EM Wave-Based Hand Gesture Recognition for Astronauts Using 3D Memristive Neural Network

Abstract

The astronauts' spacesuit helmet is generally fitted with a communications carrier assembly (CCA), which has a critical role in ensuring the safety of the astronauts by enabling clear communication during spacewalks. While on spacewalks, often hand gestures are used to communicate between crew members. In this paper, to automatically recognize the hand gestures, the classification of electromagnetic (EM) waves from a patch antenna placed on the hand of an astronaut is performed using a three-dimensional memristive Artificial Neural Network (3D-ANN). Performance characteristics of Ku-band microstrip patch antennas on glass, PET (Polyethylene terephthalate), and FR4 (Flame retardant-4) substrates are analyzed in this work. In the case of FR4 and glass substrate, copper is deposited as the patch, while graphene is deposited as the patch on the PET substrate. The work is proposed for the space suite of astronauts as an alternative for communications carrier assembly (CCA), and hence simulations and experiments are performed for standalone antenna, standalone antenna on Body model, ON-Body to ON-Body, and ON-Body to OFF-Body scenarios. Four hand gestures are performed and classified using a three-dimensional memristive Artificial Neural Network (3D-ANN) based on Skywater 130 nm PDK (SKY130) for the ON-body to OFF-body scenario with an accuracy of 80%. Variability analysis is also performed in the 3D-ANN classifier.