Skip to Main Content
This paper presents a novel algorithm for efficient feature extraction using mutual information (MI). In terms of mutual information, the optimal feature extraction is creating a new feature set from the data which jointly have largest dependency on the target class. However, it is not always easy to get an accurate estimation for high-dimensional MI. In this paper, we propose an efficient method for feature extraction using two-dimensional MI estimates. A new feature is created such that the MI between the new feature and the target class is maximized and the redundancy is minimized. The effectiveness of the proposed algorithm is evaluated by using the classification of EEG signals. The tasks to be discriminated are the imaginative hand movement and the resting state. The results demonstrate that the proposed mutual information- based feature extraction (MIFX) algorithm performed well in several experiments on different subjects and can improve the classification accuracy of the EEG patterns. The results show that the classification accuracy obtained by MIFX is higher than that achieved by full feature set.