 This paper proposes two novel frameworks based on an improved two-dimensional non-linear Fitsu Nugumo, FHN, neuron system to extract features from single-channel motor imagery, MI. The proposed methodology was evaluated using an open-access database, BCI Competition for Dataset 2A, an offline database, and a 10-fold cross-validation procedure. Results show that the improved non-linear FHN system can transfer the energy of noise into MI, thereby effectively enhancing the time-frequency energy. Compared with traditional methods, the proposed methods achieved higher classification accuracy and robustness. This article was authored by Reichi Uenshin, Guanghua Su, Yagwang Jia, and others.