Table 1. Classification results for different PSD-based feature selection methods: REF, ReliefF, PCA, and mRMR with different classifiers: DNN, KNN, RF, and SV.

Feature selection	Classifiers	Accuracy (%)	Recall (%)	Precision (%)	F-measure (%)
RFE	LDA	94	94	95	94
	SVM	95	95	95	95
	RF	100	100	100	100
	DNN	98	98	98	98
	KNN	90	90	90	90
ReliefF	LDA	95	95	95	95
	SVM	97	97	97	97
	RF	100	100	100	100
	DNN	99	99	99	99
	KNN	94	94	94	94
PCA	LDA	55	55	30	39
	SVM	81	81	83	82
	RF	82	82	84	83
	DNN	90	90	90	90
	KNN	81	81	82	81
mRMR	LDA	95	95	96	95
	SVM	97	97	97	97
	RF	100	100	100	100
	DNN	99	99	99	99
	KNN	94	94	94	94
Importance (10 features)	LDA	93	93	94	93
	SVM	98	98	98	98
	RF	100	100	100	100
	DNN	97	97	97	97
	KNN	94	94	94	94
Importance (30 features)	LDA	94	94	94	94
	SVM	96	96	96	96
	RF	97	97	97	97
	DNN	98	98	98	98
	KNN	93	93	93	93
None	LDA	70	70	71	70
	SVM	86	86	87	86
	RF	93	93	93	93
	DNN	93	93	93	93
	KNN	78	78	79	78