Portable Brain-Computer Interface for the Intensive Care Unit Patient Communication Using Subject-Dependent SSVEP Identification

Table 3

Our system’s performance versus GMM-MAP-SVM and GMM-MAP-SSVM identification accuracies using 3 different numbers of mixtures.

Subjects	CCA	PSDA	GMM-MAP-SVM			GMM-MAP-SSVM
			4 mixtures	8 mixtures	16 mixtures	4 mixtures	8 mixtures	16 mixtures
Subject 1	83%	67%	92.00%	95.30%	94.40%	98.30%	100.00%	100.00%
Subject 2	77%	83%	93.60%	96.70%	92.10%	96.50%	98.70%	93.20%
Subject 3	40%	26%	91.70%	93.60%	90.60%	90.80%	98.40%	94.90%
Subject 4	59%	19%	89.40%	90.20%	89.70%	87.20%	97.70%	94.20%
Subject 5	67%	41%	92.30%	93.90%	93.80%	93.70%	98.80%	97.60%
Subject 6	55%	35%	90.00%	91.50%	92.50%	92.00%	96.90%	96.60%
Subject 7	62%	29%	90.20%	90.60%	93.80%	96.60%	100.00%	99.70%
Subject 8	71%	39%	86.50%	89.50%	88.70%	87.80%	98.50%	96.50%
Subject 9	47%	17%	89.80%	92.00%	89.30%	92.70%	99.20%	95.80%
Subject 10	61%	17%	87.70%	95.80%	91.90%	95.40%	98.80%	96.30%
Average	62%	37%	90.32%	92.91%	91.60%	93.10%	98.70%	96.40%