Table 10.

Confusion matrix of $experimen{t}_4$ for the user-general model. Rotation of the bracelet $=YES$ (on training and test set), orientation correction $=YES$. Best result with four synchronization gestures $(sync=4)$ to select the maximum average energy sensor $s_x$. Classification accuracy $=81.22\%$.

	Targets						Predictions Count (Precision%)
	waveIn	waveOut	Fist	Open	Pinch	noGesture
waveIn	6651	138	336	302	725	217	8369 79.47%
waveOut	207	6550	83	416	264	139	7659 85.52%
fist	359	29	6614	262	656	160	8080 81.86%
open	147	849	391	6165	1034	162	8748 70.47%
pinch	95	30	126	295	4424	95	5065 87.34%
noGesture	191	54	100	210	547	6877	7979 86.19%
Targets Count (Sensitivity%)	7650 86.94%	7650 85.62%	7650 86.46%	7650 80.59%	7650 57.83%	7650 89.9%	45,900 81.22%