Inertial Sensing for Human Motion Analysis: Processing, Technologies, and Applications

Abstract

Human motion has always attracted significant interest and curiosity. In particular, the last two centuries have seen a fast and great development of innovative techniques and technologies for the scientific analysis of human motion. If initially this was mainly due to the large interest in biomedical fields, a growing number of other leading applications has kept this interest alive until today. These applications emerge, for instance, in sport, entertainment, and industrial contexts.

The first motion capture systems, appeared along the nineteenth century, were typically based on optical technologies and their development was profoundly interlaced with the contemporary development of photography and cinematography. Since then, many other different technologies have been employed to develop new motion capture systems, such as (but not limited to) inertial, mechanical, magnetic, and acoustic. In particular, inertial motion capture systems, based on the use of inertial sensors (such as the accelerometer, which measures the acceleration, and the gyroscope, which measures angular velocity), are likely to replace the previous ones and become a standard technology. This is mainly favored by the recent great improvement in the large-scale development of accurate inertial sensors ever cheaper.

When referring to inertial human motion analysis, several application areas are driving current research and development efforts. A tentative list may include, for instance, the following: clinical and home monitoring and/or rehabilitation; ambient assisted living; computer graphics and computer animation; gaming and virtual reality; sport training; pedestrian navigation; and robotics. Furthermore, human motion analysis often implies a transversal investigation of many aspects of human motion, at different levels of abstraction and at different detail depths. For instance, one may just be interested in recognizing and estimating the pose of a person as well as in identifying the activities and/or the gestures that he/she is performing. Furthermore, one may be just interested in analyzing a restricted part of the body rather than focusing on the full body.

Due to this heterogeneity of topics and intents, this thesis does not focus on a specific application or method, but aims at investigating different aspects of inertial human motion analysis, by specifically discussing the corresponding data processing approaches and the involved technologies. Four research areas have been taken into account which correspond to four types of applications: arm posture recognition; activity classification; evaluation of functional motor tasks; and motion reconstruction. In particular, these applications have been chosen in order to cover topics with different levels of abstraction and different detail depths.