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

摘要: We analyse and compare the classification accuracies of six different classifiers for a two-class mental task arithmetic rest using functional near-infrared spectroscopy fNIRS signals. The signals tasks from prefrontal cortex region brain seven healthy subjects were acquired multichannel continuous-wave imaging system. After removal physiological noises, features extracted oxygenated hemoglobin HbO Two- three-dimensional combinations those used tasks. In classification, modalities, linear discriminant analysis LDA, quadratic QDA, k-nearest neighbour kNN, Naive Bayes approach, support vector machine SVM, artificial neural networks ANN, utilized. With these classifiers, average among 2- 3-dimensional 71.6, 90.0, 69.7, 89.8, 89.5, 91.4% 79.6, 95.2, 64.5, 94.8, 96.3%, respectively. ANN showed maximum accuracies: 91.4 96.3%. order to validate results, statistical significance test was performed, which confirmed that p values statistically significant relative all other < 0.005