Electrooculography (EOG) gives the possibility of eye tracking using biosignal measurements. Typical EOG signal consists of rapid value changes (saccades) separated by almost constant values. Additionally, the pulse shape from eyelid blinking is observed. The separation of them is possible using numerous methods, like median filtering. The proposed optimization method based on a model fitting using the variable number of parameters gives the possibility of features localization even for nearby saccades and blinking pulses.