Classification of SSVEP signals using the combined FoCCA-KNN method and comparison with other machine learning methods

Abstract

Brain–computer interfaces (BCI) can be used to decode brain activity and extract commands to control external devices. The inherent complexity of brain signals and the interconnectedness of the information processing steps of these signals have created a sense of the need to use machine learning methods and identify the appropriate pattern to increase the accuracy of the results. In this paper, a new classification approach, which combines the extraction of steady-state visual evoked potentials (SSVEP) signal properties using the Fusing Canonical Coefficients (FoCCA) method and machine learning algorithms, is proposed to increase the accuracy of SSVEP signal classification. This approach, which uses the FoCCA algorithm or other existing algorithms as a feature extractor, is considered a new approach in the processing of brain signals. The results obtained from Support vector machines (SVM), K-Nearest Neighbor (KNN), and Decision Tree algorithms are also compared with each other and has studied and compared to the existing statistical methods. Using the FoCCA method will help us to extract and select the appropriate features to present to the machine learning algorithm and classify the signals, and the application of machine learning algorithms will play an effective role in increasing the classification accuracy.