Analysis of R-C3D optimized sign language recognition and acquisition system in the context of multilingual communication
Gesture is a natural and intuitive mode of interpersonal communication,and gesture recognition is an essential key technology for achieving the new generation of human-computer interaction.Due to the diversity,polysemy,and temporal and spatial differences of gestures themselves,sign language recognition is also a current research hotspot.Research is based on static sign lan-guage recognition,using convolutional neural networks to extract feature information from gestures,and training corresponding network models.Then input the results into the regional convolutional 3D network for dynamic continuous sign language recognition.Research has shown that after using deep network feature extraction networks,compared to the original region convolutional 3D network,the performance of continuous sign language recognition is better,the loss curve converges faster,and the final loss value is lower.When improving the model for continuous sign language action recognition,the average accuracy of the entire class increased by 4.59％,the speed increased by 58.16％,and the recognition ability and accuracy were greatly improved.It can be seen that the algorithms stud-ied have significant importance in sign language recognition and image dynamic recognition.
deep learningsign language recognitionneural networkregional convolutional 3D Network
万方数据
维普
