Decoding method of motor imagery EEG signal based on OA-WMNE brain source imaging

Abstract

The present invention is based on the decoding method of the motor imagery EEG signal of OA-WMNE brain source imaging, first adopts baseline correction and the superposition average on the time domain to carry out the pretreatment of EEG signal, obtains the superposition average signal of each kind of motor imagery task; And then , using the WMNE brain-source imaging algorithm to inverse transform it into the brain-source space to obtain dipole estimates, and determine the time interval of interest (TOI) according to the difference between the two kinds of motor imagery dipole waveform changes; and then for all single motor imagery The EEG signal is inversely transformed, and all the dipole amplitudes on each sampling point in the TOI form a feature vector to obtain a set of features on the sampling point; then the features on all sampling points form a feature sample set, The zero-mean standardization process was performed on it, and the univariate feature selection method was used for feature dimensionality reduction; finally, the support vector machine was used for feature classification to obtain the highest average decoding accuracy, improve the spatial resolution of EEG, and help improve the performance of motor imagery tasks. decoding accuracy.

Description

technical field [0001] The invention belongs to the technical field of brain source space decoding of EEG signals, in particular to a decoding method for motor imagery EEG signals from the brain dipole source space in a Brain Computer Interface (BCI) system, which adopts superposition in the time domain A method combining Overlapping Averaging (OA) and Weighted Minimum Norm Estimates (WMNE) brain-source imaging techniques (denoted as OA-WMNE) decodes motor imagery EEG signals in the brain-source spatial domain. Background technique [0002] Motor Imagery Electroencephalography (MI-EEG) hides a large amount of biological information in the motor perception cortex of the brain. MI-EEG signals recorded non-invasively in the scalp provide an important reference for brain activity in the field of sensors, because they have relatively High time-frequency resolution is widely used in the fields of BCI system research and clinical rehabilitation evaluation, so the salient features o...

AI Technical Summary

Problems solved by technology

However, this method has the following problems in practical application: (1) The preprocessing process of ICA to decompose the EEG signal will cause the loss of some effective information of the original MI-EEG, which is different from the purpose of enlarging the characteristic information of the scalp signal through the inverse EEG transformation. conflict

(2) The inverse transformation of a single independent component only maps a main source signal of the scalp electrode, and does not make full use of all the physiological information contained in MI-EEG. Uniform phenomenon, affecting the accuracy of decoding

(3) The selection of the brain-source spatial activation region (Region of Interest) is limited by the relatively obvious unilateral limb motor imagery with Event Related Desynchronization (ERD). For the correlation analysis of more complex motor imagery tasks , the results of the most relevant independent component source imaging are difficult to reflect the information of the real activity of the cerebral cortex, resulting in a decrease in decoding accuracy

Images