The current and future trends of modern wireless communication systems place heavy demands on fast data transmissions in order to satisfy end users’ requirements anytime, anywhere. Such demands are obvious in recent applications such as smart phones, long term evolution (LTE), 4 & 5 Generations (4G & 5G), and worldwide interoperability for microwave access (WiMAX) platforms, where robust coding and modulations are essential especially in streaming on-line video material, social media and gaming. This eventually resulted in extreme exhaustion imposed on the frequency spectrum as a rare natural resource due to stagnation in current spectrum management policies. Since its advent in the late 1990s, cognitive radio (CR) has been conceived as an enabling technology aiming at the efficient utilisation of frequency spectrum that can lead to potential direct spectrum access (DSA) management. This is mainly attributed to its internal capabilities inherited from the concept of software defined radio (SDR) to sniff its surroundings, learn and adapt its operational parameters accordingly. CR systems (CRs) may commonly comprise one or all of the following core engines that characterise their architectures; namely, adaptive coded modulation (ACM), automatic modulation classification (AMC) and spectrum sensing (SS). Motivated by the above challenges, this programme of research is primarily aimed at the design and development of new paradigms to help improve the adaptability of CRs and thereby achieve the desirable signal processing tasks at the physical layer of the above core engines. Approximate modelling of Rayleigh and finite state Markov channels (FSMC) with a new concept borrowed from econometric studies have been approached. Then insightful channel estimation by using Kalman filter (KF) augmented with interacting multiple model (IMM) has been examined for the purpose of robust adaptability, which is applied for the first time in wireless communication systems. Such new IMM-KF combination has been facilitated in the feedback channel between wireless transmitter and receiver to adjust the transmitted power, by using a water-filling (WF) technique, and constellation pattern and rate in the ACM algorithm. The AMC has also benefited from such IMM-KF integration to boost the performance against conventional parametric estimation methods such as maximum likelihood estimate (MLE) for channel interrogation and the estimated parameters of both inserted into the ML classification algorithm. Expectation-maximisation (EM) has been applied to examine unknown transmitted modulation sequences and channel parameters in tandem. Finally, the non-parametric multitaper method (MTM) has been thoroughly examined for spectrum estimation (SE) and SS, by relying on Neyman-Pearson (NP) detection principle for hypothesis test, to allow licensed primary users (PUs) to coexist with opportunistic unlicensed secondary users (SUs) in the same frequency bands of interest without harmful effects. The performance of the above newly suggested paradigms have been simulated and assessed under various transmission settings and revealed substantial improvements.

Radio Frequency Machine Learning (RFML) is the application of ML techniques to solve problems in the RF domain as an alternative to traditional digital-signal processing (DSP) techniques. Notable among these are the tasks of specific emitter identification (SEI), determining source identity of a received RF signal, and automated modulation classification (AMC), determining the modulation scheme of a received RF transmission. Both tasks have a number of algorithms that are effective on simulated data, but struggle to generalize to data collected in the real-world, partially due to the lack of available datasets upon which to train models and understand their limitations. This thesis covers the practical considerations for systems that can create high-quality datasets for RFML tasks, how variances from real-world effects in these datasets affect RFML algorithm performance, and how well models developed from these datasets are able to generalize and adapt across different receiver hardware platforms. Moreover, this thesis presents a proof-of-concept system for large-scale and efficient data generation, proven through the design and implementation of a custom platform capable of coordinating transmissions from nearly a hundred Software-Defined Radios (SDRs). This platform was used to rapidly perform experiments in both RFML performance sensitivity analysis and successful transfer between SDRs of trained models for both SEI and AMC algorithms. / Master of Science / Radio Frequency Machine Learning (RFML) is the application of machine learning techniques to solve problems having to do with radio signals as an alternative to traditional signal processing techniques. Notable among these are the tasks of specific emitter identification (SEI), determining source identity of a received signal, and automated modulation classification (AMC), determining the data encoding format of a received RF transmission. Both tasks have practical limitations related to the real-world collection of RF training data. This thesis presents a proof-of-concept for large-scale, efficient data generation and management, as proven through the design and construction of a custom platform capable of coordinating transmissions from nearly a hundred radios. This platform was used to rapidly perform experiments in both RFML performance sensitivity analysis and successful cross-radio transfer of trained behaviors.

