无线电工程2025，Vol.55Issue(5)：959-965,7.DOI:10.3969/j.issn.1003-3106.2025.05.007
WiFi信号可视化与双流网络融合的手势识别
WiFi Signal Visualization and Dual-stream Network Fusion for Gesture Recognition
摘要
Abstract
Gesture actions can be classified into two categories:large-scale actions for basic human-computer interaction and fine-grained small-scale actions containing complex semantic information,depending on their signal fluctuation ranges.The existing WiFi gesture recognition systems not only rely on multiple independent network models when facing actions of different scales,but also need to deploy multiple transceivers in strictly restricted positions.A system combining WiFi signal visualization with a dual-stream network is proposed,which incorporates a Cross-modal Interaction Module(CIM)and a Spectral Feature Transformation(SFT)module.The proposed system adapts to varied transceiver positions and gesture scales,enabling single-input single-output scenarios.On the Widar3 dataset,the system outperforms other methods under identical conditions.In single-input single-output scenarios,for datasets containing both large-scale and fine-grained small-scale gesture actions,the system achieves an average recognition accuracy of 94.85％with fixed transceiver positions and maintains 91.26％accuracy under dynamic position conditions.关键词
WiFi信号可视化/双流网络/手势识别/跨模态交互/谱特征转移Key words
WiFi signal visualization/dual-stream network/gesture recognition/cross-modal interaction/spectral feature trans-formation分类
电子信息工程引用本文复制引用
丁昊,袁江琳,石鸿凌,龚晓龙,孙美静..WiFi信号可视化与双流网络融合的手势识别[J].无线电工程,2025,55(5):959-965,7.基金项目
国家重点研发计划(2020YFC1522900) （2020YFC1522900）
中央高校基本科研业务费专项资金项目(CZY23026) （CZY23026）
中南民族大学学术创新团队经费项目资助(XTZ24006)National Key R&D Program of China(2020YFC1522900) （XTZ24006）
Special Fund Project for Basic Scientific Research of Central Universities(CZY23026) （CZY23026）
Academic Innovation Team Funding for South-Central Minzu University(XTZ24006) （XTZ24006）
