Test–Retest Repeatability of Human Gestures in Manipulation Tasks

Abstract

The importance of performance excellence and operator’s safety is fundamental not only when operators perform repetitive and controlled industrial tasks, but also in case of abrupt gestures due to inattention and unexpected circumstances. Since optical systems work at frequencies that are too low and they are not able to detect gestures as early as possible, combining the use of wearable magneto-inertial measurement units (MIMUs) with the adoption of deep learning techniques can be useful to instruct the machine about human motion. To improve the initial training phase of neural networks for high classification performance, gesture repeatability over time has to be verified. Since the test–retest approach has been poorly applied based on MIMUs signals in a context of human–machine interaction, the aim of this work was to evaluate the repeatability of pick-and-place gestures composed of both normal and abrupt movements. Overall, results demonstrated an excellent test–retest repeatability for normal movements and a fair-to-good test–retest repeatability for abrupt movements. In addition, results suggested important information about the application of deep learning to identify the types of movements: the test showed how to improve reinforcement learning for the identification of onset gestures, whereas the retest allowed for defining the time necessary to retrain the network.