CONFIRMED: ECE Department: ELE Master of Science Thesis Defense by Christopher Gravelle
Topic: Experimental Demonstration of Sound Source Bearing Estimation at a Single Receiver Using Spectral Cues Location: Lester W. Cory Conference Room Science & Engineering Building (SENG), Room 213A Abstract: Traditionally, bearing estimation of a sound source is accomplished by exploiting relative time differences of arrival across an array of sensors. For scenarios where arrays of sensors cannot be effectively utilized due to size or cost, an alternative method for source localization must be considered. Yovel et al. [Science, 2010] discovered one such method when observing that bats steer the main response axis (MRA) of their echolocation beams askew of a target. While this strategy reduces the SNR of a received echo, it paradoxically improves the Fisher information about a target's bearing angle. This trade-off suggests there are spectral cues which improve target localization despite the lost signal power. For a single sensor with far-field frequency-dependent directivity, and a known broadband waveform of finite temporal support and sufficient SNR, it is possible to locate a sound source by the angle-dependent lowpass filtering of the signal at the receiver. Furthermore, there exists an angle which minimizes the bearing angle estimate that is near to, but not coincident with the main response angle of the receiver. This thesis presents experiments demonstrating a man-made system estimating a source's bearing from signal spectral information using a single directional sensor with frequency dependence over the bandwidth of the received signal. A linear FM chirp from a source in the far field is recorded, measuring the beampattern of the receiver. Maximum likelihood estimates (MLE) of source bearing are calculated in a Monte Carlo algorithm, comparing noise-corrupted recordings to a randomized dictionary of template recordings. Mean-squared error (MSE) is computed as a function of source angle and compared to the Cramer-Rao lower bound (CRLB). Experimental results support that the MSE is not proportional to angle-dependent SNR, rather there are local variance minima away from the receiver MRA where the received signal power is attenuated. The MSE local minima are consistent with optimal angles observed in previous studies simulating the exploitation of spectral cues on target localization [Kloepper et al., JASA-EL 2018] [Tidwell & Buck, IEEE SSPD 2019]. Advisor(s): Dr. John R. Buck, Chancellor Professor, Department of Electrical & Computer Engineering, UMASS Dartmouth Committee Members: Dr. Dayalan P. Kasilingam, Professor, Department of Electrical & Computer Engineering, UMASS Dartmouth; Dr. Paul J. Gendron, Associate Professor, Department of Electrical & Computer Engineering, UMASS Dartmouth 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 email at jbuck@umassd.edu Zoom Conference Link: https://umassd.zoom.us/j/91744864902 Meeting ID: 917 4486 4902 Passcode: 844492
Science & Engineering Building, Lester W. Cory Conference Room: Room 213A
ECE: Electrical & Computer Engineering Department
508.999.9164
https://umassd.zoom.us/j/91744864902
Cost: Free