EE Seminar: Non-Asymptotic Performance Bounds of Eigenvalue based Detection of Signals in Non-Gaussian Noise

~~Speaker: Ron Heimann
M.Sc. student under the supervision of Prof. Anthony J. Weiss, Prof. Amir Leshem and Prof. Ephraim Zehavi

Wednesday, February 10th, 2016 at 15:30
Room 011, Kitot Bldg., Faculty of Engineering

Non-Asymptotic Performance Bounds of Eigenvalue based Detection of Signals in Non-Gaussian Noise

Abstract

The core component of both cognitive radio and centralized multistatic passive radar is its detector. When a device is equipped with multiple antennas, the detection method is often based on an eigenvalue analysis of the sample covariance matrix. This work explores the performance of the common largest eigenvalue detector, for the case of a narrowband transmitted signal and calibrated i.i.d. white receiver noise. In contrast to popular Gaussian assumption, our performance bounds are valid for a wide family of signals and noise, such as sub-Gaussian and constant envelope random processes. Moreover, the results are given in closed-form for any finite number of observations and antennas, in contrast to the standard asymptotic analysis approach. Finally, the new error bounds are compared to simulation results for several practical examples, and prove to be informative.