Analysis of Different Classification Techniques for Two-Class Functional Near-Infrared Spectroscopy-Based Brain-Computer Interface

Table 5

Classification accuracy using SVM among all subjects.

Feature combination	S1	S2	S3	S4	S5	S6	S7
Mean & slope	93.2	95.5	92.5	95.7	95.2	93.0	94.2
Mean & peak	97.0	98.5	98.5	98.7	97.7	98.5	98.7
Mean & variance	97.7	98.0	97.7	98.0	97.7	97.0	97.2
Slope & peak	92.7	92.7	88.9	96.0	94.2	90.2	91.7
Slope & variance	98.0	98.0	97.5	97.7	96.2	95.2	96.0
Peak & variance	98.0	98.7	97.5	97.5	98.0	96.7	98.5
Peak & skewness	92.7	85.2	83.9	93.7	90.5	86.9	83.4
Mean & skewness	89.5	88.9	84.9	88.2	88.4	86.2	87.7
Slope & skewness	82.4	86.4	84.9	84.7	86.4	82.7	86.7
Kurtosis & skewness	54.5	54.8	51.3	52.5	54.8	52.8	50.8
Variance & skewness	98.0	98.0	97.0	97.5	96.2	95.7	96.2
Peak & kurtosis	90.2	83.4	81.2	94.0	86.2	84.2	83.2
Mean & kurtosis	87.7	89.2	85.4	89.2	88.9	84.4	87.7
Slope & kurtosis	82.7	86.7	83.9	83.2	86.4	81.2	86.2
Variance & kurtosis	98.5	98.5	97.2	97.7	96.7	95.7	96.7