Channel component correlation analysis for multi-channel EEG feature component extraction

Wenqiang Yan ^1,2* Qi Luo ¹ ChengHang Du ¹

• ¹ Xi'an Jiaotong University, Xi'an, China
• ² State Key Laboratory of Robotics, Shenyang Institute of Automation (CAS), Shenyang, Liaoning Province, China

Electroencephalogram (EEG) analysis has shown significant research value for brain disease diagnosis, neuromodulation and brain-computer interface (BCI) application. The analysis and processing of EEG signals is complex since EEG are nonstationary, nonlinear, and often contaminated by intense background noise. Principal component analysis (PCA) and independent component analysis (ICA), as the commonly used methods for multi-dimensional signal feature component extraction, still have some limitations in terms of performance and calculation. In this study, channel component correlation analysis (CCCA) method was proposed to extract feature components of multi-channel EEG. Firstly, empirical wavelet transform (EWT) decomposed each channel signal into different frequency bands, and reconstructed them into a multi-dimensional signal. Then the objective optimization function was constructed by maximizing the covariance between multi-dimensional signals. Finally the feature components of multi-channel EEG were extracted using the calculated weight coefficient. The results showed that the CCCA method could find the most relevant frequency band between multi-channel EEG. Compared with PCA and ICA methods, CCCA could extract the common components of multi-channel EEG more effectively, which is of great significance for the accurate analysis of EEG. The CCCA method proposed in this study has shown excellent performance in the feature component extraction of multi-channel EEG and could be considered for practical engineering applications.