Automatic Identification of Sensor Localization on the Upper Extremity

In this paper, we use 3D data from accelerometers and gyroscopes to identify the segment on which each inertial sensor is attached. An automatic procedure will reduce the time and skills needed to acquire motion capture data with inertial sensors, as well as reduce potential error.

Our objective was to make a computationally low-cost algorithm that can be applied on a variety of models. We propose an algorithm based on features extracted from 3D accelerometer and gyroscope data. Our algorithm does not rely on training data or standard classifiers, and is based only on basic mathematical operations allowing for a fast (0.2s on average) and light procedure that is executable on nearly all platforms.

We first tested our code on a wooden mockup of the upper extremity and subsequently applied it on data from three healthy subjects. Twelve features were selected for analysis on four different movements. Each motion was performed at three different speeds to assess the potential use of the algorithm on patient populations.

The results indicate that this algorithm can be used to identify the segments on which the inertial sensors are located. The only information required prior to execution of the algorithm is the number of segments that are involved in the model.