Application of robust data processing methods to the analysis of eye movements

Area of Science:

• Biomedical Engineering
• Signal Processing
• Neuroscience

Background:

• Accurate estimation of eye velocity from eye position signals is crucial for understanding visual-motor control and diagnosing neurological disorders.
• Traditional linear filters often introduce artifacts like ringing and are susceptible to noise, limiting their efficacy in real-world biomedical applications.

Purpose of the Study:

• To develop and evaluate a robust, nonlinear differentiating digital filter for accurate eye velocity estimation.
• To demonstrate the filter's applicability to other biomedical signals and its superiority over existing methods.

Main Methods:

• A novel nonlinear differentiating digital filter, termed the Robust Differentiator (RD), was developed.
• The RD utilizes an odd number of two-point differences followed by a median operation to calculate the derivative.
• The filter's performance was assessed against impulse and Gaussian noise, comparing it with FIR differentiating filters.

Main Results:

• The RD effectively estimates eye velocity from eye position signals, applicable to diverse biomedical data.
• The filter eliminates the "ringing" artifact common in linear filters and exhibits no impulse response.
• The RD demonstrates superior performance in the presence of impulse noise and is equally effective against broadband and narrowband Gaussian noise.

Conclusions:

• The Robust Differentiator offers a robust and effective solution for estimating derivatives of biomedical signals, particularly eye position.
• Its nonlinear, order-statistic approach provides significant advantages in noise suppression and artifact elimination compared to linear filters.
• The adjustable "bandwidth" of the RD enhances its versatility for various signal processing tasks.