Inertial motion capture for ambulatory analysis of human movement and balance

Sammanfattning: Inertial sensors (accelerometers and gyroscopes) are ubiquitous in today’s society, where they can be found in many of our everyday mobile devices. These sensors are capable of recording the movement of the device, and by extension, the movement of humans carrying or interacting with the device. Human motion capture is frequently used for medical purposes to assess individual balance performance and movement disorders. Accurate and objective assessment is important in the design of individualized interventions and rehabilitation strategies.The increasing availability of inertial sensors, combined with their mobility and low cost, has made inertial motion capture highly relevant as a more accessible alternative to the laboratory based gold standard. However, mobile solutions need to be adopted for plug-and-play use with the end user in mind. Methods that automatically calibrate the sensors, and methods that detect and record relevant motions are required.This thesis contributes to the development of human inertial motion capture as a plug-and-play technology. A method for accelerometer calibration, which allows for compensation of systematic sensor errors, is proposed. The method fuses accelerometer and gyroscope data to allow dynamic rotation of the sensor during the calibration procedure. Other proposed methods handles sensor-to-segment calibration in a biomechanical model. The position of a joint center and the direction of a hinge joint’s rotation axis, are identified in each sensor’s intrinsic coordinate system. This is done by fitting recorded motions to the kinematic constraints of the underlying biomechanical model. The methods are evaluated on real sensor data collected from mechanical joint systems that mimics human limbs.The state of the current knowledge regarding objective human balance assessment isstudied in the form of a systematic review, that includes methods for modeling and identifying neuromuscular control of human balance. A similar modeling framework is then applied to identify feedback controllers in physical and artificial (simulated) systems. Finally, inertial sensors are applied in tremor quantification in Parkinson’s disease (PD) and essential tremor (ET). The method uses only the inertial sensors in a standard smart phone, and is applied on data from human subjects with PD or ET.