FMCW Radar-Based Hand Gesture Recognition Technique Using Frequency-Weighted MTI

Abstract

This paper proposes a signal processing technique that recognizes hand motions using a 60-GHz frequency-modulated continuouswave radar system. In the case of hand motion recognition with a radar, the recognition accuracy decreases owing to a clutter signal generated by the upper body and differences between the hand sizes of different people. To solve this problem, this paper proposes a moving target indicator filter with a weighted frequency for each frequency, which improves hand motion recognition accuracy, removes self-interference clutter signals, and uses a range profile and Doppler profile. Neural networks can recognize hand movements; hence, the proposed five hand gesture recognition techniques exhibited an 87.3% accuracy, regardless of the user.