Towards Human Machine Interaction: Exploring RF based Gesture Recognition

Gestures have been an integral part of communication and have been used for centuries to communicate effectively with other humans regardless of their culture. Using these natural responses to control and interact with a device is called gesture recognition. Gesture recognition uses various sensors such as cameras, radar, or other sensors to interpret and accept the gesture as a command. This work investigates gesture recognition utilizing radio frequency (RF) signals to identify the unique spatial information of the different gestures caused by the internal displacements of the wrist's tissues and muscles. Experiments were conducted to characterize the sensitivity of the single and dual RF probe to the internal variations of the wrist's tissues while performing a gesture in three frequency bands (C, X, Ku). A simple simulation model of the human wrist was designed in Ansys High Frequency Structure Simulator (HFSS), and the backscattering from the RF probe was compared with the measurements. Additionally, this work characterizes two prototype sensors based on Substrate Integrated Waveguide Leaky Wave Antenna (SIW-LWA), which was developed to work using the same principle. The prototype sensors are designed to operate in the 8-18 GHz frequency range and radiate into the human wrist. This work investigates the sensor's sensitivity to the internal variations of the wrist's tissues while performing a gesture. Furthermore, this work compares the accuracy of the prototype sensors, and the gesture classification is evaluated using the KNN classifier and RMSE.