Real-time human motion estimation using biomechanical models and non-linear state-space filters

Area of Science:

• Sports Biomechanics
• Rehabilitation Engineering
• Motion Analysis

Background:

• Real-time computation of human joint angular displacements and derivatives from video is crucial for sports biomechanics and rehabilitation.
• Biofeedback protocols in rehabilitation can significantly benefit from accurate, real-time motion analysis capabilities.

Purpose of the Study:

• To develop and validate a system for real-time computation of angular data and derivatives of human joints using biomechanical models and state-space filters.
• To achieve high accuracy in motion analysis for applications in sports and rehabilitation.

Main Methods:

• Utilized biomechanical models, including kinematic chains and surface envelopes, coupled with state-space filters for real-time computation.
• Employed TV cameras to minimize distances between 2D marker projections and back-projected markers for automatic model configuration updates.
• Applied a multidimensional extension of Stirling's interpolation formula for derivatives of the observation model and developed algorithms for calibration, initialization, and data labeling.

Main Results:

• Achieved high accuracy in real-time computation, with a maximum angular error of less than 1 degree and maximum point reconstruction error of less than 1 mm.
• Demonstrated system reliability and robustness against false matching caused by marker occlusions.
• Reduced orientation artifacts due to skin motion by a factor of 50%.

Conclusions:

• The developed system reliably computes human joint angular data and derivatives in real-time from video sequences.
• The system offers significant improvements in accuracy and robustness for motion analysis in sports biomechanics and rehabilitation engineering.
• The approach effectively minimizes artifacts, enhancing the utility of biofeedback protocols.