Virginia Tech
Deep Learning Empowered Unsupervised Contextual Information Extraction and its applications in Communication Systems
Abstract
dc:description.abstractThere has been a rapid growth in the amount of data produced by wireless systems. With the advancements in 5G technologies, it has become essential to process this data efficiently. Specifically, when there is too much flow of information, there is a practical need to find useful information (i) for efficient utilization of communication resources and (ii) for better decision-making. This thesis investigates how to efficiently extract useful features from large-dimensional observations in wireless communication-oriented problems. Semantic communications reduce bandwidth consumption by transmitting only the important features of the input data. Unlike conventional communications, semantic communications aim to achieve accurate recovery of the intended meaning rather than bit-by-bit accuracy. Consequently, feature extraction is a crucial step for semantic communications. In this work, we develop a deep learning based contextual video information processing framework, called Kratos, that accurately extracts the critical and statistically significant information (such as enemies' locations on the battlefields) from the collected images and/or videos while suppressing the background information; thus enabling semantic communications in Internet-of-battlefield (IoBT) networks. Rapid advancements in radio technology have also spiked demand for spectrum resources burdening already limited spectrum availability. With the surge in malicious actors who want to exploit the spectrum, it has become essential to identify such intrusions so they can be rectified quickly. We develop a deep-learning based framework called WANDA to identify such anomalies. Different from the existing literature, Kratos and WANDA utilize the soft Hirschfeld-Gebelein-Renyi (HGR) correlation method to extract the important features from the multi-modal, high-dimensional and statistically correlated observations. For demonstration purposes, a battlefield video game is simulated, and Kratos is applied to recover the states and positions of the enemy spaceships, and WANDA was used to identify anomalies in a Wifi Signal spectrum. Our experiments show that Kratos can accurately track the states and positions of the enemies in real time, even in the presence of entirely unknown background information. Similarly, WANDA also performed significantly better than its counterparts in all SNR ranges maximizing the True Positives at a low cost of False Positives.
Degree
thesis:*- Name thesis:degree_name
- Master of Science
- Level thesis:degree_level
- masters
- Discipline thesis:degree_discipline
- Computer Science and Applications
- Department dc:contributor.department
- Computer Science and Applications
- Grantor dc:publisher
- Virginia Tech
- Year dc:date.issued
- 2023
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Gusain, Kunal
- Chairs dc:contributor.committeechair
-
- Reed, Jeffrey H.
- Lou, Wenjing
- Committee members dc:contributor.committeemember
-
- Ramakrishnan, Narendran
- Hasan, Shaddi Husein
- Shah, Vijay K.
Subjects
dc:subject × 9Rights
dc:rights- Statement dc:rights
-
- Creative Commons Attribution 4.0 International
- Licence dc:rights.uri
- Language dc:language.iso
- en
Identifiers
dc:identifier.*- Dc Identifier Other
- vt_gsexam:36386