Researching And Developing One Of The Most Powerful Hand Gesture Recognition System With Short Range Digital Radio Technology( UWB Radar)

This system marks a significant advance in research, standing out as one of the pioneering studies using FMCW MIMO Radar (77-81GHz) for hand gesture recognition.

Our PowerGate Research team recognizes the challenges posed by lighting variations and dust on gesture recognition accuracy. To tackle this, we conducted an in-depth exploration of Frequency-Modulated Continuous-Wave (FMCW) radar, which shows promise for gesture recognition. Unlike traditional cameras, radar isn’t as affected by ambient light or physical obstacles, making it a valuable alternative.
In this study, we outline a method for identifying ten distinct hand gestures by analyzing data collected through Texas Instruments’ radar system. We process the raw radar data using Fast Fourier Transform (FFT) to produce the Range-Doppler Heatmap (RDH). Then, we apply micro-Doppler analysis to extract gesture data based on velocity over time from the RDH.

Additionally, we introduce a lightweight Convolutional Neural Network (CNN) model tailored to capture intricate patterns from the processed data, including velocity, angles, and phase characteristics. Trained on a dataset featuring eight dynamic hand gestures, our CNN model achieves an impressive average accuracy of 98.35% across various gestures.

Date

April 22, 2024

Category

R&D, Researching And Developing, UWB Radar Deep Learning