Gesture recognition for millimeter wave radar based on LocalPVT
Shuo Zhao, Zhaocheng Wang, Hailong Kang, Ruonan Wang, Guangxuan Hu, Guang Zhang · IET conference proceedings. · 2024
Benefiting from the increasing development of deep learning, the deep learning-based millimeter wave radar gesture recognition method has been widely used in military and civil fields. Nonetheless, it faces problems such as scarcity of public datasets and lack of global information in traditional networks. To solve the above problems, a millimeter wave radar gesture recognition method based on LocalPVT is proposed in this paper. The proposed method consists of two stages: constructing the millimeter wave radar gesture recognition (MWRGR) dataset and implementing gesture recognition. In the first stage, we utilize a 77 GHz millimeter-wave radar to capture the echoes of gestures and perform preprocessing to generate the Micro-Doppler time spectrograms of dynamic gestures to facilitate the construction of the dataset. In the second stage, we propose a LocalPVT network for gesture recognition tasks that can not only capture global but also local information. We improve the global and local perception of time spectrograms of dynamic gestures in the LocalPVT network by introducing locality in the Transformer, which has only global perception. Experimental results based on the constructed MWRGR dataset demonstrate the effectiveness of the proposed method.