Application of accelerometer-based gait recognition to adjuvant clinical gait analysis

Area of Science:

• Biometric technology
• Mobile health
• Wearable sensors

Background:

• Gait recognition leverages unique and stable gait patterns from wearable accelerometers for user identification.
• This technology is particularly suited for mobile environments, offering a non-intrusive biometric solution.

Purpose of the Study:

• To develop analytical models for distinguishing normal and pathological gait functions.
• To evaluate gait features for reliable identity recognition and functional analysis.

Main Methods:

• Comparison of various hybrid filter methods by combining four classical filtering techniques.
• Evaluation of feature matching using six classical distance discrimination functions.
• Estimation of gait cycle abnormality impact using standard deviation.

Main Results:

• The study confirms the inherent stability and invariance of gait periodic patterns.
• Optimal recognition rate of 96.67% achieved with MF hybrid filter and Correlation distance on small samples.
• Minimal time consumption of 0.038 seconds for analysis, demonstrating efficiency.

Conclusions:

• Provides evidence for selecting optimal filter and discrimination methods for efficient gait feature extraction.
• Suggests approaches for analyzing clinical gait patterns, aiding future research and applications.