Motor imagery electroencephalogram signal processing method, device, and storage medium

A method for processing a motor imagery (MI) electroencephalogram (EEG) signal, performed by an electronic device, the method comprising:

• acquiring a to-be-recognized MI-EEG signal, the to-be-recognized MI-EEG signal being an MI-EEG signal collected from a target object;
• inputting the to-be-recognized MI-EEG signal to a target feature extraction model to obtain to-be-recognized MI features representing the to-be-recognized MI-EGG signal, and
• inputting the to-be-recognized MI features to a target classification model to obtain a target classification result outputted by the target classification model,
• wherein a similarity between feature distributions of different MI features extracted by the target feature extraction model for different MI training signals is greater than or equal to a first similarity threshold, the different MI training signals comprising a target MI-EEG signal and a source MI-EEG signal corresponding to a first known action,
• an action predicted to be performed in the source MI-EEG signal that is represented by a classification result outputted by the target classification model being consistent with the first known action, the classification result being a classification result determined by the target classification model according to source MI features extracted from the source MI-EEG signal by the target feature extraction model, and the target classification result being used for representing an action predicted to be performed in the to-be-recognized MI-EEG signal;
• and
• controlling, according to the target classification result, a target device matching the target object to perform the action predicted to be performed in the to-be-recognized MI-EEG signal.

An electronic device, comprising a memory and a processor, the memory storing a computer program, and the processor being configured to execute the computer program to cause the electronic device to perform steps, comprising:

• inputting a source MI-EEG signal belonging to a source domain and a target MI-EEG signal belonging to a target domain to an initial feature extraction model to obtain first source MI features representing the source MI-EEG signal and first target MI features representing the target MI-EEG signal;
• inputting the first source MI features to an initial classification model to obtain a first classification result outputted by the initial classification model, the first classification result representing an action predicted to be performed in the source MI-EEG signal;
• and
• adjusting a model parameter of the initial feature extraction model and/or a model parameter of the initial classification model to obtain a target feature extraction model and a target classification model when
• a first known action corresponding to the source MI-EEG signal is inconsistent with the action predicted to be performed of the source MI-EEG signal, or
• a similarity between a feature distribution of the first source MI features and a feature distribution of the first target MI features is less than a first predetermined threshold,
• wherein a similarity between a feature distribution of second source MI features representing the source MI-EEG signal and outputted by the target feature extraction model and a feature distribution of second target MI features representing the target MI-EEG signal and outputted by the target feature extraction model is greater than or equal to the first predetermined threshold, and an action predicted to be performed, represented by a second classification result outputted by the target classification model, of the source MI-EEG signal that is consistent with the first known action.