Performance analysis of QR-decomposition RLS and householder sliding window RLS for noise elimination of EEG

• Published in: 2017 IEEE Region 10 Humanitarian Technology Conference (R10-HTC) pp. 594 - 597
• Main Authors: Khanom, Samiha, Islam, Md. Rabiul
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.12.2017
• Subjects: Adaptive filters; Algorithm design and analysis; Electroencephalography; Filtering algorithms; Gaussian noise; householder sliding window; mean square error; Microsoft Windows; normalised correlation coefficient; QR-decomposition; Signal to noise ratio; Welch PSD
• ISSN: 2572-7621
• DOI: 10.1109/R10-HTC.2017.8289030

The paper conducts with the performance analysis of QR-decomposition (QRD) and householder sliding window (HSW) recursive least square (RLS) algorithm of adaptive filter for elimination of noise from EEG signal. Previous works show that the conventional RLS algorithm is best suited for noise elimination but it is more susceptible to noise with the increase of the number of iteration and coefficients of filter. QRD-RLS and HSW-RLS are more robust and numerically stable than conventional RLS. For the performance analysis, The EEG signal is collected from a database and mixed with random noise, white Gaussian noise and coloured noise. The noise corrupted EEG signal is filtered adaptively with the QRD-RLS and HSW-RLS and the performance is measured with mean square error, signal-to-noise ratio, Welch power spectral density estimation and normalised correlation coefficient. With these parameters, the QR-decomposition RLS algorithm is found more efficient for the elimination of noise from EEG signal.