Research on Motor Imagery Based on Optimal Frequency Band Selection
Electroencephalogram(EEG)signals consist of multiple frequency bands,with motor imagery primarily involving the mu and beta bands.When the required frequency bands are clearly defined,classification tasks can generally be performed smoothly.However,frequency band selection has not been a major focus of research.Thus,developing an application capable of screening the optimal frequency bands for classification tasks is particularly important.This research analyzes the classification performance of EEG data related to visual cue-based left-hand and right-hand motor imagery based on the BCI Competition IV dataset 2b.The experiments utilize a 10-fold cross-validation approach to evaluate the models,aiming to reduce randomness and provide stable performance metrics.This research applies two models of BP-CNN and AR-CNN,and compares them with a CNN model that extends the frequency bands within known ranges.Even within unknown frequency ranges,effective motor imagery recognition is achieved through precise frequency band selection and model optimization.It provides a theoretical foundation for brain-computer interface technology and offers significant reference value for future research in related fields.
motor imagerybrain-computer interfacefrequency band selectionConvolutional Neural Networks
国家科技期刊平台
NSTL
万方数据