With the increase in spectrum congestion, intelligent spectrum sensing systems have become more important than ever before. In the field of Radio Frequency Machine Learning (RFML), techniques like deep neural networks and reinforcement learning have been used to develop more complex spectrum sensing systems that are not reliant on expert features. Architectures like Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN) have shown great promise for applications like automated modulation classification, signal detection, and specific emitter ID. Research in these areas has primarily focused on "one size fits all" networks that assume a fixed signal length in both training and inference. However, since some signals are more complex than others, due to channel conditions, transmitter/receiver effects, etc., being able to dynamically utilize just enough of the received symbols to make a reliable decision allows for more efficient decision making in applications such as electronic warfare and dynamic spectrum sharing. Additionally, the operator may want to get to the quickest possible decision. Recurrent neural networks have been shown to outperform other architectures when processing temporally correlated data, such as from wireless communication signals. However, compared to other architectures, such as CNNs, RNNs can suffer from drastically longer training and evaluation times due to their inherent sample-by-sample data processing. While traditional usage of both of these architectures typically assumes a fixed observation interval during both training and testing, the sample-by-sample processing capabilities of recurrent neural networks opens the door for "decoupling" these intervals. This is invaluable in real-world applications due to the relaxation of the typical requirement of a fixed time duration of the signals of interest. This work illustrates the benefits and considerations needed when "decoupling" these observation intervals for spectrum sensing applications. In particular, this work shows that, intuitively, recurrent neural networks can be leveraged to process less data (i.e. shorter observation intervals) for simpler inputs (less complicated signal types or channel conditions). Less intuitively, this works shows that the "decoupling" is dependent on appropriate training to avoid bias and insure generalization. / Master of Science / With the increase in spectrum congestion, intelligent spectrum sensing systems have become more important than ever before. In the field of Radio Frequency Machine Learning (RFML), techniques like deep neural networks and reinforcement learning have been used to develop more complex spectrum sensing systems that are not reliant on expert features. Architectures like Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN) have shown great promise for applications like automated modulation classification, signal detection, and specific emitter ID. Research in these areas has primarily focused on "one size fits all" networks that assume a fixed signal length in both training and inference. However, since some signals are more complex than others, due to channel conditions, transmitter/receiver effects, etc., being able to dynamically utilize just enough of the received symbols to make a reliable decision allows for more efficient decision making in applications such as electronic warfare and dynamic spectrum sharing. Additionally, the operator may want to get to the quickest possible decision. Recurrent neural networks have been shown to outperform other architectures when processing temporally correlated data, such as from wireless communication signals. However, compared to other architectures, such as CNNs, RNNs can suffer from drastically longer training and evaluation times due to their inherent sample-by-sample data processing. While traditional usage of both of these architectures typically assumes a fixed observation interval during both training and testing, the sample-by-sample processing capabilities of recurrent neural networks opens the door for "decoupling" these intervals. This is invaluable in real-world applications due to the relaxation of the typical requirement of a fixed time duration of the signals of interest. This work illustrates the benefits and considerations needed when "decoupling" these observation intervals for spectrum sensing applications. In particular, this work shows that, intuitively, recurrent neural networks can be leveraged to process less data (i.e. shorter observation intervals) for simpler inputs (less complicated signal types or channel conditions). Less intuitively, this works shows that the "decoupling" is dependent on appropriate training to avoid bias and insure generalization.

One of the fundamental hurdles in realizing new spectrum sharing allocation policies is that of reliable spectrum sensing. In this thesis, three research thrusts are presented in order to further research in this critical area. The first of these research thrusts is the development of a novel asynchronous and noncoherent modulation classifier for PSK/QAM modulated signals in flat-fading channels. In developing this classifier, a novel estimator for the unknown channel gain and fractional time delay is proposed which uses a method-of-moments based estimation approach. For the second research thrust of this thesis, the developed method-of-moments based estimation approach is extended to estimate the signal-to-noise ratio of PSK/QAM modulated signals in flat-fading channels, in which no a priori knowledge of the modulation format and channel parameters is assumed. Finally, in the third research thrust, a distributed spectrum sensing approach is proposed in which a network of radios collaboratively detects the presence, as well as the modulation scheme, of a signal through the use of a combination of cyclic spectrum feature-based signal classification and an iterative algorithm for optimal data fusion. / Master of Science

One component of wireless communications of increasing necessity in both civilian and military applications is the process of automatic modulation classification. Modulation of a detected signal of unknown origin requiring interpretation must first be determined before the signal can be demodulated. This thesis presents a novel architecture for a modulation classifier that determines the most likely modulation using Grey Relational Analysis with the extraction and combination of multiple signal features. An evaluation of data preprocessing methods is conducted and performance of the classifier is investigated with the addition of each new signal feature used for classification. / Master of Science

