User Identification from Gait Analysis Using Multi-Modal Sensors in Smart Insole.

Area of Science:

• Biometrics and Human Identification
• Sensor Technology and Signal Processing
• Machine Learning and Pattern Recognition

Background:

• Gait patterns are unique identifiers for individuals.
• Various sensors (vision, acceleration, pressure) capture gait data.
• Existing methods for gait recognition have limitations.

Purpose of the Study:

• To propose a novel method for individual identification using gait patterns.
• To apply null-space linear discriminant analysis (NLDA) for gait recognition.
• To evaluate the effectiveness of multi-modal sensor data (pressure and acceleration) from smart insoles.

Main Methods:

• Collected time-series pressure and acceleration data using smart insoles during walking.
• Employed null-space linear discriminant analysis (NLDA) for feature extraction and classification.
• Compared identification accuracy using acceleration data, pressure data, and a combination of both.

Main Results:

• The proposed multi-modal approach achieved high identification accuracy (>95%) for 14 participants.
• Combined pressure and acceleration data yielded superior identification performance compared to individual modalities.
• NLDA effectively distinguished individuals based on their unique gait patterns captured by smart insoles.

Conclusions:

• Null-space linear discriminant analysis is a powerful technique for gait-based individual identification.
• Smart insoles with multi-modal sensors offer a promising solution for accurate biometric recognition.
• Gait analysis using combined pressure and acceleration data can significantly enhance identification accuracy.