Human activity recognition based on millimeter-wave radar

Due to the rapid development of human-computer interaction, human activity recognition has become a hot research topic. Millimeter-wave radar is an advanced device for collecting high-quality data. In this paper, we propose a human activity recognition system utilizing millimeter-wave radar. Firstly, we collect spectrograms caused by human motion using millimeter-wave radar. Then, we adopt residual neural networks to process data and complete the classification task. Based on our proposed system, we classify four activities and achieve an average recognition accuracy of 91.75%. In the end, we discuss some factors that influence the experimental results, and provide the future development of this research field.