A Data Augmentation Method for Human Activity Recognition Based on mmWave Radar Point Cloud
Zhiming Wang, Dechen Jiang, Bin Sun, Yong Wang
Abstract
Numerous methods have been proposed to address the issue of insufficient data for human activity recognition based on millimeter wave (mmWave) radar. However, few of these methods have studied data augmentation of mmWave radar point cloud while taking into account the characteristics of mmWave signals. In this letter, a data augmentation method for human activity recognition based on mmWave radar point cloud is proposed. The method includes three operations: distance translation, degree rotation, and velocity simulation, which generate samples with varying distances, angles, and human motion velocities. The experimental results demonstrate that the combination of all three data augmentation operations achieves the best performance. The proposed method effectively improves the model's generalization performance in human activity recognition for different scenarios in the radar field.