Effective Approach for Detecting the Single Trial ERP Using Rational Gaussian Neural Network
Minfen Shen1, Yuzheng Zhang, Yisheng Zhu
1Key Lab. of Guangdong, Shantou University, Guangdong 515063, China.
Summary
This study introduces a novel rational Gaussian network to detect single-trial event-related potentials (ERPs). The method effectively suppresses noise and enhances signal-to-noise ratio (SNR) in biomedical signal processing.
Area of Science:
- Biomedical Signal Processing
- Neural Networks
- Signal Detection
Background:
- Detecting single-trial event-related potentials (ERPs) is challenging due to low signal-to-noise ratios (SNR).
- Traditional methods often struggle with effective noise suppression, limiting accuracy in real-time applications.
Purpose of the Study:
- To propose a novel modified Radial Basis Function (RBF) neural network, termed a rational Gaussian network, for enhanced single-trial ERP detection.
- To evaluate the performance of this network in terms of noise suppression and signal tracking capabilities.
Main Methods:
- A modified RBF neural network (rational Gaussian network) was developed.
- The Gaussian RBF was normalized to optimize noise suppression, particularly at low SNR.
- Performance was assessed using Mean Squared Error (MSE) and signal tracking ability.
Main Results:
- The proposed rational Gaussian network demonstrated significant enhancement of the SNR.
- The method successfully tracked signal variations, including specific ERP components.
- Experimental results with real ERP signals validated the effectiveness of the approach.
Conclusions:
- The novel rational Gaussian network offers a robust solution for single-trial ERP detection.
- This approach significantly improves SNR and signal tracking in biomedical signal processing.
- The method shows promise for advanced applications in analyzing neural signals.
