Skip to main content
. 2020 Mar 11;20(6):1557. doi: 10.3390/s20061557

Table 3.

Overall accuracy (Acc), specificity (true negative rate (TNR)), sensitivity/recall (true positive rate (TPR)) and precision (positive predictive value (PPV)) values of the posture recognition when using the four different kernel functions considering the trunk–lower limb (T-Ll) and upper body (UpB) datasets in the LL task.

LL Task
1 kg 2 kg 5 kg
SVM kernels Acc TNR TPR PPV Acc TNR TPR PPV Acc TNR TPR PPV
T-Ll Linear 98.7 98.7 98.7 98.7 94.9 94.9 94.9 94.9 98.1 98.7 97.4 98.7
Quadratic 95.5 96.2 94.9 96.1 99.4 98.7 100 98.7 97.4 96.2 98.7 96.3
Cubic 96.8 97.4 96.2 97.4 98.7 98.7 98.7 98.7 96.8 96.2 97.4 96.2
Gaussian 98.1 98.7 97.4 98.7 99.4 100 98.7 100 97.4 96.2 98.7 96.3
UpB Linear 73.7 70.5 76.9 72.3 72.4 67.9 76.9 70.6 71.2 73.1 69.2 72.0
Quadratic 76.9 76.9 76.9 76.9 71.8 70.5 73.1 71.3 67.3 67.9 66.7 67.5
Cubic 62.2 56.4 67.9 60.9 62.8 55.1 70.5 61.1 70.5 69.2 71.8 70.0
Gaussian 69.2 66.7 71.8 68.3 73.1 70.5 75.6 72.0 69.2 69.2 69.2 69.2