Convolutional Neural Networks with 3D Input for P300 Identification in Auditory Brain-Computer Interfaces

Table 4

Results for each subject in those models with the highest accuracy for each trial length considering the CNN models with both the 3D and 2D input approach. The subject’s number appears on the first column to the left.

	500 ms		400 ms		300 ms
	3D	2D	3D	2D	3D	2D
(1)	0.880	0.72	0.877	0.736	0.859	0.685
(2)	0.873	0.823	0.866	0.78	0.878	0.744
(3)	0.896	0.774	0.908	0.794	0.859	0.774
(4)	0.828	0.812	0.83	0.766	0.831	0.785
(5)	0.864	0.78	0.851	0.792	0.826	0.697
(6)	0.828	0.808	0.828	0.763	0.825	0.685
(7)	0.881	0.788	0.858	0.752	0.886	0.796
(8)	0.879	0.792	0.904	0.77	0.847	0.721
(9)	0.863	0.736	0.847	0.76	0.853	0.722