TRAINING METHOD FOR ELECTROENCEPHALOGRAPHY MODE CLASSIFICATION MODEL, CLASSIFICATION METHOD AND SYSTEM

摘要

Disclosed is a training method for an electroencephalography mode classification model, applicable in tasks for personal identification (PI) and fatigue state while driving detection, comprising: acquiring, preprocessing, and marking electroencephalography (EEG) data to produce a training dataset having markers; inputting each piece of the EEG data into an attention mechanism-based convolutional neural network, extracting to produce a mode feature; and correcting, on the basis of the mode feature and markers of the EEG data, a parameter for use in an electroencephalography classification model. Also disclosed are a corresponding method for electroencephalography mode classification and a system for electroencephalography mode classification. Used in the classification of multiple driving-related tasks, such as the tasks for personal identification (PI) and fatigue state while driving detection, the average classification precision is high, a great balance is strived between classification precision and time costs, and potential application values are provided in the classification of multiple tasks of biomedical signals.