A two-stage denoising framework for EOG signals has been proposed.

In Stage I, the algorithm uses one of the four techniques – band-pass filter, EMD, SWT and median filter, which gives the highest SNR estimate.

An important finding of this work is the use of recursive estimators, which helps in improving the SNR of the EOG data in Stage II.

Recursive estimators take reasonably higher CPU time and hence are called upon only in case the Stage I methods do not provide a significant SNR.

The developed algorithm can be extended to other biomedical signals with proper modelling.