Speech imagination electroencephalogram signal recognition method based on multi-granularity information fusion

By introducing a multi-grained information fusion method of view importance, sample importance and feature importance in verbal imagination EEG signal recognition, combined with multi-view analysis, self-step learning and feature weighting technology, the problem of insufficient decoding accuracy and robustness in the existing technology is solved, and the classification effect of higher accuracy and stronger robustness of EEG signal is achieved.

CN120105157AActive Publication Date: 2025-06-06HANGZHOU DIANZI UNIV
8 Cites 1 Cited by

Patent Information

Application Number
CN202510259499.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-03-06
Publication Date
2025-06-06
Estimated Expiration
2045-03-06

AI Technical Summary

Technical Problem

The prior art has shortcomings in the decoding accuracy and robustness of verbal imagination EEG signals. Especially when processing complex multi-dimensional EEG data, it is difficult to effectively integrate signals from different frequencies, time, space and brain regions, and low-quality samples have a great impact.

Method used

A multi-grained information fusion method based on view importance, sample importance and feature importance is adopted. Through multi-view analysis, self-step learning and feature weighting technology, a machine learning model is constructed to integrate multi-dimensional information of EEG signals, highlight task-related features and suppress the impact of low-quality samples.

Benefits of technology

It significantly improves the decoding accuracy of EEG signals and the robustness of the model, enhances the overall performance of the brain-computer interface, and maintains good classification results in a variety of environments.

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

Abstract

The invention discloses a speech imagination electroencephalogram signal recognition method based on multi-granularity information fusion, and the method comprises the steps: firstly collecting electroencephalogram data when a subject imagines multiple tasks, carrying out the preprocessing and feature extraction of the collected electroencephalogram data, and obtaining a sample vector and a label corresponding to the sample vector; secondly, a machine learning model combining view importance, sample importance and feature importance is constructed, and a joint optimization objective function is obtained; and then initializing weights of each view importance, sample importance and feature importance, and performing iterative optimization through an alternate optimization method according to a target function. And finally, after iterative optimization, inputting the sample vector into a machine learning model to obtain a prediction classification category of the speech imagination electroencephalogram signals. According to the method, information related to tasks can be effectively extracted from complex electroencephalogram signals, and it is ensured that a good classification effect can still be kept in a changeable environment.
Need to check novelty before this filing date? Find Prior Art