Figure 3 | Scientific Reports

Figure 3

From: A new and general approach to signal denoising and eye movement classification based on segmented linear regression

Figure 3

Sample time series of a simulated linear smooth pursuit (solid black line) separated by saccades, and the denoising results of the three benchmarked algorithms with simulated sampling noise level of 1.5° (gray data points). NSLR (solid red line) identifies four segments. Note that it approximates the intersaccadic slow movement interval as two separate linear segments, with the first corresponding to the post saccadic oscillation. Total Variation (dotted blue line) and Wiener (dotted green line) filters reduce the measurement error considerably as well. Total Variation produces an approximately piecewise constant signal, which is characteristic to this algorithm. Wiener filter recovers the PSO well but exhibits ringing effects during the more linear part. Supplementary Figures S4S6 show NSLR with different simulated noise levels using the simulated data.

Back to article page