FIG 2 (IMAGE)
Caption
Figure 2 The results of emotion recognition experiments. a-d. The training-testing set splitting method, confusion matrix, macro-averaged ROC curve, and average F1 score for the 8-fold a, LOO b, 4-fold c, and LTO d cross-validation methods. The accuracy difference between the 8-fold and LOO cross-validation was 1.04%, while the difference between the 4-fold and LTO cross-validation was 2.75%. AUC of these four cross-validation methods are 0.93, 0.93, 0.93 and 0.92, respectively. The maximum F1 score difference for the same emotion were 0.07 in both comparative groups. e. The average accuracy of each subject under each cross-validation method.
Credit
OET
Usage Restrictions
Credit must be given to the creator.
License
CC BY