New Approaches in Nonlinear Analysis and Modeling of Multichannel Signals

In this thesis, we present novel machine-learning-driven approaches for nonlinear analysis and modeling of complex multichannel signals, focusing on two key applications: Electroencephalography (EEG) classification and analysis, and phase retrieval-based image reconstruction. EEG serves as an effective diagnostic tool for mental disorders and neurological abnormalities and improving its analysis can enhance classification performance. We propose a novel EEG data representation that leverages the spatial layout of sensors and preserves their topology to improve classification accuracy and computing cost of EEG analysis. Compared to traditional one-dimensional channel concatenation, our model consistently boosts accuracy by 5–8% across multiple machine learning algorithms in different EEG-based problems. The phase retrieval problem is an inverse problem which consists of recovering a constrained image from the magnitude of its Fourier transform. It’s fundamental in a variety of fields and imaging systems such as Crystallography, X-ray, optical imaging, and astronomical imaging. Although traditional algorithms such as the hybrid input-output (HIO) method, and continuous hybrid input–output (CHIO) are widely used to solve this problem, their reconstruction performance can be improved. We introduce a new hybrid model that consists of CHIO methods and deep neural networks to solve this inverse problem. The new model achieves better reconstruction performance with minimal additional computational cost.

Author dc:creator

• Lubna Shibly Mokatren (24400091)

• EEG Analysis with Spatial Correlation Using Deep Learning
• Phase Retrieval

Statement dc:rights

• In Copyright
• Open Access after 2028-05-01

DOI dc:identifier

https://doi.org/10.25417/uic.32995136.v1

OAI identifier oai:identifier

oai:figshare.com:article/32995136