Multiview Representation Learning for Human Activity Recognition