A Novel Feature Fusion and Reprocessing Technique of Brain-Computer Interface for Motion Imagination

Brain-computer interface (BCI) refers to the direct communication and control channels established between human brain and computer or other electronic devices. As a novel way of man-machine interface, BCI enables paralyzed patients to see the new hope of autonomous interaction with the outside world. The background of this paper is BCI based on motion imagination. Different features such as discrete wavelet transform (DWT), power spectral density (PSD) and autoregressive (AR), can describe useful information from different angles, feature selection can reflect the classification results of BCI. In this paper, four BCI feature fusion techniques are proposed, and the results are compared. To further improve the classification, we employ different feature reprocessing techniques (PCA, ICA and KPCA) to deal with the fusion matrix. The classification result is the best when the feature fusion method is weighted addition of DWT, PSD and AR, while KPCA is the reprocessing technique.