Date of Award

1-1-2020

Language

English

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

College/School/Department

Department of Computer Science

Content Description

1 online resource (xii, 112 pages) : illustrations (some color)

Dissertation/Thesis Chair

Mariya Zheleva

Committee Members

Petko Bogdanov, Hany Elgala

Keywords

Radio resource management (Wireless communications), Cognitive radio networks, MIMO systems, Spectrum analysis, Modulation spectroscopy, Radio frequency allocation

Subject Categories

Computer Sciences

Abstract

With recent advances in emerging Dynamic Spectrum Access (DSA) and Cognitive Radio technologies, modulation recognition (ModRec) has emerged as a critical problem with importance to spectrum-sharing applications. Existing approaches, target modulation recognition as if a packet will be decoded in full and thus, pose stringent requirements on spectrum sensing and transmitter behavior: (i) a transmitter's bandwidth should be scanned alone and in full, (ii) for MIMO ModRec, the sensor should have at least same as many antennas as the transmitter, (iii) modulation symbol representation should be uniform and (iv) prior knowledge of the transmitter's technology should be available. These stringent requirements will not be readily met by future spectrum sensing infrastructures, which will use low-cost, sweep-based spectrum sensing to intermittently scan radio bands in support of DSA. Our research bridges the gap between modulation recognition requirements and future spectrum sensing capabilities. We design new signal features that retain information of the underlying modulation even when the sensing requirements are not met. We design lightweight, yet accurate machine learning frameworks around these features for accurate and future-proof modulation recognition.

Share

COinS