Feasibility of the Hand–Eye Algorithm in Sensor-to-Sensor Calibration of Wearable Devices

In the field of wearable devices, the accuracy of mutual alignment between diverse motion sensors is a critical issue, particularly when manual mounting is involved. This study examines the feasibility of the use of a sensor-to-sensor hand–eye calibration (HEC) for consumer-grade motion tracking sensors, used in wearable applications. The HEC was implemented on three consumer-grade sensors: an inertial-odometer (VIO) vision system, a virtual reality tracker based on structured infrared light, and an inertial measurement unit (IMU). During in-lab calibration, each sensor was mounted with a known relative misalignment onto the end-effector of a cobot used as a reference positioning system. The cobot executed a series of three elementary rotations (180° each) along the roll, pitch, and yaw axes. The algorithm consistently identified the known misalignment for all three sensors with good accuracy. Results revealed that applying the alignment correction derived from the procedure led to a reduced root-mean-square error (RMSE) for the orientation of the VIO (
$\sim 0.9^{\circ }$
) and IMU (
$\sim 1.7^{\circ }$
) systems. The VR tracker showed lower RMSE values (
$\sim 0.8^{\circ }$
) regardless of the calibration application. In addition, a case study involving three healthy subjects performing a motor sit-to-stand task while wearing the sensors was analyzed. The study yielded two key findings: 1) the calibration was capable of estimating accurately the mutual alignment among different motion tracking devices, without affecting the nominal intrinsic accuracy of the sensors, and 2) the calibration was also feasible in the field, by using the subject’s functional movements as calibration data, without compromising accuracy.