Applications of markerless motion capture in gait recognition

Area of Science:

• Biomechanics
• Human Motion Analysis
• Forensic Science

Background:

• Human gait analysis is crucial in forensic science, but traditional methods face limitations.
• Marker-based motion capture is precise but can be cumbersome and time-consuming.
• Developing accurate markerless motion analysis techniques is essential for efficient gait studies.

Purpose of the Study:

• To explore human gait variability.
• To develop and validate precise markerless kinematic parameter estimation methods for forensic gait analysis.
• To assess the accuracy and reliability of a novel markerless motion capture system.

Main Methods:

• Conducted gait studies in a custom-built laboratory using eight synchronized cameras for markerless motion analysis.
• Processed stereovision-based algorithms for accurate 3D participant reconstructions during normal walking.
• Extracted kinematics using manual and automatic methods, comparing automatic results to marker-based motion capture for validation.

Main Results:

• The proposed markerless motion capture method demonstrated precision comparable to marker-based methods in frontal and sagittal planes.
• Manual annotations revealed high discriminatory power of shoulder kinematics and shank length for gait recognition.
• Identified high inter-observer variability in manual annotations, emphasizing the need for consistent expert application.

Conclusions:

• Markerless motion capture offers a reliable basis for gait recognition using kinematic parameters.
• Shoulder kinematics and shank length are significant discriminatory factors in gait analysis.
• New regression equations were developed to correct bias and account for sex differences in pelvis, improving joint center predictions.