Real-Time Baseline Wander Removal from Electrooculography Using Probabilistic Baseline Prediction

Electrooculography (EOG) is a convenient method for analysing eye movement. A human eye acts like a dipole where the back of the eye is relatively negative than the front. EOG measures the field generated by this dipole-like property of an eye with electrodes. It is low cost, easily installed, and most importantly, non-invasive. However, EOG is not conventionally adopted because of difficulties in removing baseline wander (BW) noise. Many studies have successfully removed this noise in repetitive and predictable signals such as electrocardiography (ECG), but not many accomplishments have been made in non-predictable signals such as EOG. This paper proposes an algorithm that removes BW from EOG in real-time by predicting the baseline. The algorithm constantly differentiates the signal when the subject is looking forward, referred to as the reference data in this paper, from the signals when the subject is looking elsewhere using standard deviation. Then, the average of certain window of reference data is taken to predict the baseline. A simulation comparison between some of the other methods used for ECG and EOG BW removal is presented which shows that the proposed algorithm performs exceptionally well, especially considering that it is one of the few algorithms that can remove the BW in real-time.