Figure 4.

Classifier performance in differently grouped subsets to be left out in the cross-validation (CV). In within-subject CV, SVM performs considerably better than LSTM. Compared to across-subject analysis (leave-one-subject-out CV), both SVM and LSTM classification results are only slightly affected by a smaller amount of training when using two-fold CV (regardless of the grouping characteristics, i.e., matched hand size or age). In addition, grouping participants in two hand size groups and two age groups revealed similar DAs compared to two hand size/age-matched groups using the same amount of training data. Note that the chance level is at 16.6%.