UTD-MHAD: A multimodal dataset for human action recognition utilizing a depth camera and a wearable inertial sensor | Synapse