Date of Award

2022-08-01

Degree Name

Doctor of Philosophy

Department

Computational Science

Advisor(s)

Virgilio Gonzalez

Abstract

Long Term Evolution or LTE has gained interest for new applications that can benefit society including mobile broadband services like 802.11 or Wi-Fi. Both LTE and Wi-Fi uses similar modulation technique Orthogonal Frequency Division Multiplexing or OFDM. 6GHz sub carrier bands are crowded with LTE users. As wireless communications technology continues to develop, LTE technology in unlicensed bands (LTE-U) is a viable solution to the lack of spectrum resources. The competition between LTE-U and Wi-Fi will seriously impair their communication quality, so the friendly coexistence of both become an important research topic. This paper discusses the use of a software-defined radio (SDR) testbed at UTEP in order to rapidly prototype and classify radio frequency (RF) signals using deep learning (DL) techniques with validation accuracy as high as 96.67%. SDR testbed data is processed and fed into the Convolutional Neural Network (CNN), which performs feature extraction and trains the network to classify RF signals. The proposed method differentiates LTE-U and Wi-Fi signals effectively and allows them to coexist. The spectrum sensing function plays a key role in the coexistence. Several transfer learning algorithms are tested to increase the performance of the classification and to minimize the loss probability. The CNN extract features from the observations belonging to different class of RF signals for training, and finally validates the training set. The performance of the proposed transfer learnings were tested over the air using SDRs for variable signal-to-noise ratio with noise uncertainty. The UTEP SDR testbed is unique in several ways, including the extensive use of SDR technology, the use of industry-grade hardware and software-based systems, and the ability to design experiments in accordance with the user's preferences.

Language

en

Provenance

Recieved from ProQuest

File Size

110 p.

File Format

application/pdf

Rights Holder

Mirza Mohammad Maqbule Elahi

Share

COinS