Theses & Dissertations

Abstract

This thesis addresses the fundamental challenge of reliable, high-spectral- efficiency communication in high-mobility wireless environments, where doubly-dispersive channels cause severe inter-carrier interference that renders conventional orthogonal frequency division multiplexing (OFDM) inadequate. The work develops novel index modulation frameworks built upon affine frequency division multiplexing (AFDM), a chirp-based waveform that achieves full delay-Doppler diversity with significantly lower pilot overhead than orthogonal time frequency space (OTFS). Several distinct index modulation architectures are proposed, including enhanced spatial modulation for OTFS, generalized dual-mode index modulation for OTFS, and multiple AFDM-based variants encompassing generalized, quadrature, dual-mode, and multi-mode index modulation schemes. Each architecture exploits additional information dimensions beyond conventional QAM modulation to progressively improve spectral efficiency.

Abstract

This research proposes a machine learning framework to improve the classification and interpretation of consciousness levels using rs-EEG. Variational mode decomposition (VMD) is employed to enhance feature extraction, followed by a stacking ensemble regressor integrating gradient boosting, random forest, and support vector regression. Shapley additive explanations (SHAP) ensured interpretability by identifying key predictive features, supporting a simplified linear model aligned with the coma recovery scale–revised (CRS-R) score. Functional connectivity analyses employing amplitude envelope correlation and related measures were utilized to guide deep learning models and graph-theoretical assessments, thereby establishing associations between network topology and glasgow outcome scale–extended (GOSE) scores. These results contribute to the development of interpretable, data-driven methodologies for enhancing diagnostic precision and prognostic evaluation in disorders of consciousness.