Electrical and Computer Engineering ETDs

Publication Date

Spring 1-25-2021


Source localization with sensor arrays have found applications across domains beginning with radar and sonar, astronomy, acoustics, bio-medical devices and more recently in autonomous cars and adaptive communication systems. The knowledge of the spatial spectrum not only provide information about the source and interference but also assists in increasing signal integrity and avoid interference. This provides an added degree of freedom in the form of spatial diversity. This research investigates spatial spectrum estimation of waveforms from the signals sampled by arbitrarily distributed sensors. Conventional high resolution algorithms such as root-MuSiC fails to perform accurate source localization due to the reliance of the polynomial rooting technique on the Vandermonde structure of the array manifold. Spatial interpolation methods to estimate the uniform equivalent from the irregularly sampled signal has been widely used to address the irregularity in sampling. We investigate the limitations of the current approaches and propose more accurate and robust interpolation methods. We formulate a complex Gaussian Process (GP) regression formulation to address the complex variable in communication signals as an alternative to the existing methodology. We further explore support vector regression (SVR) and introduce a complex SVR formulation endowed with Mercer’s Kernels to perform spatial interpolation. From conventional machine learning we move on to investigate deep learning strategies and propose an encoder-decoder architecture for spatial interpolation and denoising. We integrate the interpolation methods with a high resolution localization technique in the form of root-MuSiC and develop novel algorithms to perform source localization with arrays of any shape and size. From two step methodologies of spatial interpolation and subsequent angle of arrival estimation we delve into data driven approaches and propose an end-to-end framework for source localization using a Long Short Term Memory (LSTM) architecture. The proposed framework is entirely data driven and can perform seamless source localization with sensor arrays of any given geometry.


Machine Learning, Deep Learning, DOA estimation, Array Signal Processing, LSTM, SVM, Gaussian Processes, Signal Processing

Document Type




Degree Name

Electrical Engineering

Level of Degree


Department Name

Electrical and Computer Engineering

First Committee Member (Chair)

Prof. Christos G. Christodoulou

Second Committee Member

Prof. Manel Martinez Ramon

Third Committee Member

Prof. Jose Luis Rojo-Alvarez

Fourth Committee Member

Dr. David Murrell