A Framework for the Analysis of Biomechanical Loading Using Human Motion Tracking

Abstract

In this work, a feedback system for biomechanical load analysis based on joint angles and execution duration of recognized motions is described. For automatic analysis, the system must be set up by experts for the respective application case. The system can be trained individually to recognize motion sequences and parameterized with critical joint angle ranges and times for the evaluation. The system processes Cartesian joint positions, which can be captured by different types of motion tracking systems. The data processing steps filtering, normalization, feature extraction, classification, and segmentation are described. For classification, a Support Vector Machine with polynomial kernel is used that achieves a recognition accuracy up to 87% for 19 different gymnastic motions. In conclusion, a system with an associated user interface is shown, which is able to assist and analyze the user in motion sequences.