EEG Decoding via Non-Negative CP Decomposition and 2-DPCA
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing EEG decoding methods fail to effectively extract and identify time component characteristics from EEG data in boundary avoidance tasks, neglecting the interaction among EEG modes.
Innovation Solution
An EEG decoding method based on non-negative CP decomposition, combined with 2-DPCA for optimizing characteristic dimensions and support vector machine classification, to extract and decode time component characteristics of EEG data for left and right hand movements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional tensor discriminant analysis algorithm is used to extract frequency component of single motion imagination, then recognition effect of motion imagination intention is improved, but interaction among EEG modes is ignored
Solution Approach 1:
The patent transforms the traditional 3-order tensor (channel, frequency, time) into a 4-order tensor by adding the subject dimension. This dimensional expansion enables the model to capture interactions across all four modes simultaneously, resolving the limitation of ignoring mode interactions while maintaining recognition accuracy.
Solution Approach 2:
The patent employs nested decomposition by first applying non-negative CP decomposition to the 4-order tensor to extract subject-specific features, then applying 2-DPCA to the extracted time components for further feature optimization. This nested approach allows progressive extraction of information at different levels while preserving mode interactions.
2Measurement precision
If EEG data from multiple subjects is processed separately, then individual characteristics are captured, but data efficiency and generalization are reduced
Solution Approach 1:
The patent merges EEG data from multiple subjects into a unified 4-order tensor structure, enabling joint processing that captures both individual characteristics and cross-subject patterns. The non-negative CP decomposition then separates subject-specific components from shared patterns, achieving both individualization and data efficiency simultaneously.
Solution Approach 2:
The patent segments the combined multi-subject tensor into subject-specific component matrices through non-negative CP decomposition. This segmentation allows extraction of individual characteristics from the aggregated data, maintaining personalization while utilizing the full dataset for improved generalization.
3Measurement precision
If more EEG modes and dimensions are utilized, then more comprehensive characteristics are extracted, but computational complexity increases
Solution Approach 1:
The patent changes the mathematical parameters by using non-negative constraints in CP decomposition and applying 2-DPCA for dimensionality reduction. These parameter transformations simplify the computational structure while preserving the essential information from all four modes, reducing complexity without sacrificing extraction completeness.
Solution Approach 2:
The patent extracts only the essential time component characteristics from the 4-order tensor through non-negative CP decomposition, then further optimizes these extracted features using 2-DPCA. This selective extraction approach processes comprehensive multi-mode data while maintaining computational efficiency by focusing on the most informative components.
Data Source
AI summary
This disclosure provides an EEG decoding method based on a non-negative CP decomposition model. The method extracts time component characteristics of the EEG of the different subjects in the boundary avoidance task, optimizes a characteristic dimension by using a 2-DPCA, and takes classification by using a support vector machine, so that differences of the EEG of subjects in different states can be reflected, and the EEG classification of the single subject has a great accuracy. The time component characteristics of the EEG can be obtained by using the channel components and the frequency components based on the non-negative CP decomposition model and by means of the interaction between the EEG modes. The characteristics of the obtained EEG time components have good separability, and the dimensions of the characteristics are optimized, so that the EEG of left and right hand movements in the boundary avoidance tasks can be effectively decoded.