Would a thermal sensor improve arm motion classification accuracy of a single wrist-mounted inertial device?

Table 3 Model performance with personalized classification model

Model	Average accuracy, (%)	Average evaluation time to classify 24 motions, (s)
SVM	90.61	4.06
KNN	74.15	0.04
LDA	93.55	0.28
Classification trees	82.08	0.65

Average accuracy is calculated upon the 11 participants’ classification accuracies. Average evaluation time is calculated by taking the average time required to classify a set of 24 motions across the participants

ISSN: 1475-925X