Fakecatcher: detection of synthetic portrait videos using biological signals

This invention detects fake portrait videos, such as deepfakes, by analyzing subtle biological signals present in a person's skin. It works by generating 'signal maps' from the video and then uses artificial intelligence, specifically a convolutional neural network or a support vector network, to determine if these signals are spatially coherent and temporally consistent, which indicates authenticity. The core idea is that real videos contain these consistent biological signals, while fakes do not.