Video and inertial (VIMU) motion capture systems for human movement assessment

点击翻译按钮获取中文摘要

For years now, optoelectronic systems have been considered the ground truth for human kinematics assessment. Yet, these tools are expensive, cumbersome, require trained specialists, and provide a restricted volume of acquisition; their use is thus limited to laboratory settings. To overcome these limitations, new approaches for kinematics assessment arose in the past decades, including wearable inertial measurement units (IMU). Composed of accelerometers, gyroscopes and magnetometers, IMU allow movement to be assessed in a variety of environments such as lab, clinic, and home. Although promising, IMU are prone to various reliability issues due to environmental effects, calibration problems, and data fusion algorithms optimization process. Hence, their true potential for human kinematics is currently limited. Human pose can also be estimated from videos, though these methods were mainly developed for animation purposes; their ability to provide accurate and reliable mobility features from a clinical perspective is thus questionable. The complementarity of video and inertial data has been highlighted in different fields such as robotic navigation, but its potential has not truly been investigated for mobility assessment yet. The long-term objective of my research program is to develop new systems and algorithms to allow better capture and understanding of mobility issues. This research proposal aims at developing innovative systems for human kinematics assessment based on a combination of video and inertial data (VIMU). Specifically, two complementary frameworks of applications will be considered: an observer-based VIMU system and a wearable VIMU tool. Both will make use of action sports cameras with embedded IMU. The first approach will tackle the challenges associated with VIMU movement assessment from an observer's standpoint. Thus, dynamic tracking and assessment of a person's motion in various environments will be enabled, including complexed conditions (e.g., flat/inclined walking paths, stair ascent/descent). The second approach considers the development of a wearable VIMU tool for mobility assessment. Explicitly, the VIMU system will track environmental features and combine it with inertial information to determine changes in posture. The wearable approach will first concentrate on semi-controlled conditions, including clinically recognized mobility tests, moving towards free-living objective mobility assessment. Both approaches will tackle technological challenges associated with calibration process (cameras extrinsic/intrinsic parameters, sensors tuning, reference frame identification), human movement assessment algorithms (human recognition, task identification and segmentation, kinematics parameters assessment), and include development of autonomous quality feedback algorithms. This proposal thus addresses the technical issues to enable development of innovative VIMU systems for human mobility assessment, and to democratize their use.