Third-order volterra model based on DUPSO for EEG signal denoising

In order to gain a high-performance analysis for electroencephalogram (EEG) signal, denoising has become a hot topic in this field. Due to the good performance of Volterra model, it has been widely applied to remove the noise in EEG signals. However, the inherent dimensionality problem of Volterra series makes the existing research mostly choose low order Volterra, which will sacrifice some accuracy. In addition, the realistic acquisition of EEG signals cannot satisfy the ideal conditions of phase space reconstruction, which makes it difficult to reconstruct phase space in the analysis process. For the purpose to solve these two questions, we introduce a third-order Volterra filter (TOVF) model to study the denoising problem of EEG and apply a dissipative uniform searching particle swarm optimization (DUPSO) algorithm to optimize the model's coefficients. Then a denoising model based on DUPSO third-order Volterra filter (DUPSO-TOVF) can be obtained. Simulating results show that the DUPSO-TOVF model has a significant increase in the SNR and decrease in the MSE when compared with UPSO second-order Volterra filter (UPSO-SOVF) and PSO-SOVF. Besides, the calculation time of DUPSO-TOVF model is not much different from the two compared models', which means the proposed model not only has the highest precision among the compared models but also can avoid the dimension disaster effectively.