Speech2Action: Cross-Modal Supervision for Action Recognition | IEEE Conference Publication | IEEE Xplore