virtual-augmentation

This paper proposes a coverage-aware virtual IMU augmentation framework to address data scarcity in human activity recognition, generating and selecting synthetic sensor data to improve model performance with limited labeled data.