first problem. We then used a long short-term memory (LSTM) model to train with the key points extracted from the video using the above-mentioned model for the second problem. Besides, we have built a model for recognizing the sign language of the Vietnamese and combined it into sentences with an accuracy of 96.25% for 16 actions by combining the two models and input data being video actions. The simulation results show that the model can be applied to education and daily life.