A Hybrid Visual Transformer for Efficient Deep Human Activity Recognition | IEEE Conference Publication | IEEE Xplore