Table 4.

Leave-one-subject-out (LOSO) evaluation with a Random forest classifier with 100 estimators for different data representation and window overlapping.

Data Representation	Overlap	Sensitivity	Specificity	AUC	EER (%)
Mazilu	0%	0.824	0.821	0.900	18.0
	50%	0.829	0.826	0.903	17.4
	75%	0.826	0.825	0.904	17.5
MFCCS	0%	0.824	0.824	0.909	17.6
	50%	0.831	0.827	0.911	17.3
	75%	0.835	0.834	0.913	16.7
FFT	0%	0.842	0.839	0.916	16.1
	50%	0.841	0.839	0.919	16.1
	75%	0.849	0.847	0.921	15.3