Browsing by Author "Vorobyov, Sergiy A. Prof., Aalto University, Department of Signal Processing and Acoustics, Finland"

Now showing 1 - 2 of 2

Advances and New Applications of Spectral Analysis
(2023) Esfandiari, Majdoddin
School of Electrical Engineering | Doctoral dissertation (article-based)
Spectral analysis is a mathematical tool for modeling signals and extracting information from signals. Among the areas where it finds applications are radar, sonar, speech processing, and communication systems. One of applications of spectral analysis is to model random signals with the so-called rational models like autoregressive (AR) model. Spectral analysis is used for estimating direction-of-arrival (DOA) in array processing as well. In addition, spectral analysis is used for channel estimation in communication systems such as massive/mmWave MIMO systems. It is because channels in these systems can be modeled with multipaths' gains and angles. This thesis proposes methods for the problems of noisy AR parameter estimation, DOA estimation in unknown noise fields, and massive/mmWave MIMO channel estimation and data detection with one-bit ADCs. The AR model offers a flexible yet simple tool for modeling complex signals. In practical scenarios, the observation noise may contaminate the AR signal. In this thesis, five methods for estimating noisy AR parameter estimation are proposed. In developing the methods, the concepts such as eigendecomposition (ED) and constrained minimization are used. The most common assumption about the structure of the observation noise in DOA estimation problem is the uniform noise assumption. According to it, the noise covariance matrix is a scaled identity matrix. However, other noise covariance matrix structures such as nonuniform or block-diagonal are more accurate in some practical situations. A generalized least-squares (GLS)-based DOA estimator that takes into consideration the signal subspace perturbation and also enjoys a properly designed DOA selection strategy is proposed for the case of uniform sensor noise. For the case of nonuniform noise, a non-iterative subspace-based (NISB) method is developed which is computationally efficient compared to state-of-the-art competitors. Moreover, a unified approach to DOA estimation in uniform, nonuniform, and block-diagonal sensor noise is presented. The use of one-bit ADCs instead of high-resolution ADCs is considered as an elegant solution for reducing power consumption of large-scale systems such as massive/mmWave MIMO and radar systems. In this thesis, we use the analogy between binary classification problem and one-bit parameter estimation to develop algorithms for massive MIMO and mmWave systems. In this regard, a method called SE-TMR is developed for one-bit mmWave UL channel estimation. Another method called L1-RLR-TMR is also offered for one-bit mmWave UL channel estimation. At last, the concept of AdaBoost is combined with Gaussian discriminant analysis (GDA) for developing computationally very efficient channel estimators and data detectors. It is shown that the proposed methods which use approximate versions of GDA as weak classifiers in iterations of the AdaBoost-based algorithms are exceptionally efficient, specifically in large-scale systems.
Methods for Convolutional Sparse Coding and Coupled Feature Learning with Applications to Image Fusion
(2023) G. Veshki, Farshad
School of Electrical Engineering | Doctoral dissertation (article-based)
The sparse approximation model, also known as the sparse coding model, represents signals as linear combinations of only a small number of elements (atoms) from a dictionary. This model is used in many applications of signal processing, machine learning, and computer vision. In many tasks, the use of dictionaries adapted to signal domains has led to significant improvements. The process of finding domain-adapted dictionaries is called dictionary learning. Structured sparse approximation and dictionary learning has been successfully used in applications such as image fusion, where it is required to find correlated patterns in multi-measure and multimodal signals. Image fusion is the problem of combining multiple images, for example, acquired using different imaging modalities, into a single, more informative image. A shift-invariant extension of the standard sparse approximation model that can describe the entire high-dimensional signals is referred to as convolutional sparse coding (CSC). It has been demonstrated in several studies that the CSC model is superior to its standard counterpart in representing natural signals such as audio and images. A majority of the leading CSC and CDL algorithms are based on the alternating direction method of multipliers (ADMM) and the Fourier transform. There is only one significant difference between these methods, which is in the way they address a convolutional least-squares regression subproblem. In this thesis, we propose a novel solution for this subproblem that improves the computational efficiency of the existing algorithms. Additionally, we present an efficient ADMM-based approximate online CDL algorithm that can be used in applications that require learning large dictionaries over high-dimensional signals. Next, we propose new methods and develop computationally efficient algorithms for learning correlated features (called coupled feature learning (CFL) in this thesis) in multi-measure and multimodal signals based on sparse approximation and dictionary learning. The presented CFL algorithms potentially apply to signal and image processing tasks where a joint analysis of multiple correlated signals (e.g., multimodal images) is essential. We also propose CSC-based extensions and variations of the proposed CFL algorithm. Based on the proposed CFL methods, we develop multimodal image fusion algorithms. Specifically, the learned coupled dictionary atoms, representing correlated visual features, are used to generate unified enhanced images. We address multimodal medical image fusion, infrared and visible-light image fusion, and near-infrared and visible-light image fusion problems. This thesis contains representative experimental results for all proposed algorithms. The effectiveness of the proposed algorithms is demonstrated based on comparisons with state-of-the-art methods.