Sign language recognition from skeletal data using graph and recurrent neural networks

Within this field, Isolated Sign Language Recognition (ISLR) focuses on recognizing single, context-free signs, which can be seen as the building blocks for continuous sign language understanding. While ISLR provides an environment to study the visual and temporal characteristics of signs, it also poses significant challenges due to intra-class variations, such as differences in signing speed, hand orientation, and signer-dependent styles. Addressing these challenges is essential for building reliable models that can generalize across diverse signers and recording conditions. Moreover, the recognition of isolated signs provides a basis for Continuous Sign Language Recognition (CSLR), where the boundaries between consecutive signs are not explicitly defined. ISLR is typically formulated as a classification problem, where each video segment corresponds to a single label representing a sign.