Table 3 Validation set accuracy for different window lengths and overlapping, best result displayed in bold.

From: Empowering stroke recovery with upper limb rehabilitation monitoring using TinyML based heterogeneous classifiers

Window size (ms)

Stride (ms)

Stride: WindowSize Ratio (%)

Accuracy (%)

Loss

4000

1000

25

92.2

0.45

3600

900

25

87.0

0.58

3200

800

25

84.4

0.47

2800

700

25

85.1

0.69

4000

2000

50

90.4

0.26

3600

1800

50

93.0

0.2

3200

1600

50

83.9

0.51

2800

1400

50

84.6

0.48