A Low-Complexity Hand Gesture Recognition Framework via Dual mmWave FMCW Radar System

Abstract

In this paper, we propose a novel low-complexity hand gesture recognition framework via a multiple Frequency Modulation Continuous Wave (FMCW) radar-based sensing system. In this considered system, two radars are deployed distributively to acquire motion vectors from different observation perspectives. We first independently extract reflection points of the interested target from different radars by applying the proposed neighboring reflection points detection method after processing the traditional 2-dimensional Fast Fourier Transform (2D-FFT). The obtained sufficient corresponding information of detected reflection points, e.g., distances, velocities, and angle information, can be exploited to synthesize motion velocity vectors to achieve a high signal-to-noise ratio (SNR) performance, which does not require knowledge of the relative position of the two radars. Furthermore, we utilize a long short-term memory (LSTM) network as well as the synthesized motion velocity vectors to classify different gestures, which can achieve a significantly high accuracy of gesture recognition with a 1600-sample data set, e.g., 98.0%. The experimental results also illustrate the robustness of the proposed gesture recognition systems, e.g., changing the environment background and adding new gesture performers.

Adaptive Trajectories’ Constant False Alarm Rate Mirror Filters and Elevation Angle Evaluation for Multiple-Input Multiple-Output Radar-Based Hand Gesture Recognition

“…There are several existing radar-based gesture recognition algorithms. They are based on different hardware architectures, including the continuous wave (CW) radar [10][11][12], the frequency-modulated continuous wave (FMCW) radar [7,8,[13][14][15][16][17][18][19][20], and the multipleinput multiple-output (MIMO) radar [8,21,22]. The CW radar is simple to design and has excellent Doppler sensitivity, but it lacks the resolution of range.…”

Millimeter wave gesture recognition using multi-feature fusion models in complex scenes

“…Visual-based CNN method is the real optical reflection of limbs, but it doesn’t avoid the defects of visualization-based methods. The motion information parameters of FMCW radar are saved as images, and they are used to recognition by CNN so that avoiding the defects of visualization-based methods 43 – 46 .…”