Developing and validating an ambulatory system for evaluating postures and upper limb movements under real working conditions

Outline of reasons and objectives
The objective of this study was to develop and validate an ambulatory system for evaluating postures and movement of the upper limb, on the basis of inertial sensors. The use of such systems requires kinematic modelling of the upper limb, associated with such sensors (the modelling depending on the selected calibration method), and validation of the kinematic output data (angles of the joints and positions of the segments), in order to develop a reliable system.

Approach
An experimental protocol was conducted in the laboratory. 10 subjects were equipped with 5 inertial sensors (hand, forearm, arm, shoulder blade, thorax) and they performed a calibration session aimed at building the correspondence between the segment axes and the technical axes of the inertial sensors, and a test session making it possible to validate the best calibration procedure. The calibration session was made up of 7 posture holds/functional movements and the test session was made up of 8 different posture holds/functional movements. For each subject, all of these sessions were repeated with 3 different experimenters. Kinematic modelling of the upper limb made it possible to compare the angular data (wrist, elbow, and shoulder) and the position of the hand in space, as calculated from the inertial sensors, with the values obtained by means of an optoelectronics system, considered as a reference system.
The angular value comparison criteria concerned accuracy (relative to the reference), fidelity (reproducibility), and interpretability (compliance with physiological limits of the joints). Positioning error of the hand was characterised by calculating the accuracy.

Main results
The results showed that a calibration combining a static posture and a functional flexion/extension movement of the elbow was more appropriate in view of the various comparison criteria chosen. The following were observed, in particular: a joint angle fidelity of about 5-10°, and a positioning error of the hand in space of 7-15 cm depending on the various test conditions considered. These results are consistent with the existing scientific literature on the subject.
In addition, it was shown that, by following the experimental procedure rigorously (using written instructions), it was possible to limit the experimenter effect (observer-expectancy effect). Finally, a graphical interface was developed, offering viewing of the joint angles of the wrist, of the elbow, and of the shoulder, and a realistic 3D representation of the movement of the upper limb.

Discussion
This study, which was the subject of a PhD thesis defended on 8 December 2015, contributed to making available an ambulatory system for evaluating postures and upper limb movements with a view to assessing upper limb MSD risks in working environments. Based on the use of inertial sensors, this system will ultimately make it possible to avoid having to implement an optoelectronic system and having to cope with the lack of accuracy of conventionally used goniometers. Recommendations on the anatomic calibration procedure were formulated on the basis of the criteria of accuracy, fidelity, and interpretability. The limits of current kinematic modelling were identified and prospects for improvement were proposed. Those prospects concern taking account of an inertial sensor placed on the shoulder blade, and characterising the system in a magnetically disturbed environment. The results of this study have been the subject of several presentations at national and international conferences, and were published in the peer-reviewed
international journal “Sensors” published in July 2015.