An Improved Noise Elimination Model of EEG Based on Second Order Volterra Filter

Recently, electroencephalogram (EEG) is widely applied for physiological research and clinical diagnosis of brain diseases. Therefore, how to eliminate noise to gain a pure EEG signal becomes a common difficulty in this field. As a typical method for chaotic time series, Volterra is widely used to study EEG signal. However, the calculation of Volterra coefficients is likely to cause dimensionality disaster. In addition, EEG signals collected in real environment are not easy to extract the prior information, which is related to the quality of the reconstructed phase space. In order to overcome these two problems, we introduce a uniform searching particle swarm optimization (UPSO) algorithm to optimize the coefficients of Volterra then a noise elimination method based on UPSO second order Volterra filter (UPSO-SOVF) can be constructed. The proposed model can improve the quality of phase-space reconstruction by implicating the phase space reconstruction process in the model solving process and then get the embedding dimension and delay time dynamically. In this paper, some experiments are made on different EEG signals and compared with the particle swarm optimization second order Volterra filter (PSO-SOVF). The result shows that the proposed model has a better performance in avoiding the dimensional disaster and can better reflect regularities of the EEG signal series than PSO-SOVF. It can fully meet the requirements for noise elimination of EEG signal.