表 1. Performance comparison of classification accuracy (%) for meditation and resting states.

多个模型在冥想静息状态脑电分类任务的准确率（%）比较

实验设置	方法	准确率	精确率	召回率	F1分数
注：粗体数字表示当前实验设置下的最优结果，各指标数值越大结果越好
被试依赖	EEGT	64.75 ± 9.24	68.02 ± 12.56	62.15 ± 20.89	61.60 ± 17.43
	MLP	66.82 ± 13.46	66.84 ± 13.06	68.77 ± 18.84	66.79 ± 14.37
	SVM	68.32 ± 15.12	67.45 ± 15.60	72.29 ± 19.07	69.12 ± 15.72
跨被试	DANN（成对，平均）	53.04 ± 5.96	53.16 ± 6.30	53.80 ± 14.04	50.99 ± 9.10
	MLP	55.86 ± 7.48	45.53 ± 25.53	46.46 ± 31.60	44.16 ± 25.60
	SVM	57.22 ± 9.59	59.33 ± 11.27	46.02 ± 26.81	47.43 ± 20.57
	EEGT	58.00 ± 7.18	56.30 ± 15.14	54.17 ± 24.72	53.00 ± 17.56
	DANN	60.63 ± 8.53	61.96 ± 7.75	56.57 ± 21.25	57.24 ± 13.02
	MFSAN	65.97 ± 8.59	67.62 ± 10.46	69.00 ± 20.30	65.69 ± 11.30
有标定的跨被试	DANN（成对，堆叠集成）	59.32 ± 13.20	58.61 ± 12.90	64.12 ± 21.14	60.12 ± 14.97
	DANN（微调）	65.87 ± 10.41	62.13 ± 17.82	67.92 ± 22.91	64.14 ± 18.66
	DANN（微调，成对，平均）	66.35 ± 13.72	65.95 ± 13.14	67.81 ± 19.54	65.83 ± 15.24
	MLP（微调）	66.98 ± 11.62	66.07 ± 10.84	72.18 ± 20.72	67.56 ± 13.65
	MFSAN+	71.71 ± 10.10	71.34 ± 9.88	73.84 ± 13.71	72.02 ± 10.47
	提出的方法CMAAN	73.09 ± 11.19	73.38 ± 11.91	76.48 ± 12.32	74.14 ± 9.71