TABLE 2.

Performance evaluation results for various data augmentation strategies, as conducted on subjects with high MI proficiency. The table presents validation set accuracy Ac and the accuracy improvement ΔAc obtained for TP feature classification.

DA strategy	Augmented set Ac	Small set ΔAc = Ac − 0.836	Full set ΔAc = Ac − 0.882
small (33%) + 33% NFT	0.858	0.022, p = 0.013	−0.024, p = 0.169
small (33%) + 67% NFT	0.857	0.021, p = 0.003	−0.025, p = 0.134
small (33%) + 167% NFT	0.853	0.017, p = 0.012	−0.029, p = 0.092
small (33%) + 33% NFT jitter [t 0 ]	0.85	0.014, p = 0.003	−0.032, p = 0.06
small (33%) + 67% NFT jitter [t 0 ]	0.85	0.014, p = 0.017	−0.032, p = 0.07
small (33%) + 167% NFT jitter [t 0 ]	0.848	0.012, p = 0.102	−0.034, p = 0.041
small (33%) + 67% noise	0.847	0.011, p = 0.327	−0.035, p = 0.114