Table 1: Classification test accuracy results. Test 1 only uses time-domain features. In test 2, only frequency domain features are added. In test 3, time-frequency-domain features are added. In test 4, only the IEMG feature is used. Test 5, only time-domain features are used, and no loads are considered in classification. |
||||||
Type of model |
Kernel function |
Test 1 Accuracy [%] |
Test 2 Accuracy [%] |
Test 3 Accuracy [%] |
Test 4 Accuracy [%] |
Test 5 Accuracy [%] |
DT |
Fine tree |
73.8 |
54.4 |
86.4 |
64 |
88.5 |
DT |
Medium tree |
62.3 |
45.7 |
69.2 |
52 |
79.1 |
DT |
Coarse tree |
38.8 |
30.4 |
43 |
30.8 |
63.3 |
SVM |
Linear SVM |
84.2 |
26.7 |
88.5 |
67.3 |
85.8 |
SVM |
Quadratic SVM |
98.4 |
19.4 |
98.3 |
84.9 |
97.5 |
SVM |
Fine Gaussian |
99.8 |
50.7 |
99.7 |
99.5 |
99.3 |
SVM |
Medium Gaussian |
95.9 |
44 |
96.8 |
90.5 |
94.4 |
SVM |
Coarse Gaussian |
81.2 |
37 |
88 |
75.6 |
86.4 |
KNN |
Fine KNN |
99.8 |
71.3 |
99.8 |
99.8 |
99.8 |
KNN |
Medium KNN |
99.7 |
71.3 |
99.8 |
99.5 |
95.7 |
KNN |
Coarse KNN |
94.4 |
62.6 |
95.7 |
90.2 |
95.7 |
KNN |
Cosine KNN |
99.8 |
37.8 |
99.8 |
99.6 |
99.8 |
KNN |
Weighted KNN |
99.8 |
71.9 |
99.8 |
99.8 |
99.8 |
Ensemble |
Boosted Trees |
66.2 |
47.6 |
83.9 |
56.9 |
84.7 |
Ensemble |
Bagged Trees |
99.8 |
76.9 |
99.8 |
99.8 |
99.8 |
Ensemble |
Subspace Discriminator |
57.1 |
23.6 |
57.1 |
45.3 |
73 |
Ensemble |
RUsBoosted Trees |
63.9 |
44.3 |
83.6 |
57.1 |
78.7 |
Number of classes |
|
14 |
14 |
14 |
14 |
6 |
DTs = Decision trees. SVM = support vector machines. KNN = k-Nearest Neighbours. |