A Domain Adaptation Method for EEG Signals Based on Riemannian Manifold Coordinate Alignment

Abstract

The present invention claims to protect an EEG signal domain adaptation method based on Riemannian manifold coordinate alignment, which belongs to a transfer learning domain adaptation method, especially a cross-session and cross-object classification data alignment method in a brain-computer interface. The method includes the following steps: first, perform preprocessing such as de-averaging, filtering, and normalization on the collected EEG data; then use the covariance matrix to process it, and convert the EEG data into a symmetric positive definite (SPD) matrix; The positive definite matrix constructs the Riemannian manifold space; then, the EEG data points in the manifold space are mapped to the corresponding tangent space for coordinate alignment; finally, the aligned EEG data points are mapped back to the Riemannian manifold space and vectorized deal with. The invention can reduce the difference between the EEG signals of different subjects in the brain-computer interface system, so that the data distribution among different subjects tends to be consistent.

Description

technical field [0001] The invention belongs to a migration learning domain adaptation method, in particular to a data alignment method for cross-session and cross-object classification in a brain-computer interface. Background technique [0002] Machine learning has been successfully applied in many fields, but the cost of collecting and labeling samples with the same distribution as the target data is high. When the distribution of the source domain dataset and the target domain dataset are different, the source domain cannot achieve good prediction results on the target domain. Migration learning is a machine learning method that solves the distribution difference between the source domain and the target domain. Its core is to find the similarity between the source domain and the target domain, and use the similarity to apply the knowledge obtained in the source domain to the target domain. . [0003] Domain adaptation is a subproblem under the domain of transfer learni...

AI Technical Summary

Problems solved by technology

[0006] The invention aims to solve the time-consuming problem in the calibration stage caused by the large differences in the EEG signals of different subjects in the field of brain-computer interface

Images