Search

The rise in the use of wireless communication has led to the problem of spectrum scarcity in licensed bands. The popularity of Internet of Things (IoT) requires innovative solutions that maximize the use of available spectrum to support the increasing number of connected devices. This thesis tackles two significant problems in wireless communication: the need for efficient spectrum sensing techniques and the scarcity of large, diverse raw in-phase (I) and quadrature (Q) datasets. Adviser: Mehmet Can Vuran The ability to detect and classify modulation of the signals efficiently can enable a cognitive radio to monitor the spectrum activity in real time and utilize unused frequencies. In this work, we propose an end-to-end framework to detect and classify narrowband signals from wideband IQ samples. The end-to-end framework reduces the sensing time for narrowband signals by 10 times. The training strategy proposed in the framework to train modulation classifiers on extracted signals increases the classification accuracy from 41% to 99%. Given the dynamic nature of wireless channels, comprehensive datasets gathered from various environments are crucial for the success of data-driven solutions. To tackle this issue, we have designed a systematic process to collect and centralize large-sized wireless samples by using the Nebraska Experimental Testbed of Things (NEXTT) testbed. The result is an accumulation of 123 TB of raw IQ samples across a frequency range of 54 to 2,590 MHz from three distinct wireless environments.

This thesis proposes a technique for blind modulation classification for spectrum sensingapplications by leveraging second-order cyclostationarity and frequency domain characteristics of signals. A detailed analysis of different cyclostationary features of commonly used digital modulations using the Strip Spectral Correlation Analyzer (SSCA) is carried out. Based on the one-dimensional cyclic features derived from the SSCA estimates, a hierarchical decision tree-based classifier is developed. The algorithm also extracts important signal parameters such as symbol rate. The proposed method follows a pre-processing pipeline similar to a practical spectrum sensing system, and its accuracy is rigorously evaluated with respect to Signal to Noise Ratio and Energy to Noise Ratio of the signals in both simulated and over-the-air data scenarios.

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.

The growing need for connectivity is a significant challenge to fulfill as the spectrum is a limited resource. Underutilized spectrum such as TV can be used to address this challenge. By employing an opportunistic technology where unlicensed users can make use of the channel in the absence of the licensed user, can make an enormous difference. Cognitive Radio (CR) has been invented from this idea of sharing the channel. CR Network (CRN) can make it possible to utilize the spectrum to t’s full potential.Along with over the air communication, deep space communication also looking into the CRN technology to maximize the channel utilization. NASA had deployed a testbed known as SCaN testbed attached to the International Space Station, to accommodate researches on deep space communications. In deep communication lower scheme modulations are used because of the low SNR value to maintain the signal quality. Reconfigurable radio technology like CRN can be used to increase the throughput by changing the modulation scheme w.r.t better SNR value. Whenever, the SNR value for the channel improves, the radio can change the modulation to improve the channel throughput. However, CRN comes with all the traditional wireless security threats and some additional security challenges such as PUEA and SSDF that exploits the vulnerabilities of Spectrum Sensing and cognitive capability of the CRN. To ensure a robust technology to use in the space communication, these challenges need to be addressed and mitigation techniques need to implemented. PUEA is such an attack where an adversary emulates the primary signal and deprives other SUs from discovering the channel. In a CRN, it is important to identify the signal for threat assessment which can help the ICS to address the PUEA attack by taking security measures. In our work, we are exploring the signal identification process by providing studies on modulation classification.As a part of funded project, we got access to captured data from NASA SCaN testbed Ka-Band channel. This dataset contains two types of modulation scheme, which we have processed to classify with Deep learning methods like Feed Forward fully connected MLP Network and Convolutional Neural Network (CNN). To extract different features from the raw complex datset we have transformed it to three other datasets and trained them with two different model. We compared the results to find which feature works better with which deep learning model. MLP resulted in an F-1 score of 0.96 and 0.95 for IQ and FFT representation of the complex dataset, where CNN resulted in 0.97 and 0.98, respectively.

