Physiological Signal Classification via Spatial Filtering Matrix
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional methods for classifying physiological electric signals are cumbersome due to the need for transfer learning, which requires simultaneous use of source and target domain data, making it difficult to operate effectively across different subjects.
Innovation Solution
A method involving data alignment and spatial feature extraction using a target spatial filtering matrix generated from aligned training signals, allowing for classification without prior acquisition of target user signals, by reducing distribution differences and maximizing feature discrimination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If transfer learning method is used for physiological electric signal classification, then classification accuracy can be improved by using source domain information, but the operation becomes cumbersome requiring simultaneous use of source and target domain data
Solution Approach 1:
The method segments the classification process into two independent phases: (1) training phase using only source domain data to build a pre-trained model, and (2) application phase using the pre-trained model with target domain data. This segmentation eliminates the need to simultaneously use both source and target domain data during operation, making the system easier to deploy while maintaining classification accuracy through the pre-trained model's learned features
Solution Approach 2:
The method performs preliminary training action in advance using source domain data to pre-train the deep learning model before actual target domain classification. This preliminary action prepares the model with generalized features that can be directly applied to target domain signals without requiring simultaneous access to source domain data during operation, thus improving operational convenience while preserving classification accuracy
2Reliability
If traditional transfer learning is applied, then model performance can be enhanced, but the requirement for simultaneous source and target domain data makes it difficult to operate effectively across different subjects
Solution Approach 1:
The method creates a universal pre-trained model through domain adaptation training on source domain data that can function across multiple target subjects. The model learns subject-invariant features during pre-training that enable it to generalize to different subjects without requiring retraining or simultaneous access to source data, thus enhancing cross-subject applicability while maintaining model performance through the learned universal representations
Solution Approach 2:
The method creates a copy of the source domain knowledge embedded in the pre-trained model that can be applied to target domain subjects. Instead of requiring direct access to source domain data during target domain operation, the essential patterns and features are copied into the model's weight parameters during pre-training, enabling the model to function effectively across different subjects with only target domain data available at deployment
Data Source
AI summary
A physiological electric signal classification processing method includes: performing data alignment on an initial physiological electric signal corresponding to a target user identity based on target signal spatial information corresponding to the target user identify to obtain a target physiological electric signal; performing spatial feature extraction on the target physiological electric signal based on a target spatial filtering matrix to obtain a target spatial feature, the target spatial filtering matrix being generated based on target training physiological electric signals corresponding to a plurality of training user identities respectively and training labels corresponding to the target training physiological electric signals, the target training physiological electric signals being obtained by performing data alignment on initial training physiological electric signals based on training signal spatial information corresponding to the training user identities; and obtaining a classification result corresponding to the initial physiological electric signal based on the target spatial feature.