Table 1. Classification results (%)

Feature set		Accuracy	ASD			TD
			Precision	Recall	F1-score	Precision	Recall	F1-score
eGeMAPS	All (88)	88.08	88.12	99.90	93.64	78.57	2.66	5.14
	RFE (70)	88.29	88.62	99.43	93.72	65.31	7.73	13.82
KDSFs	All (29)	91.25	92.23	98.33	95.18	76.85	40.10	52.70
	RFE (19)	90.78	94.55	94.99	94.77	62.50	60.39	61.43
Subset of KDSFs	Pitch+MFCC (18)	91.08	93.05	97.09	95.03	69.37	47.58	56.45
	VQ+MFCC (18)	91.90	95.45	95.32	95.39	66.51	67.15	66.83
	SR+MFCC (21)	92.13	95.40	95.66	95.53	67.98	66.67	67.32

The best performances are in bold. The number in parenthesis denotes the number of features used for training the classifiers.
ASD, autism spectrum disorder; TD, typical development; eGeMAPS, extended Geneva Minimalistic Acoustic Standard Parameter Set; RFE, recursive feature elimination; KDSFs, knowledge-driven speech features; VQ, voice quality; MFCC, Mel-frequency cepstral coefficient; SR, speech rate.