Hand position estimation using inertial sensors

Magneto-Inertial Measurement Units sensors (MIMU) are more and more used in various applications, including ergonomics, as they allow for ambulatory measurements. Despite much interest from the scientific community in this topic, several aspects of research deserve more investigation. Since MIMU only deliver 3D orientation data, the determination of MIMU-based segment positions remains challenging. This work proposes the assessment of MIMU-based hand positioning error. Ten subjects participated in this study. They were equipped with four MIMU, placed on the thorax, right upper arm, right forearm and right hand segments. Each subject repeated five experimental sessions, each consisting in four test conditions: two static poses (rest-pose, T-pose) and two series of movement (maximal shoulder flexion, circular multi-joint movement). The MIMU-based upper limb kinematic chain was obtained by associating information of (1) 3D segment orientations, determined during a static pose calibration, and (2) segment lengths, estimated by regression rules based on the subject height. The hand positioning error was calculated by using an optoelectronic system as reference (REF) and by aligning both MIMU and REF thorax segments together. The root mean square error (RMSE) was extracted for each experimental session and was then averaged over them (N = 50). Results showed a global RMSE in the range of 7 – 15 cm confounding the four test conditions. This range of error is consistent with values mentioned in other studies, performed on a robotic arm or without considering the wrist motion. 3D visualization of the upper limb represents a complementary source of information to joint angles data, especially for shoulder joint, in order to ease the interpretation of postures and movements in a general context of musculoskeletal disorder prevention. If a better precision for upper limb positioning is expected, a more advanced kinematic chain and dedicated calibration steps represent interesting approaches to investigate.