Fig. 3
Results of the subject-independent phase dissimilarity model. A: The average accuracy for each subject for the subject-independent model and random classifier that utilizes trials from other subjects as templates, excluding trials from the same subject (grand average accuracy is 74.96%). B: The confusion matrices of the model and random classifier summarize predictions for all audiobooks (i.e., stimuli). The best model’s classification belongs to audiobook #17, with an accuracy of 92.16%. C: Evaluating classification performance across different duration lengths for each subject indicates a decrement in accuracy when the length of the audiobooks decreases. At least 20 s of audiobooks are needed to classify EEG signals above the chance level.
