University Events

Description: Topic: Robust Spectral Estimation and Adaptive Beamforming with Linear Combinations
of Order Statistics

Location: Science & Engineering Building (SENG), Room 212

Zoom Conference Link: https://umassd.zoom.us/j/96475537559
Meeting ID: 964 7553 7559 Passcode: 448179

Abstract:
The sample mean is widely employed in parameter estimation problems, but the sensitivity to outliers in data limits its scope of use. In contrast, estimators based on order statistics avoid the bias caused by outliers. This dissertation proposal will study the properties of order statistics and their combinations applied to estimation problems in both spectral analysis and array signal processing. In spectral analysis, the loud transients often present in acoustic recordings are the outliers. Those transients introduce bias to Welch's spectral estimator. Schwock and Abadi [2021] showed that a modified Welch estimator based on the 80th sample percentile is the minimum variance unbiased estimator based on a single order statistic. This work developed a hybrid approach between Welch's and order statistics estimators by performing a weighted sum of the quietest subset of ordered samples of the periodograms. By discarding the loudest samples of the periodogram, the truncated linear order statistics filter (TLOSF) reduces the bias caused by loud transients. By combining multiple order statistics into the estimate, the TLOSF achieves a lower variance than the 80th percentile estimator. Employing parameter estimators based on order statistics can also be useful in array processing of narrowband signals. When applied to the Dominant Mode Rejection adaptive beamformer, median filtering provides a more accurate estimate of the background noise power than the sample mean. Similar to the Welch percentile spectral estimator, this work plans to investigate which order statistics other than the median serves as an unbiased or minimum variance estimator of the noise power. In addition, a weighted sum of order statistics should further reduce the variance of the noise power estimator while keeping its accuracy.

Advisor(s): Dr. John R. Buck, Chancellor Professor, Department of Electrical & Computer Engineering, UMASS Dartmouth

Committee Members: Dr. Paul J. Gendron, Associate Professor, Department of Electrical & Computer Engineering, UMASS Dartmouth; Dr. Ruolin Zhou, Assistant Professor, Department of Electrical & Computer Engineering, UMASS Dartmouth; Dr. Shima Abadi, Associate Professor, University of Washington

NOTE: All ECE Graduate Students are ENCOURAGED to attend.
All interested parties are invited to attend. Open to the public.

*For further information, please contact Dr. John R. Buck at 508.999.9237 or via email at jbuck@umassd.edu.