Hand Gesture Controlled Mobile Robot Navigation Guidance Using Constant-Q Transform

With the rapid development of collaborative and assistive robotics over the past few decades, robotic systems have been integrated into day-to-day domestic and industrial applications largely. Hand gesture-guided approaches offer quite an effective, seamless, and intuitive interface of human-robot interaction (HRI). Monocular vision-based systems provide hassle-free interfaces for gesture data collection from human participants but at the cost of several photometric challenges in the working space. In this paper, constant-Q transform (CQT) has been utilized for useful feature extraction from the hand gesture dataset, especially under challenging environments such as inadequate illumination levels. An exhaustive study of different components and their combinations in the CQT-feature space has been conducted. This demonstrates the effectiveness of the same in extracting reliable features from the photometrically degraded vision sensor data and consequently its successful implementation into real-world HRI environments.