Walk Identification using a smart carpet and Mel-Frequency Cepstral Coefficient (MFCC) features.

Area of Science:

• Engineering
• Computer Science
• Gerontology

Background:

• Independent living for elders is a growing concern.
• Existing activity monitoring systems can be intrusive or lack accuracy.
• There is a need for unobtrusive, real-time monitoring solutions for elder care.

Purpose of the Study:

• To develop and analyze a novel floor-based sensor system for in-home activity monitoring.
• To assess the system's ability to recognize individuals based on their gait.
• To evaluate the effectiveness of Mel-Frequency Cepstral Coefficients (MFCC) for person identification in activity monitoring.

Main Methods:

• Development of a context-aware, unobtrusive floor-based sensor system.
• In-depth analysis of sensor-generated waveform characteristics (power spectrum, pulse width, signal shape).
• Feature extraction using Mel-Frequency Cepstral Coefficients (MFCC) for classification.
• Evaluation using data from 10 subjects under various environmental conditions.

Main Results:

• The system successfully monitors activities like walking and falling.
• Analysis of analog signal characteristics provided insights into sensor data.
• MFCC features were effective in distinguishing between different individuals.
• An average accuracy of 82% was achieved in person identification.

Conclusions:

• The developed sensor system offers a promising solution for unobtrusive elder activity monitoring.
• Computational analysis of sensor waveforms, particularly MFCC, enables accurate person identification.
• This technology can significantly contribute to supporting the independent living of elderly individuals.