I. Introduction
When the brain is performing motor imagery tasks, the Electroencephalogram (EEG) signals obtained by the BCI system are called the Motor Imagery Electroencephalogram (MI-EEG)[1]. Different brain activity contents will reflect different voltage fluctuations [2]. The MI-EEG signal is usually a non-stationary random signal with a low signal-to-noise ratio [3], [4]. In addition, there are great differences between different subjects, and it has a great relationship with the interference of the signal acquisition experiment site environment. Therefore, how to improve the quality of the MI-EEG signals, so that it can most accurately represent the intention of the motor task performer, this is still the focus of researchers in the field of BCI system.