Computer Vision-Based Gait Recognition on the Edge: A Survey on Feature Representations, Models, and Architectures

Area of Science:

• Computer Science
• Biometrics
• Artificial Intelligence

Background:

• Computer vision-based gait recognition (CVGR) is a non-invasive biometric technology with potential in healthcare and human-computer interaction.
• Current CVGR systems rely on cloud servers, leading to latency issues.
• Edge computing offers a decentralized approach to reduce response times and enable real-time applications.

Purpose of the Study:

• To review the state-of-the-art in CVGR systems suitable for edge computing.
• To explore gait data acquisition, feature representation, and model architectures for edge deployment.
• To identify limitations and future research directions for edge-based CVGR.

Main Methods:

• Review of current literature on CVGR systems and edge computing.
• Analysis of gait data acquisition modalities and feature representations.
• Examination of machine learning models and architectures for edge inference.

Main Results:

• Advancements in low-cost microcomputers facilitate edge deployment of CVGR.
• Edge computing reduces latency and enhances real-time capabilities for CVGR.
• The paper provides a comprehensive overview of edge-ready CVGR techniques.

Conclusions:

• Edge computing presents a promising paradigm for CVGR, offering reduced latency and improved real-time performance.
• Further research is needed to address limitations and optimize CVGR systems for edge environments.
• The integration of CVGR with edge computing opens new avenues for biometric and healthcare applications.