Off-campus UMass Amherst users: To download campus access dissertations, please use the following link to log into our proxy server with your UMass Amherst user name and password.

Non-UMass Amherst users: Please talk to your librarian about requesting this dissertation through interlibrary loan.

Dissertations that have an embargo placed on them will not be available to anyone until the embargo expires.

Author ORCID Identifier

https://orcid.org/0000-0002-4982-3433

AccessType

Campus-Only Access for Five (5) Years

Document Type

dissertation

Degree Name

Doctor of Philosophy (PhD)

Degree Program

Electrical and Computer Engineering

Year Degree Awarded

2023

Month Degree Awarded

February

First Advisor

Ramakrishna Janaswamy

Second Advisor

Robert W. Jackson

Third Advisor

Marco Duarte

Fourth Advisor

Pamela Loring

Subject Categories

Electromagnetics and Photonics | Signal Processing

Abstract

Antenna arrays are applied widely in our everyday life, they play an important role in satellite communication, self-driving car, radar and so on. In this dissertation, we will apply antenna array techniques to track movement of the birds as well as estimate locations of RF emitters.

In the first part, an antenna array is deployed to track the movement of birds. We implemented a static method which ignores the interrelation between detection samples and a dynamic method which includes the movement model and measurement model to track the bird. The movement model describes the change in bird's position with respect to time and the measurement model links the observed measurement to the bird's position to improve prediction. In the movement model, several parameters are used to describe the movement of the avian target. However the values of these parameters are difficult to estimate because they are dependent on many factors, such as the species of the bird, the weather condition, the flying area and so on. In order to estimate the values of these parameters, one method which is based on the maximum likelihood estimation is proposed in this dissertation. With the measurement model, the prior estimate for the bird from the movement model can be updated with the given measurement. However the measurement is corrupted with the environment noise and subject to device non-linearities. This will introduce errors in the position we want to estimate. In this dissertation, an unscented transformation is used in a Kalman filter to predict and update the estimation of the bird’s position for improving the prediction accuracy.

In the second part, incoming radio signals are estimated using a linear antenna array. The position of the elements in the array is assumed to be known and we want to estimate the incoming signals via the received signals. We start with exploring the possibility of solving the direction of arrival (DoA) problem as a multi-class classification problem. However, if the incoming signal are from directions simultaneously, the classifier does not work well and the training data set is very large for the multiple sources case. Inspired by the concept of extracting feature in the image classification problem, we also tried to use the back propagation neural network to extract the feature of the received signals during the training period; the result of using the neural network is presented in the dissertation. We also convert the DoA problem into the regression problem. The relevance vector machine (RVM) has previously been applied for the direction of arrival (DoA) estimation by Carlin et al. In this dissertation, in order to improve the efficiency and estimation accuracy of the start-of-the-art method, a Bayesian learning with Laplace prior is first applied for the DoA problem. Compared with using regular RVM, one advantage of using Laplace prior is that less iterations are required to get the estimation results. In addition, the relationship between the real and imaginary part of the incoming signals is taken into account via the proposed pairwise learning algorithm, which can improve the accuracy of estimation and further improve the estimation efficiency. Then, in order to improve the robustness of our method in the noisy environment, we extend our method to the situation with multiple snapshots. In each snapshot, the sources are fixed but the locations of receivers could change with time. Our method is also applied with the adaptive grid, we present the comparison for the estimation result between the on-grid and adaptive grid. Furthermore, we derive the Cram\'er-Rao lower bounds (CRLB) of the proposed DoA estimation method, the CRLBs are derived under different scenarios: (i) if the unknown parameters consist of deterministic and random variables, a hybrid CRLB is derived; (ii) if all the unknown parameters are random, a Bayesian CRLB is derived; and the marginalized Bayesian CRLB is obtained by marginalizing out the nuisance parameter. We present the simulation results of our proposed method under different conditions such as various of number of receiving element in the antenna array, different number of snapshots and the various signal to noise ratio (SNR).

DOI

https://doi.org/10.7275/33072671

Share

COinS