Distilling Cross-Temporal Contexts for Continuous Sign Language Recognition

Abstract

Continuous sign language recognition (CSLR) aims to recognize glosses in a sign language video. State-of-the-art methods typically have two modules, a spatial perception module and a temporal aggregation module, which are jointly learned end-to-end. Existing results in [9, 20, 25, 36] have indicated that, as the frontal component of the over-all model, the spatial perception module used for spatial feature extraction tends to be insufficiently trained. In this paper, we first conduct empirical studies and show that a shallow temporal aggregation module allows more thor-ough training of the spatial perception module. However, a shallow temporal aggregation module cannot well capture both local and global temporal context information in sign language. To address this dilemma, we propose a cross-temporal context aggregation (CTCA) model. Specifically, we build a dual-path network that contains two branches for perceptions of local temporal context and global temporal context. We further design a cross-context knowledge distil-lation learning objective to aggregate the two types of con-text and the linguistic prior. The knowledge distillation en-ables the resultant one-branch temporal aggregation mod-ule to perceive local-global temporal and semantic context. This shallow temporal perception module structure facili-tates spatial perception module learning. Extensive exper-iments on challenging CSLR benchmarks demonstrate that our method outperforms all state-of-the-art methods.