SF-Net: Structured Feature Network for Continuous Sign Language Recognition

TL;DR

SF-Net is a novel structured feature learning framework for continuous sign language recognition that effectively captures multi-level semantic information and outperforms previous methods without requiring pre-training.

Contribution

The paper introduces SF-Net, a new end-to-end trainable model that encodes features at multiple semantic levels for improved sign language recognition.

Findings

Outperforms previous sequence supervision methods in accuracy

Effective in learning multi-level semantic features

Demonstrates strong adaptability across datasets

Abstract

Continuous sign language recognition (SLR) aims to translate a signing sequence into a sentence. It is very challenging as sign language is rich in vocabulary, while many among them contain similar gestures and motions. Moreover, it is weakly supervised as the alignment of signing glosses is not available. In this paper, we propose Structured Feature Network (SF-Net) to address these challenges by effectively learn multiple levels of semantic information in the data. The proposed SF-Net extracts features in a structured manner and gradually encodes information at the frame level, the gloss level and the sentence level into the feature representation. The proposed SF-Net can be trained end-to-end without the help of other models or pre-training. We tested the proposed SF-Net on two large scale public SLR datasets collected from different continuous SLR scenarios. Results show that the proposed SF-Net clearly outperforms previous sequence level supervision based methods in terms of both accuracy and adaptability.