A Zero-Padding Frequency Domain Convolutional Neural Network for SSVEP Classification.

Area of Science:

• Neuroscience
• Computer Science
• Signal Processing

Background:

• Steady-state visual evoked potential (SSVEP) based brain-computer interfaces (BCI) are crucial for human-computer communication.
• Nonlinear relationships in SSVEP signals pose a significant challenge to BCI performance.

Purpose of the Study:

• To propose a novel Convolutional Neural Network (CNN) algorithm to enhance SSVEP BCI performance.
• To address the nonlinearities and improve the accuracy of SSVEP signal processing.

Main Methods:

• Utilized discrete Fourier transform to compute the power spectral density (PSD) of SSVEP signals.
• Implemented zero-padding in the time domain to refine PSD and ensure frequency point intervals match stimulation frequency gaps.
• Developed a Zero-Padding Frequency Domain Convolutional Neural Network (ZPFDCNN) model incorporating CNN's nonlinear transformation capabilities.

Main Results:

• Experimental validation on an SSVEP dataset demonstrated the effectiveness of the proposed ZPFDCNN method.
• The ZPFDCNN model significantly improved the information transfer rate (ITR) for high-speed SSVEP-based BCI.
• The method showed superior performance in handling complex SSVEP signal characteristics.

Conclusions:

• The ZPFDCNN model offers a promising approach for enhancing SSVEP BCI performance.
• This advancement has substantial potential for real-world BCI applications, particularly for high-speed communication.
• The study highlights the efficacy of combining signal processing techniques with deep learning for improved BCI systems.