Recognizing human actions by fusing spatio-temporal appearance and motion descriptors | IEEE Conference Publication | IEEE Xplore