The present invention provides a
deep learning model construction method based on MEMD (Multiple Empirical Mode
Decomposition). The method comprises the steps of: performing band-pass filtering of multi-channel brain electrical signals measured by a
brain control device to obtain p-dimensional brain electrical signals, performing
signal segmentation of the p-dimensional brain electrical signals bytaking 2 seconds as a unit to obtain a plurality of
signal samples, and performing multi-element empirical mode
decomposition of each
signal sample; obtaining different intrinsic mode function component numbers n for the signal samples, taking the minimum n as c, and retaining the first c intrinsic mode functions; for each signal sample, adding the c intrinsic mode functions obtained through multi-element empirical mode
decomposition with the trend of a data sequence X(t) for stacking to obtain three-dimensional data samples with the sizes of q*p*(c+1), and taking the three-dimensional data samples as input of the
deep learning model; and constructing a
deep learning model applied in the
motor imagery. The deep learning model construction method based on the MEMD and the application thereof in
motor imagery are high in computing efficiency in the deep learning model, output an instruction which can control the motion of devices such as a mechanical arm, and are good in timeliness andgood in utilization potentiality.