Signal detection, parameter estimation and modulation classification are widely applied to many areas and plays a very important role in civilian and military, such as bio-science, criminal psychology, communication engineering, radar system, electronic warfare and so on. In the civilian field, with the increasing number of wireless electronic devices and higher transmission data rate demand, the problem of spectrum congestion becomes more and more highlighted and urgent. In recent years, wireless industry has shown great interest in Cognitive Radio (CR) and Dynamic Spectrum Access (DSA) network, whose primary function is to use limited frequency bands to transmit own signals without any interference with other primary users. Hence, the accuracy of signal detection and parameters estimation are particularly important and can provide reliable communication performance for cognitive radio users. In the military field, electronic warfare is crucial important part in modern war, such as own signal needs to be hidden, securely transmitted and received, enemy’s signals need to be identified, located and jammed. Thus, in such a non-cooperative environment, signal detection, parameter estimation and modulation classification technologies become more and more important and challenging. In the past few decades, several signal detection methods have been proposed, such as energy-based detection, matched filter-based detection and cyclostationary feature based detection. Energy based detection is simple to implement, but poorly performing at low SNR. Although the matched filter-based detection is the optimal detector, it needs to accurately know the prior information of the detected signal. Hence, matched filter-based detection is impractical to implement in real environment, such as non-cooperative environment. Cyclostationary feature based signal detection has high computational complexity, but it can be used for high-precision signal detection in low SNR environments. In recent years, there are many researchers show their interest and effort in signal detection, parameter estimation and modulation classification technologies. Most of them are working with single signal detection, parameter estimation and modulation classification. A few people consider time and frequency mixed signals as their target signals. In particular, some people assume that there is no any overlap between co-exited signals in time domain and frequency domain. In such case, we can easily separate those co-existed signals with a band-pass filter in frequency domain. Meanwhile, we can easily know the number of co-existed signals, estimate each signal’s parameters and classify their modulation types. However, in a spectrum congested environment, such as cognitive radio and electronic warfare, several signals are often mixed together with plenty of overlap in both time domain and frequency domain. In some special case, several signals are entirely overlapped in time domain and frequency domain, such as in-band full duplex communication signals. It is more challenge to enumerate and classify those kinds of mixed signals. Hence, studying mixed signal detection, parameter estimation and modulation classification is more practical significance. In this dissertation, we employ signal energy-based, mainly employ signal cyclostationary feature and machine learning technology-based methods to detect, estimate and classify mixed signal, which have significant overlap in both time domain and frequency domain. In particular, we employ energy-based detection to preliminary detect the signal is existed or not existed in the channel and use spectrum analysis roughly locate the interesting frequency band. Meanwhile, we employ different order signal cyclostationary features to detect, estimate and classify four popular digital communication signals, which includes low-order modulation type BPSK signal, high-order modulation type QPSK signal, 8-PSK signal and 16-QAM signals. According to our previous work, we can use second-order cyclostationary feature to detect and classify mixed signals, such as mixed BPSK signals, mixed QPSK signals. However, since some signals have no second-order cyclostationary feature, we unable to precisely estimate and classify them by using low order cyclostationary feature, such as we cannot use Spectral Correlation Function (SCF) to classify mixed QPSK signal and 16-QAM signal. So, in this dissertation, we consider some more challenged cases, include detecting, estimating and classifying mixed higher-order modulation signals, such as 16-QAM and 8-PSK signals, classifying mixed signals, which have similar cyclostationary features, such as QPSK and 16-QAM mixed signals, and analyze heavily overlapped mixed signals, such as two signals have same carrier frequency. Moreover, we employ low-order and high order cyclostationary features, i.e., cyclic moment and cyclic cumulants, to detect, estimate and classify more different combinations of mixed signals, such as BPSK and QPSK mixed signal, two QPSK mixed signal, BPSK and 16-QAM mixed signal, etc. In this dissertation, we also provide detailed performance analysis to demonstrate our proposed method can effectively detect mixed signals, estimate mixed signals’ parameters, such as carrier frequency, symbol rate and power, and classify mixed signals’ modulation types. In addition, our performance analysis is based on AWGN channel, flat fading channel and multi-path fading channels.

Deep learning is a subset of machine learning that contributes widely to the contemporary success of artificial intelligence. The essential idea of deep learning is to process complex data by abstracting hierarchical features via deep neural network structure. As one type of deep learning technique, deep belief network (DBN) has been widely used in various application fields. This thesis proposes an approximation based hardware realization of DBNs that requires low hardware complexity. This thesis also explores a set of novel applications of the DBN-based classifier that will benefit from a fast implementation of DBN. In my work, I have explored the application of DBN in the fields of automatic modulation classification method for cognitive radio, Doppler radar sensor for detection and classification of micro unmanned aerial systems, cyber security applications to detect false data injection (FDI) attacks and localize flooding attacks, and applications in social networking for prediction of link properties.The work in this thesis paves the way for further investigation and realization of deep learning techniques to address critical issues in various novel application fields.

