Method of Sign Language Recognition Based on Temporal Shift Module and Two-Stream Networks
CAI Chang
LIN Jingyu
Abstract：In the existing sign language recognition methods,multimodal images are widely used,but the multimodal data is complex and difficult to operate.In addition,the existing sign language recognition methods can not effectively aggregate the global information of human body and the local information of motion region.In order to improve the sign language recognition method,this paper proposes a sign language recognition method based on residual temporal shift module and two-stream networks,which only us-es RGB image.The two branches of two-stream networks are improved to global image branch and local branch of motion region.The semantic segmentation algorithm is used to solve the problem of hand localization.The two branches effectively aggregate the global information and the motion region information through data fusion.Experiments on SLR500 open source dataset show that the recognition rate of this method is up to 94.7%.
Keywords：sign language recognitiontwo-stream networkstemporal shiftglobal featurelocal semantic segmentation of moving regiondata fusion
Publication Date：2023-12-20
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：6( 2841-2845,2851 )