Study of Kalman filter based Noise Estimation in Artifactual EEG and their Quantification

This paper presents an approach to estimate the noise in electroencephalographic (EEG) recordings. The approach is based on the Kalman filter technique. The parameters of noise signals are estimated using the Kalman filter. The idea behind this approach is to denoise the EEG signals and to preserve the stationarity by considering different window sized EEG time series. The results are validated in different window sizes of 300 milliseconds, 1 minute, 2 minutes, and 3 minutes up to 4 minutes data in different EEG. The entropy of the original signal and the estimated signal is also calculated, which shows the content of the signal. The Mean Square Error (MSE) and Signal to Noise ratio (SNR) in different channels are calculated for different window size EEG data. From the obtained results, we observe that with an increase in window size data, the MSE decreases and SNR increases and the entropy of the estimated signal is almost same as that of the original signal. Depending on the requirement, the window size is chosen. The smaller window gives better visibility, and the larger window gives less error. Estimation of noise quantifies the different types of inband artifacts contaminated the EEG signal.