affiliation not provided to SSRN
Multimodal Deepfake detection, non-critical phoneme-viseme, VA correlation learning, self-supervised learning