Micro Gesture Recognition with Multi-Dimensional Feature Fusion and CQ-MobileNetV3 Using FMCW Radar

Wei Xue ^1,2,3, Rui Wang ^1,2,3, Jianyun Wei ^1,2,3

⁰School of Automation, China University of Geosciences, Wuhan 430074, China.

Sensors (basel, Switzerland) +

|

November 27, 2025

View abstract on PubMed

Summary

This study introduces a novel radar-based micro gesture recognition system using multi-dimensional feature fusion and a lightweight neural network. The method achieves high accuracy for contactless human-computer interaction with low computational costs.

Area Of Science

Human-Computer Interaction
Radar Signal Processing
Deep Learning

Background

Contactless human-computer interaction (HCI) is advancing, with radar-based gesture recognition gaining traction.
Recognizing micro gestures presents challenges due to small motion amplitudes and short durations, impacting feature extraction and accuracy.
Balancing high recognition accuracy with low computational and storage demands is crucial for practical applications.

Purpose Of The Study

To develop an efficient and accurate micro gesture recognition method for radar-based HCI.
To address the challenges of feature extraction and computational complexity in micro gesture recognition.
To propose a novel approach combining multi-dimensional feature fusion with a lightweight deep learning network.

Main Methods

Constructing range-time, velocity-time, and angle-time maps from radar data.
Refining these maps through normalization and adaptive filtering, then fusing them into a comprehensive range-velocity-angle-time map.
Designing and implementing a lightweight CQ-MobileNetV3 network, optimizing MobileNetV3 with attention modules (CBAM, SA) for enhanced accuracy and reduced complexity.

Main Results

The proposed method achieved a 97.16% recognition accuracy for 14 distinct micro gestures.
The CQ-MobileNetV3 network demonstrated remarkable efficiency with only 0.207 M parameters and 0.027 GFLOPs computational complexity.
Experimental results using a 77 GHz FMCW radar validated the superior performance compared to other deep neural networks.

Conclusions

The developed multi-dimensional feature fusion and CQ-MobileNetV3 network offer a highly accurate and computationally efficient solution for radar-based micro gesture recognition.
This approach effectively captures micro gesture characteristics, overcoming limitations in feature extraction and enabling practical contactless HCI.
The study highlights the potential of lightweight deep learning models and advanced feature fusion techniques in advancing radar-based interaction systems.

Keywords:

deep neural network frequency-modulated continuous wave (FMCW) radar micro gesture recognition multi-dimensional feature fusion

Related Concept Videos

IR Frequency Region: Fingerprint Region

IR spectra are divided into two main regions: the diagnostic region and the fingerprint region. The diagnostic region of the spectrum lies above 1500 cm−1. The absorptions resulting from single-bond vibrations of the N–H, C–H, and O–H stretch at higher wavenumbers and appear on the left side of the spectrum. The stretching absorptions of the C≡C and C≡N occur between 2100–2300 cm−1. In contrast, those arising from stretching absorptions of the...