 This paper presents a novel approach called Interactive Frequency Comvolutional Neural Network, IFNET, for motor imagery, MI, decoding. It uses cross-frequency interactions to enhance the representation of MI characteristics, which are extracted from low and high-frequency bands. The proposed method outperforms existing methods on two benchmark datasets, demonstrating its superiority in terms of classification accuracy and speed. This article was authored by Jiahung Wang, Lin Yao, and Yu Ming Wang.