Fast reconstruction of EEG signal compression sensing based on deep learning

Area of Science:

• Biomedical Engineering
• Signal Processing
• Machine Learning

Background:

• Traditional electroencephalography (EEG) signal acquisition generates large datasets, posing challenges for data transmission and storage due to bandwidth, delay, and memory limitations.
• Compressed Sensing (CS) offers a solution to transmission pressure, but iterative reconstruction algorithms are computationally complex and slow, hindering rapid EEG monitoring.
• Existing CS reconstruction methods for EEG signals face limitations in speed and computational efficiency for real-time applications.

Purpose of the Study:

• To develop a non-iterative, fast algorithm for reconstructing EEG signals using compressed sensing (CS) and deep learning.
• To overcome the computational complexity and slow data processing speed associated with traditional iterative CS reconstruction algorithms.
• To enable rapid and accurate reconstruction of EEG signals for enhanced monitoring systems.

Main Methods:

• A novel non-iterative algorithm combining compressed sensing (CS) and deep learning techniques for EEG signal reconstruction.
• Utilized an improved residual network model incorporating one-dimensional dilated convolution for feature extraction.
• Learned the nonlinear mapping between measured values and original EEG signals for direct reconstruction.

Main Results:

• The proposed deep learning-based CS algorithm demonstrated significantly higher reconstruction accuracy compared to traditional CS methods.
• Achieved a markedly faster reconstruction speed than both traditional iterative CS algorithms and existing deep learning reconstruction algorithms.
• Successfully validated through simulations on the open BCI contest dataset, proving its efficacy.

Conclusions:

• The developed non-iterative deep learning algorithm provides a rapid and accurate solution for EEG signal reconstruction.
• This method effectively addresses the limitations of traditional CS reconstruction, enabling practical application in rapid EEG monitoring systems.
• The approach offers superior performance in both accuracy and speed, advancing the field of EEG signal processing.