Person Recognition Based on Deep Gait: A Survey.

Area of Science:

• Computer Vision
• Biometrics
• Machine Learning

Background:

• Gait recognition, or walking pattern recognition, is a biometric technique for remote individual identification.
• Deep learning methods have shown promise since 2014 for automatic feature extraction in gait recognition.
• Accurate gait recognition faces challenges from environmental variability, covariate factors, and human body representation.

Purpose of the Study:

• To provide a comprehensive overview of advancements in deep learning for gait recognition.
• To analyze existing gait datasets and evaluate state-of-the-art technique performance.
• To present a taxonomy of deep learning methods and identify their limitations in gait recognition.

Main Methods:

• Literature review of gait recognition datasets and performance analysis of current techniques.
• Development of a taxonomy to categorize and organize deep learning approaches in gait recognition.
• Identification and discussion of limitations inherent in deep learning methods for this task.

Main Results:

• Analysis of various gait datasets and performance benchmarks of leading deep learning techniques.
• A structured taxonomy categorizing deep learning methods applied to gait recognition.
• Identification of fundamental limitations of current deep learning approaches in gait recognition.

Conclusions:

• Deep learning has advanced gait recognition, but significant challenges remain.
• A structured overview and taxonomy aid in understanding the research landscape and limitations.
• Future research should focus on addressing identified challenges to improve gait recognition performance.