Machine Learning Approaches for Sign Language Recognition: An Analytical Review

sign language recognition (SLR) is developing quickly and has significant promise to close communication barriers. Four promising strategies are examined in this review: Dynamic hand gestures and their shapes are well captured by 3D Convolutional Neural Networks (3DCNNs); BLSTM-3D ResNet, a bi-directional long short-term memory with residual networks, efficiently processes sign language’s sequential structure; Graph Convolutional Attention-Residual Networks (GCAR) use hand skeleton data and spatial relationships to achieve remarkable accuracy on large datasets, while Sequence-to-Sequence Video-to-Text (S2VT) models with Tensor Train Decomposition (TTD) provide efficient computation, especially for complex languages like Chinese Sign Language. Together, these developments are opening the door for transportable, scalable, and real-time SLR systems, which should improve accessibility and communication support for the deaf community.