A Novel Deep Learning Approach for Recognizing Stereotypical Motor Movements within and across Subjects on the Autism Spectrum Disorder

Table 4

Results of experiments 2, 3, and 4 in time and frequency domains on 6 subjects of Study1 and Study2 datasets.

		Study1						Study2
	Experiments	S1	S2	S3	S4	S5	S6	S1	S2	S3	S4	S5	S6	Mean
Time domain	Cross-domain transfer learning	71.66	74.40	66.80	90.69	61.87	92.19	81.35	58.13	35.66	88.44	73.98	_	72.29
	Within-domain transfer learning	75.37	76.44	56.53	91.74	63.37	92.76	84.86	62.97	41.60	93.32	80.55	_	74.50
	Simple CNN	74.88	74.52	71.90	91.56	64.46	93.95	87.48	67.62	50.21	92.61	81.96	_	77.38
Frequency domain	Cross-domain transfer learning	74.50	91.56	43.77	93.11	76.03	94.20	85.16	74.67	66.98	93.66	83.99	_	79.78
	Within-domain transfer learning	90.54	97.22	83.86	95.24	86.19	98.45	92.71	90.49	84.99	97.99	92.22	_	91.81
	Simple CNN	94.40	80.85	60.17	93.32	79.66	94.73	88.59	83.78	66.67	94.77	85.29	_	83.84