Electroencephalogram signal heterogeneous label space transfer learning method based on Riemannian manifold

Abstract

The invention provides an electroencephalogram signal heterogeneous label space transfer learning method based on Riemannian manifold, which comprises the following steps: segmenting a single frequency band of electroencephalogram data of a subject A; respectively splicing the electroencephalogram data of the tag 1 and the tag 2 in the electroencephalogram data of each sub-band into a new tag 1 and a new tag 2 according to different sequences; taking the new tag 1 and the new tag 2 as source domains, and taking the tag 3 and the tag 4 as target domains; respectively calculating average covariance matrixes of the electroencephalogram data of the tag 1, the tag 2, the tag 3 and the tag 4; aligning source domain data on the Riemannian manifold of each sub-band to a target domain through a linear transformation matrix constructed by an average covariance matrix; source domain and target domain data on the Riemannian manifold after alignment of the sub-bands are mapped to a tangent space through logarithms, and high-dimensional tangent space features are reduced to low-dimensional features through mRMR; and fusing the cut space features of each sub-band after dimension reduction, and inputting the cut space features into an LDA classifier for classification. According to the method, the application range of the electroencephalogram data can be expanded.

Description

technical field [0001] The invention relates to the technical field of signal processing, in particular to a Riemannian manifold-based EEG signal heterogeneous label space transfer learning method. Background technique [0002] A brain-computer interface is a system that converts brain activity into control signals to command external devices. With the help of the BCI system, brain activity can be translated into commands to drive the device without relying on surrounding nerves and muscles. Many applications of the BCI system have been developed as a rehabilitation tool for people with communication disorders, and it is also seen as an enhancement tool for people in good health. EEG signals have the characteristics of low cost and high temporal resolution, and are widely used in EEG interfaces. Motor imagery (MI) is a spontaneously generated EEG signal that does not require external stimulation, and is especially suitable for rehabilitation training and motor control of p...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the above defects or improvement requirements of the prior art, the present invention provides a Riemannian-based EEG signal heterogeneous label space transfer learning method, the purpose of which is to solve, in the EEG signal calibration process, when the source domain data and When the label space of the target domain data is different, the source domain data cannot be used. Collecting a large number of EEG data with new labels for calibration will greatly increase the burden on the subjects and lengthen the calibration time.

Images