An Attention Self-supervised Contrastive Learning based Three-stage Model for Hand Shape Feature Representation in Cued Speech

TL;DR

This paper introduces a three-stage model combining self-supervised contrastive learning, fine-tuning with limited labeled data, and sequential feature learning for hand shape recognition in Cued Speech, achieving high accuracy and dataset expansion.

Contribution

It proposes a novel multi-stage approach integrating self-supervised contrastive learning and sequential modeling for improved CS hand shape feature extraction.

Findings

Achieved over 90% accuracy in hand shape recognition.

Improved CS phoneme recognition correctness by over 8-10%.

Built a new British English CS dataset with 5 native speakers.

Abstract

Cued Speech (CS) is a communication system for deaf people or hearing impaired people, in which a speaker uses it to aid a lipreader in phonetic level by clarifying potentially ambiguous mouth movements with hand shape and positions. Feature extraction of multi-modal CS is a key step in CS recognition. Recent supervised deep learning based methods suffer from noisy CS data annotations especially for hand shape modality. In this work, we first propose a self-supervised contrastive learning method to learn the feature representation of image without using labels. Secondly, a small amount of manually annotated CS data are used to fine-tune the first module. Thirdly, we present a module, which combines Bi-LSTM and self-attention networks to further learn sequential features with temporal and contextual information. Besides, to enlarge the volume and the diversity of the current limited CS datasets, we build a new British English dataset containing 5 native CS speakers. Evaluation results on both French and British English datasets show that our model achieves over 90% accuracy in hand shape recognition. Significant improvements of 8.75% (for French) and 10.09% (for British English) are achieved in CS phoneme recognition correctness compared with the state-of-the-art.