Predicting body movements for person identification under different walking conditions

Area of Science:

• Biometrics
• Forensic Science
• Human Motion Analysis

Background:

• Human walking motion contains unique biometric information.
• Gait analysis is crucial in forensic science for identification.
• Identifying individuals from gait data under varying conditions presents challenges.

Purpose of the Study:

• To develop and validate methods for personal identification using gait.
• To assess the effectiveness of gait identification when walking conditions change.
• To establish a functional relationship between gait vectors under normal and altered conditions.

Main Methods:

• Collected 3D motion capture data of 28 subjects walking normally and with a 5% body weight tote bag.
• Generated gait vectors from 17 anatomical landmarks over two gait cycles.
• Developed linear transformation and partial least squares regression methods to relate gait vectors.

Main Results:

• Achieved high prediction scores: 96.4% for linear transformations and 95.0% for partial least squares regression.
• Successfully predicted tote-bag gait vectors from normal gait vectors.
• Demonstrated the robustness of gait identification across different walking conditions.

Conclusions:

• Personal identification based on walking motion is feasible, even with variations in conditions.
• The developed linear transformation methods show significant potential for forensic gait science.
• This research supports the use of gait biometrics in real-world identification scenarios.