CNN-Based Personal Identification System Using Resting State Electroencephalography

Table 4

Comparison with other EEG-based identification systems using PhysioNet dataset.

Reports	Approach	Session	Subjects	Channels	Sampling rate (Hz)	Window length (s)	Stride (s)	Rank-1 (%)	EER (%)
[22]	PSD and spectral coherence	EO and EC	108	56	160	10	—	100	—
[23]	Eigenvector	EO and EC	109	64	160	12	—	96.90	4.40
[24]	Eigenvector	EO and EC	109	64	160	12	—	—	1.42
[37]	Wavelet coefficients	T1-T4	108	9	160	30	15	99.00	4.50
[26]	CNN	EO and EC	109	64	160	12	0.125	—	0.19
[25]	Eigenvector	EO and EC	109	56	—	12	—	98.93	0.73
[27]	1D-Conv. LSTM	EO and EC, T1-T4	109	16	160	1	—	99.58	0.41
Proposed	CNN	EO and EC	109	14	160	0.5	0.25	99.32 ± 0.60	0.18 ± 0.15

The key parameters used in the proposed system are highlighted in bold.