Person identification from gait analysis with a depth camera at home

Area of Science:

• Biomedical Engineering
• Computer Science
• Gerontology

Background:

• Existing fall detection and frailty evaluation systems using depth imaging lack individual identification capabilities.
• Personal identification is crucial for ambient home monitoring systems to distinguish between monitored individuals, family members, and caregivers.
• Previous work established algorithms for fall detection, daily activity recognition, and gait analysis using depth data and Hidden Markov Models (HMMs).

Purpose of the Study:

• To develop and evaluate a novel method for identifying individuals based on their gait patterns from depth image sequences.
• To enhance markerless home monitoring systems for the elderly by enabling personal recognition.
• To differentiate between known (monitored individuals, caregivers) and unknown persons in a home environment.

Main Methods:

• Utilizing gait sequences extracted using Hidden Markov Models (HMMs) for activity recognition.
• Assessing person visibility within gait sequences based on sequence likelihood.
• Identifying individuals using their height and gait patterns from fully visible walking sequences.
• Modeling individual gait patterns with person-specific HMMs built from the trajectory of the center of mass.
• Testing the algorithm with both known and unknown individuals.

Main Results:

• The proposed method accurately differentiates between known and unknown individuals.
• The algorithm correctly identifies known individuals based on their unique gait patterns.
• The system demonstrated effectiveness in distinguishing the person of interest among a mix of known and unknown individuals.

Conclusions:

• The developed method provides accurate individual identification using gait analysis from depth images.
• This technology can significantly improve the safety and personalization of home-based elderly monitoring systems.
• The system's ability to identify both known and unknown individuals enhances its applicability in real-world domestic settings.