Fig. 3: Multivariate neural decoding of natural and deepfake speaker identities.
From: Cortical-striatal brain network distinguishes deepfake from real speaker identity

Multivariate decoding accuracies revealed by group-averaged confusion matrices comparing the frequency of the predicted sound class with the true sound class for a the right NAcc, b,c six subregions of the bilateral AC, d the IFG pars triangularis, and e TP. We anatomically defined region-of-interest (ROI) maps. Colored frames indicate decoding accuracies significantly above chance (n = 24, chance level 25%, one-sampled t-test, p < 0.001, Bonferroni corrected for the number of subregions per ROI).