Cross-period brain grain recognition method and system based on single-source domain generalization

By constructing a cross-time period brainprint recognition model, and utilizing a data alignment module, a multi-scale spatiotemporal feature extractor, and a Transformer encoder, the problems of time-varying and inter-domain differences in EEG signals were solved, achieving efficient cross-time period brainprint recognition.

CN120974254APending Publication Date: 2025-11-18HANGZHOU DIANZI UNIV
0 Cites 0 Cited by

Patent Information

Application Number
CN202510952160.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-07-10
Publication Date
2025-11-18

AI Technical Summary

Technical Problem

In existing technologies, the time-varying and non-stationary nature of EEG signals leads to low accuracy in cross-time period brainprint recognition, large differences in distribution between domains, and traditional deep learning architectures are unable to effectively model long-term time dependencies, resulting in a decline in recognition performance.

Method used

A cross-time period brainprint recognition method based on single-source domain generalization is adopted. Through a data alignment module, a multi-scale spatiotemporal feature extractor, a feature fusion module, and a Transformer encoder, a cross-time period brainprint recognition model is constructed to achieve domain-invariant feature learning.

Benefits of technology

Using only EEG data from a single time period, it can achieve stable recognition in other time periods, with recognition performance superior to existing methods. It improves the accuracy and robustness of cross-time period brainprint recognition and effectively decouples the distribution differences of EEG signals across time periods.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention discloses a cross-period brain grain recognition method and system based on single-source domain generalization, and the method achieves the brain grain recognition of a target object through a constructed cross-period brain grain recognition model on the premise of only using single-period electroencephalogram data as a source domain. A constructed cross-time-period brain grain recognition model comprises a data alignment module, a multi-scale spatial-temporal feature extractor, a feature fusion module, a Transform encoder and a linear classification module which are connected in sequence. The data alignment module is used for performing data alignment processing on the electroencephalogram data to obtain an aligned electroencephalogram sample; the multi-scale space-time feature extractor is used for extracting time features and space features of electroencephalogram data of different brain regions, and the time features and the space features are fused through the feature fusion module and then sent to the Transform encoder to extract global information of the electroencephalogram data; and the linear classifier classifies the extracted global information and outputs a brain grain recognition result.
Need to check novelty before this filing date? Find Prior Art