BCI Decoding via Point-Position Augmentation for SSVEP
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current brain-computer interface (BCI) decoding methods for steady-state visual evoked potentials (SSVEP) require long training times and can cause user discomfort due to prolonged stimulation, limiting their usability and efficiency.
Innovation Solution
A brain-computer interface decoding method based on point-position equivalent augmentation, which involves data preprocessing, point-position equivalent augmentation, task-related component analysis, and the use of a naive Bayes method for classification, to enhance decoding accuracy and reduce calibration time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If classical SSVEP-BCI decoding methods are used, then decoding accuracy can be achieved, but long training time and prolonged stimulation are required
Solution Approach 1:
The patent creates virtual copies of existing SSVEP signal samples by performing point-position equivalent augmentation. Sampling points from different cycles are rearranged to generate augmented training sets, effectively copying and transforming existing data to increase sample size without requiring additional actual stimulation time, thereby maintaining decoding accuracy while reducing training time
Solution Approach 2:
The patent performs point-position equivalent augmentation on the training set before decoding. By pre-processing the training data to create augmented samples with the same physical meaning but different temporal arrangements, the system prepares enhanced training data in advance, enabling accurate decoding with less actual stimulation time
2Measurement precision
If prolonged flickering stimulus is applied to achieve better decoding effects, then decoding accuracy improves, but visual discomfort and user fatigue increase
Solution Approach 1:
The patent generates virtual signal copies through point-position rearrangement of existing SSVEP samples. By creating augmented training sets from existing data without requiring prolonged actual stimulation, the system achieves better decoding accuracy while minimizing visual discomfort and user fatigue that would result from extended flickering stimulus
Solution Approach 2:
The patent changes the temporal arrangement parameters of existing SSVEP signal samples through point-position equivalent augmentation. By rearranging sampling points within the same stimulation cycle to create equivalent augmented samples, the system improves decoding performance without increasing stimulation duration, thereby reducing visual discomfort and fatigue
3Measurement precision
If more training samples are collected to improve decoding performance, then decoding accuracy increases, but stimulation time and user burden increase
Solution Approach 1:
The patent creates multiple virtual copies of limited SSVEP training samples through point-position equivalent augmentation. By rearranging sampling points from the same stimulation cycles to generate augmented training sets, the system effectively increases sample size without requiring additional stimulation time or increasing user burden
Solution Approach 2:
The patent performs point-position equivalent augmentation as a preliminary processing step on the training set. This pre-processing creates enhanced training data with increased effective sample size from existing data, enabling improved decoding accuracy without requiring users to undergo longer stimulation sessions
Data Source
AI summary
The present disclosure discloses a brain-computer interface decoding method and apparatus based on point-position equivalent augmentation. According to the method, a point-position equivalent transformation is performed on sampling points to augment training data and generate arrangement sets. The task-related component analysis is performed on the augmented data to generate spatial filter. Afterwards, a full-frequency directed rearrangement is performed on verification signals or test signals according to the equivalent arrangement sets. After spatial filtering, Pearson correlation coefficients between the rearranged signals and the decoding templates are calculated. These correlation coefficients will be classified and voted by using a naive Bayes method. The verification module will generate the coefficient probability density functions and a threshold, and the test module will finally output the predicted label based on these information.