Articles via Databases
Articles via Journals
Online Catalog
E-books
Research & Information Literacy
Interlibrary loan
Theses & Dissertations
Collections
Policies
Services
About / Contact Us
Administration
Littman Architecture Library
This site will be removed in January 2019, please change your bookmarks.
This page will redirect to https://digitalcommons.njit.edu/theses/1655 in 5 seconds

The New Jersey Institute of Technology's
Electronic Theses & Dissertations Project

Title: Blind source separation using dictionary learning over time-varying channels
Author: Ghosh, Anushreya
View Online: njit-etd2019-009
(x, 64 pages ~ 1.1 MB pdf)
Department: Department of Electrical and Computer Engineering
Degree: Master of Science
Program: Electrical Engineering
Document Type: Thesis
Advisory Committee: Haimovich, Alexander (Committee chair)
Abdi, Ali (Committee member)
Kliewer, Joerg (Committee member)
Date: 2019-05
Keywords: Blind source separation
Dictionary learning
Doppler effect
Hidden Markov model
MIMO
Time coherence
Availability: Unrestricted
Abstract:

Distributed sensors observe radio frequency (RF) sources over flat-fading channels. The activity pattern is sparse and intermittent in the sense that while the number of latent sources may be larger than the number of sensors, only a few of them may be active at any particular time instant. It is further assumed that the source activity is modeled by a Hidden Markov Model. In previous work, the Blind Source Separation (BSS) problem solved for stationary channels using Dictionary Learning (DL). This thesis studies the effect of time-varying channels on the performance of DL algorithms. The performance metric is the probability of detection, where a correct detection is the event that the estimated value of a source exceeds a threshold at a time instant when the true source is active. Using the probability of detection when the channels are stationary as a baseline, it is shown that there is significant degradation for time-varying channels and observation intervals much longer than the time coherence. Detection performance improves when the observation time is approximately equal to the time coherence. Performance is again degraded when the observation is shorter and there is not sufficient information for the DL algorithms to learn from.


If you have any questions please contact the ETD Team, libetd@njit.edu.

 
ETD Information
Digital Commons @ NJIT
Theses and DIssertations
ETD Policies & Procedures
ETD FAQ's
ETD home

Request a Scan
NDLTD

NJIT's ETD project was given an ACRL/NJ Technology Innovation Honorable Mention Award in spring 2003