Hand Gesture Recognition based on Millimeter-Wave Radar using iFormer

Gesture recognition is an important way of human-computer interaction, and gesture recognition using millimetre wave radar has a wider range of application scenarios. Deep learning is widely used in millimetre wave radar gesture recognition, but the features extracted by existing deep learning methods are not sufficient. To address this problem, this paper applies Inception Transformer(iFormer) network to gesture recognition to extract more comprehensive features and improve the accuracy of gesture recognition. Firstly, the radar data of six gestures, namely left wave, right wave, left and right wave, push, pull, push and pull, are processed into micro-Doppler spectrograms using short-time Fourier transform, and then the high-dimensional and low-dimensional features of the spectrograms are extracted and classified using the iFormer network, and the recognition accuracy of this mothod can be up to 98.87%. In addition, a comparison is made with five methods based on AlexNet, Visual Geometry Group11(VGG11), GoogLeNet, Residual Network34(ResNet34) and Vision Transformer(VIT), and the experimental results show that the method proposed in this paper performs excellently in terms of classification accuracy and stability.