A technique designed to eliminate this problem by improving
detection of saccadic movements using the continuous wavelet
transform (CWT) is presented below. Firstly, the signal is filtered
with the wavelet transform and a linear saccadic model is used
to detect and quantify saccadic movements. Next, an RBF neural
network is used to enhance the robustness of the above model’s
capacity to detect saccadic movements over long periods of time
by differentiating between movements due to fixation problems
and other variations in the EOG signal that could be interpreted
as small saccadic movements. Fig. 4 shows this process.
shows this process.