SSVEP Detection Narrowing Frequency Window
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current SSVEP-based brain-computer interface (BCI) systems using canonical correlation analysis (CCA) are time-consuming and require significant computational resources to identify target frequencies, as they calculate correlation coefficients across a wide frequency range.
Innovation Solution
The method involves generating icons with unique target frequencies, calculating correlation coefficients within a ±0.5 Hz window of the target frequencies, and determining a confidence score based on these coefficients to efficiently identify the selected icon, reducing computational load and improving execution time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If canonical correlation analysis calculates correlation coefficients across a wide frequency range to identify target frequencies, then measurement precision is improved, but loss of time and computational resources increase
Solution Approach 1:
The patent divides the frequency analysis into two stages: first performing a coarse analysis across a wide frequency range to identify potential target frequencies, then performing a refined correlation coefficient calculation only within a narrow ±0.5 Hz window around identified frequencies. This segmentation reduces overall computation time while maintaining identification accuracy.
Solution Approach 2:
The patent performs preliminary frequency identification using a broader analysis method before conducting the detailed correlation coefficient calculation. By pre-identifying candidate frequencies through initial spectral analysis, the system avoids calculating correlation coefficients across the entire frequency range, thus reducing detection time while preserving measurement precision for the final target frequency identification.
2Measurement precision
If canonical correlation analysis calculates correlation coefficients across a wide frequency range to identify target frequencies, then measurement precision is improved, but use of computational resources increases
Solution Approach 1:
The patent segments the computational process into an initial broad frequency sweep followed by focused correlation calculations only at identified candidate frequencies. This reduces the total number of correlation coefficient calculations from potentially hundreds across the full frequency range to just a few at specific target frequencies, significantly lowering computational resource consumption while maintaining identification accuracy.
Solution Approach 2:
The system performs preliminary spectral analysis to identify candidate target frequencies before executing the computationally intensive correlation coefficient calculations. This preliminary action filters out frequencies that do not correspond to actual targets, preventing wasted computational resources on irrelevant frequency ranges while ensuring accurate identification of true target frequencies.
3Reliability
If correlation coefficients are calculated across a wide frequency range, then reliability of target frequency identification is improved, but productivity decreases
Solution Approach 1:
The patent divides the detection process into segments: initial frequency range scanning followed by focused correlation analysis at identified frequencies. This segmentation maintains reliability by ensuring thorough initial scanning while improving productivity through the subsequent focused analysis that requires fewer calculations and less time.
Solution Approach 2:
The system performs preliminary frequency identification through spectral analysis before conducting detailed correlation coefficient calculations. This preliminary action ensures that no potential target frequencies are missed in the initial sweep, maintaining identification reliability, while the subsequent focused calculations at only the identified frequencies significantly increase detection speed and productivity.
Data Source
AI summary
In accordance with one embodiment of the present disclosure, a method includes generating a plurality of icons, wherein each icon has a target frequency unique from each other, receiving brain activity data based on an epoch, generating a reference signal based on the epoch, calculating correlation coefficients between the brain activity data and the reference signal, wherein the correlation coefficients are calculated in a window that is within ±0.5 Hz of the target frequencies, including endpoints, determining a confidence score based on the correlation coefficients and the epoch, and determining a selected icon among the plurality of icons based on the correlation coefficients in response to the confidence score surpassing a threshold confidence score.