Sensor Fusion is an essential framework in many Engineering fields. It is a relatively new paradigm for integrating data from multiple sources to synthesize new information that in general would not have been feasible from the individual parts. Within the wireless communications fields, many emerging technologies such as Wireless Sensor Networks (WSN), the Internet of Things (IoT), and spectrum sharing schemes, depend on large numbers of distributed nodes working collaboratively and sharing information. In addition, there is a huge proliferation of smartphones in the world with a growing set of cheap powerful embedded sensors. Smartphone sensors can collectively monitor a diverse range of human activities and the surrounding environment far beyond the scale of what was possible before. Wireless communications open up great opportunities for the application of sensor fusion techniques at multiple levels. In this dissertation, we identify two key problems in wireless communications that can greatly benefit from sensor fusion algorithms: Automatic Modulation Classification (AMC) and indoor localization and mapping based on smartphone sensors. Automatic Modulation Classification is a key technology in Cognitive Radio (CR) networks, spectrum sharing, and wireless military applications. Although extensively researched, performance of signal classification at a single node is largely bounded by channel conditions which can easily be unreliable. Applying sensor fusion techniques to the signal classification problem within a network of distributed nodes is presented as a means to overcome the detrimental channel effects faced by single nodes and provide more reliable classification performance. Indoor localization and mapping has gained increasing interest in recent years. Currently-deployed positioning techniques, such as the widely successful Global Positioning System (GPS), are optimized for outdoor operation. Providing indoor location estimates with high accuracy up to the room or suite level is an ongoing challenge. Recently, smartphone sensors, specially accelerometers and gyroscopes, provided attractive solutions to the indoor localization problem through Pedestrian Dead-Reckoning (PDR) frameworks, although still suffering from several challenges. Sensor fusion algorithms can be applied to provide new and efficient solutions to the indoor localization problem at two different levels: fusion of measurements from different sensors in a smartphone, and fusion of measurements from several smartphones within a collaborative framework. / Ph. D. / Sensor Fusion is an essential paradigm in many Engineering fields. Information from different nodes, sensing various phenomena, is integrated to produce a general synthesis of the individual data. Sensor fusion provides a better understanding of the sensed phenomenon, improves the application or system performance, and helps overcome noise in individual measurements. In this dissertation we study some sensor fusion applications in wireless communications: (i) cooperative modulation classification and (ii) indoor localization and mapping at different levels. In cooperative modulation classification, data from different wireless distributed nodes is combined to generate a decision about the modulation scheme of an unknown wireless signal. For indoor localization and mapping, measurement data from smartphone sensors are combined through Pedestrian Dead Reckoning (PDR) to re-create movement trajectories of indoor mobile users, thus providing high-accuracy estimates of user’s locations. In addition, measurements from collaborating users inside buildings are combined to enhance the trajectories’ estimates and overcome limitations in single users’ system performance. The results presented in both parts of this dissertation in different frameworks show that combining data from different collaborative sources greatly enhances systems’ performances, and open the door for new and smart applications of sensor fusion in various wireless communications areas.

Recently, automatic modulation classification techniques using deep neural networks on raw IQ samples have been investigated and show promise when compared to more traditional likelihood-based or feature-based techniques. While likelihood-based and feature-based techniques are effective, making classification decisions directly on the raw IQ samples removes the need for expertly crafted transformations and feature extractions. In practice, RF environments are typically very dense, and a receiver must first detect and isolate each signal of interest before classification can be performed. The errors introduced by this detection and isolation process will affect the accuracy of deep neural networks making automatic modulation classification decisions directly on raw IQ samples. The importance of defining upper limits on estimation errors in a detector is highlighted, and the negative effects of over-estimating or under-estimating these limits is explored. Additionally, to date, most of the published research has focused on synthetically generated data. While large amounts of synthetically generated data is generally much easier to obtain than real-world signal data, it requires expert knowledge and accurate models of the real world, which may not always be realistic. The experiments conducted in this work show how augmented real-world signal captures can be successfully used for training neural networks used in automatic modulation classification on raw IQ samples. It is shown that the quality and duration of real world signal captures is extremely important when creating training datasets, and that signal captures made from a single transmitter with one receiver can be broadly applicable to other radios through dataset augmentation. / Master of Science / With the increasing prevalence of wireless devices in every day life, communicating between them can become more difficult because the devices must contend with each other to send and receive information. Being able to communicate in a variety of environments can be challenging and, while devices can be pre-configured for certain situations, devices that are able to automatically adjust how they communicate are more reliable and robust. The research presented in this thesis will contribute to solving this challenge by considering machine-learning based, radio frequency signal processing algorithms that are able to automatically group different communication signals. Being able to automatically group different signals is helpful because it can provide information about the wireless environment, allowing a device to make intelligent decisions based on what it detects is happening around it. However, before these algorithms can be successfully used in wireless devices, their limitations must be better understood. To this end, the work in this thesis will show how sensitive these algorithms are to imperfections in wireless devices. This work will also show how information from new environments can be captured and manipulated to allow these algorithms to scale for unseen environments and communication signals.

No description available.

No description available.

Source localization with a wireless sensor network remains an important area of research as the number of applications with this problem increases. This work considers the problem of source localization by a network of passive wireless sensors. The primary means by which localization is achieved is through direction-finding at each sensor, and in some cases, range estimation as well. Both single and multiple-target scenarios are considered in this research. In single-source environments, a solution that outperforms the classic least squared error estimation technique by combining direction and range estimates to perform localization is presented. In multiple-source environments, two solutions to the complex data association problem are addressed. The first proposed technique offers a less complex solution to the data association problem than a brute-force approach at the expense of some degradation in performance. For the second technique, the process of signal classification is considered as another approach to the data association problem. Environments in which each signal possesses unique features can be exploited to separate signals at each sensor by their characteristics, which mitigates the complexity of the data association problem and in many cases improves the accuracy of the localization. Two approaches to signal-selective localization are considered in this work. The first is based on the well-known cyclic MUSIC algorithm, and the second combines beamforming and modulation classification. Finally, the implementation of a direction-finding system is discussed. This system includes a uniform circular array as a radio frequency front end and the universal software radio peripheral as a data processor. / Master of Science