Performance Analysis of a Neuro Fuzzy Algorithm in Human Centered & Non-Invasive BCI

Brain-Computer Interface can be non-invasive devices that obtain sig-nals generated from the brain and are then manipulated to suit various applica-tions. A popular application for BCI is interfacing with robotics; and, each BCI – Robotics system employed different Machine Learning algorithms. This study aimed to present a performance analysis for a Neuro-Fuzzy algorithm, specifi-cally the Adaptive-Network-Fuzzy-Inference System (ANFIS), to classify EEG signals retrieved by the Emotiv INSIGHT. An SVM algorithm is also developed to serve as a reference vs the ANFIS’s performance. A methodology for genera-tion and acquisition of EEG signals can be used by researchers as reference. Fa-cial and Eye Gestures were utilized as means of EEG signal generation which are fed to both algorithms for simulation experiments. Results showed that the ANFIS tend to be more reliable and marginally better than of the SVM algorithm. Compared to SVM, the ANFIS took significant amounts of computational re-sources requiring higher specs and training time.