Improving Human Intention Prediction Using Data Augmentation
Shengchao Li, Lin Zhang, Xiumin Diao · 2018
One of the crucial challenges in human-robot interaction is how to enable robots to predict human intentions. In this study, we explore how data augmentation technique can contribute to human intention prediction when only limited training data is available. Specifically, we conduct experiments of predicting the intentions of a human throwing a ball towards designated targets. Prediction performances with various data augmentation methods are presented and compared. The experiment results show that prediction accuracy can be improved from 50% to 75%.