A High-Efficiency Fatigued Speech Feature Selection Method for Air Traffic Controllers Based on Improved Compressed Sensing

Table 3

Detection results based on different measurement-matrix construction methods.

Fatigue feature	Classifier	Kernel function	Measurement matrix	Average accuracy	Total time (minutes)
PH	SAE			79.67%	2.35
	SVM	RBF	Gaussian random matrix	81.61%	5.12
			Uncompressed sampling	79.36%	6.38
			ICSCA	85.11%	1.37
SWFF	SAE			82.86%	1.88
	SVM	RBF	Gaussian random matrix	91.52%	4.65
			Uncompressed sampling	89.31%	5.78
			ICSCA	94.25%	1.21