A Comparison of Approaches for Segmenting the Reaching and Targeting Motion Primitives in Functional Upper Extremity...

Area of Science:

• Biomechanics
• Human-Computer Interaction
• Rehabilitation Engineering

Background:

• Kinematic analysis of human functional upper extremity movement (FUEM) is crucial for health monitoring and rehabilitation.
• Deconstructing FUEM into activities, actions, and primitives is essential for kinematic analysis.
• Current machine learning methods show limited utility for analyzing FUEM primitives like reaching and targeting.

Purpose of the Study:

• To investigate and evaluate five distinct methods for segmenting reaching and targeting motion primitives in FUEM.
• To compare the performance of these segmentation methods across haptics simulation and real-world domains.
• To propose a novel method for automatically identifying potentially incorrect segmentation results.

Main Methods:

• Evaluation of five different segmentation algorithms for reaching and targeting motion primitives.
• Testing across two distinct data domains: haptics simulation and real-world movement capture.
• Development of an automated approach to flag segmentation errors for human review.

Main Results:

• Most evaluated segmentation methods demonstrate reasonable performance, considering current evaluation limitations.
• The proposed automated method effectively identifies potentially incorrect segmentation results.
• Performance consistency was observed between haptics simulation and real-world data.

Conclusions:

• Existing segmentation methods are largely adequate for current FUEM kinematic analysis needs.
• The developed automated review tool can significantly streamline the segmentation validation process.
• This work paves the way for more efficient and scalable kinematic data processing in clinical settings.