Figure 8.

Comparison of CNN and SVM in amputee subjects with myoelectric prostheses experience (n = 5) tested with 10 movement subset. (A) Average session-to-session accuracy in different self-recalibration scenario. Repetition 1 of movement was used as the training data, and repetitions 2 to 6 were tested one by one with or without recalibration. (B) Statistical analysis of session-to-session performance. We compare session-to-session difference among the three scenarios, as well as between CNN and SVM. ^* indicates pairwise statistically significant difference (p < 0.05).