 This paper proposes a novel approach for classifying motor imagery tasks in a brain-computer interface environment. It combines outputs from two classifiers learning on wavelet time and wavelet image-scattering features of brain signals, respectively. The experimental results demonstrate that the proposed fusion model improves the accuracy by 11% compared to the best existing classifier. This suggests that the proposed approach has great potential for developing a reliable sensor-based intervention for assisting people with neurodizability to improve their quality of life. This article was authored by Twan D. Pham.