Application of Inertial Measurement Units for Advanced Safety Surveillance System using Individualized Sensor Technology (ASSIST): A Data Fusion and Machine Learning Approach

Fatigue in the workplace, as a prevalent health and economic issue, can be controlled by monitoring and timely detection through the utilization of wearable sensors. An accurate detection with the optimal number of sensors through the estimation of whole body kinematics as well as single sensor analysis for the detection of lower extremity muscle fatigue as a root cause of slip-induced falls is important. Kalman filter was used to estimate the hip acceleration and trunk posture in gait by a posteriori data from ankle and a priori data using Fourier series approximation to relate the body kinematics considering the periodic nature of gait. The segmented gait step kinematic data from the single sensor at the ankle using a novel stepwise search-based segmentation algorithm was used for fatigue classification through a template matching pattern recognition technique (1$ Recognizer) and support vector machine (SVM). Considering 20 subject data, the best results showed the error rates of 6.5% and 3.12% for hip acceleration and trunk posture estimations, respectively. In addition, fatigue classification utilizing gait step features resulted in 90% accuracy. This study provides a framework with the use of a minimal set of sensors and data for kinematics estimation and fatigue monitoring.