Wavelet-Based Semblance Methods to Enhance the Single-Trial Detection of Event-Related Potentials for a BCI Spelling System

Table 3

Letter accuracy of single-trial detection for the 22 subjects for xDAWN, METS & SEWS-1, and METS & SEWS-2 algorithms.

Subject	METS & SEWS-1	METS & SEWS-2	xDAWN
s1	50.00	49.52	49.05
s2	66.67	64.31	60.78
s3	58.75	59.17	60.42
s4	35.56	35.56	36.19
s5	70.28	71.39	71.11
s6	38.15	40.00	31.85
s7	36.67	34.44	24.44
s8	68.52	72.59	67.78
s9	62.38	62.38	55.24
s10	44.76	43.81	44.76
s11	50.88	48.77	41.40
s12	61.67	58.33	56.11
s13	45.56	46.30	47.78
s14	52.08	52.50	43.75
s15	46.67	43.81	28.10
s16	76.41	78.97	61.54
s17	80.00	78.57	80.00
s18	57.68	58.26	58.55
s19	74.00	74.33	73.67
s20	39.65	39.30	36.49
s21	46.67	47.78	30.28
s22	68.89	70.00	63.33

The best results are highlighted in bold.