Application of Inertial Measurement Unit (IMU) in Advanced Human Health and Safety Surveillance: A Data Fusion and Machine Learning Approach
MetadataShow full item record
Accurate and reliable quantification of human physical and physiological state using wearable sensing devices is paramount in health monitoring and safety surveillance, which necessitates the measurement of relevant signals from various on-body spots. The unobtrusiveness and economic considerations urge data fusion techniques to provide proper estimations to eliminate the sensors with lower observational power and significance in body status representation. The data of the main sensing unit need to be predictive for the in-control and out-of-control physical states of a human. This significant predictive power enables us to develop the proper monitoring systems for the prediction of critical states prior to any potential safety and health incident. The present Ph.D. dissertation attempted to establish a stepping stone for the development of such a framework. In this regard, Kalman filter was employed to estimate the hip acceleration and trunk posture from a posteriori data of a single inertial measurement unit (IMU) at the ankle and a priori information from Fourier series approximation to relate body kinematics considering the periodic nature of gait. Using the ankle IMU data alone, support vector machine (SVM) classification of the kinematic profile features derived by a template matching pattern recognition technique (1$ Recognizer) was applied on the gait strides segmented through a novel stepwise search-based segmentation algorithm. A non-parametric agglomerative method for multivariate change point detection and time series pattern analysis was examined on the stride trajectory features across a prolonged daily occupational task. The research outcomes for 20 subjects suggested the error rates of 6.5% and 3.12% for hip acceleration and trunk posture estimations, respectively. The single sensor fatigue classification showed 90% accuracy utilizing gait stride features. The prolonged gait monitoring scheme presented promising results in early detection and prevention of fatigue based on the subjective declarations for 15 recruited participants. These research studies proposed a framework for safety surveillance and fatigue monitoring to prevent safety incidents utilizing a minimal set of sensors and data that can be extended to other movement monitoring application in healthcare through the ordinary smart wearable devices.