Embedding Sequential Information into Spatiotemporal Features for Action Recognition | IEEE Conference Publication | IEEE Xplore