Human behavior recognition with FMCW radar based on dual attention mechanism
To improve the classification accuracy and generalization performance of frequency modulated continuous wave(FMCW)radar-based human behavior recognition,a feature fusion method based on dual attention mechanism was proposed.Through threshold setting,the effective spectrum in the range-time spectrogram and micro-Doppler spectrogram was extracted and spliced,and then sent to AlexNet and VGG16 neural networks to extract features.Subsequently,spatial attention and enhanced channel attention module were introduced to discard redundant information,enhancing focus on crucial details to gather more interesting features for feature-fusion classification.The experimental results demonstrate that this method attains an impressive average recognition accuracy of 97.0％for six daily human behaviors.
frequency modulated continuous wave(FMCW)radarfeature fusionchannel attentionspatial attentionhuman behavior recognition
万方数据
维普
