Automatic sign language analysis: a survey and the future beyond lexical meaning

Area of Science:

• Computational Linguistics
• Computer Vision
• Human-Computer Interaction
• Artificial Intelligence

Background:

• Current automatic sign language recognition primarily focuses on lexical sign gestures in continuous signing.
• Existing research often neglects nonmanual signals and grammatical variations crucial for complete sign language understanding.
• The scalability of algorithms for large vocabularies is a key challenge in sign language processing.

Purpose of the Study:

• To survey data acquisition, feature extraction, and classification methods for sign language gesture analysis.
• To examine the analysis of nonmanual signals and their integration with hand gestures in sign language.
• To discuss challenges and future directions for robust sign recognition systems dealing with natural signing.

Main Methods:

• Review of existing literature on data acquisition, feature extraction, and classification techniques for sign language.
• Analysis of methods for modeling transitions between signs, inflectional processes, and signer adaptation.
• Examination of approaches for analyzing nonmanual signals and integrating them with manual sign gestures.

Main Results:

• Identified limitations in current sign language recognition systems that primarily focus on lexical signs.
• Highlighted the importance and challenges of incorporating nonmanual signals and grammatical variations.
• Discussed progress and remaining challenges in achieving natural sign recognition for native signers.

Conclusions:

• Successful sign language recognition requires a comprehensive approach that includes nonmanual signals and grammatical features.
• Further research is needed to develop robust methods for analyzing natural, continuous signing by native users.
• Advancements in sign recognition can benefit related fields such as linguistics and human-computer interaction.