The task of spectrum sensing, defined here to consist of signal detection, signal parameter estimation, and signal identification, is a critically important task in a wide-variety of wireless communication applications. For example, in recent years, government and research initiatives have proposed the idea of communication systems that could gain access to spectrum opportunistically when being unused by primary licensed spectrum users. In order for these opportunistic systems to be realizable, methods by which secondary spectrum users can detect and classify these primary users will be necessary. Furthermore, detection and classification among the secondary users themselves will be important for efficient spectrum usage in these systems. As another example, spectrum sensing is also of critical importance in many military applications. This is due to the inherent expectation that a priori information of hostile wireless systems will be minimal or unavailable. The goal of this dissertation is to provide both insight and solutions in the critical area of spectrum sensing. More specifically, the research contained within this dissertation deals with the development and analysis of spectrum sensing algorithms that address key issues related to channel and radio impairments that are at present underdeveloped in the literature. First, research is presented on a method-of-moments based signal parameter estimation and likelihood-based modulation classification approach for linear digital amplitude-phase modulated signals (PAM, PSK, QAM, ...) in slowly-varying flat-fading channels. Based on this work, research is then presented on a feature-based modulation classification approach which relaxes the requirements of perfect frequency synchronization and knowledge of the phase information of the received signal that the likelihood-based approach requires. Finally, research is presented on the impact that both sensor reliability and sensor correlation information have on collaborative signal detection and intelligent sensor selection.

This paper proposes a cumulant based classification means to identify waveforms for a blind receiver in the presence of time varying channels, which is built from the work done on cumulants in static channels currently in the literature. Results show the classification accuracy is on the order or better than current methods in use in static channels that do not vary over an observation period. This is accomplished by making use of second through tenth order cumulants in a signature vector that the search engine platform has the means of differentiating. A receiver can then blindly identify waveforms accurately in the presence of multipath Rayleigh fading with AWGN noise. Channel learning occurs prior to classification in order to identify the consistent distortion pattern for a waveform that is observable in the signature vector. Then using a database look-up method, the observed waveform is identified as belonging to a particular cluster based on the observed signature vector. If the distortion patterns are collected from a variety of channel types, the database can then classify both the waveform and the rough channel type that the waveform passed through. If the exact channel model or channel parameters is known and used as a limiter, significant improvement on the waveform classification can be achieved. Greater accuracy comes from using the exact channel model as the limiter.

The exponential growth in demand for mobile data has led to significant research efforts aimed at more efficient methods of utilizing the scarce RF spectrum resource. One potential solution to this scarcity problem is Cognitive Radio (CR) which involves dynamic spectrum access in which a set of unlicensed users occupy spectrum holes without causing significant degradation of performance to the incumbent users. A key enabling technology for CR networks is accurate spectrum sensing which aims to learn the radio environment in order to adapt the CR transmission. Traditional spectrum sensing techniques have mainly focused on determining only the presence or absence of a licensed user. Recent work in the past few years have shown however that more detailed knowledge pertaining to radio-scene analysis can be used to improve the performance of CR networks. The more the secondary user knows about the active licensed users, the better it can adapt its transmission strategies. In this work, we put forward the concept of advance spectrum sensing which takes a multi-dimensional approach to radio-scene analysis that estimates various parameters of the active transmitters through sensing, localization and tracking, modulation classification, PHY parameter estimation, and MAC-layer classification. In this work we investigate the elements of such an advance spectrum sensing system. Firstly, we will look at the problem of conventional spectrum sensing, or detecting the presence or absence of transmitting Primary Users. In particular, we study how detection performance could be improved through the use of cyclostationary feature detection and how it could be made robust to fading, noise uncertainty, and co-channel interferers through the optimal use of multiple sensors. Second, we attack the problem of modulation classification which we argue is a critical piece of information for future cognitive radio systems. We present a new type of pattern classification algorithm based on the concept of sampled distribution distance which offers a low computational complexity alternative to maximum-likelihood based classification. Through our extensive analysis, we have derived the optimal form of this type of classifier and applied it to the modulation classification problem. Finally, we propose a system of MAC-layer classification based on 4th-order cumulants which distinguishes between TDMA, OFDMA, CDMA and contention-based schemes. In addition, it is also able to jointly perform modulation classification with channel access method. The analysis of the statistics of the 4th-order cumulant used in our work also offers large potential for applications in different areas including modulation classification, channel estimation, and estimation of number of users.

Signal detection and modulation classification are becoming increasingly important in a variety of wireless communication systems such as those involving spectrum management and electronic warfare and surveillance, among others. The majority of the signal detection and modulation classification algorithms available in the literature assume that the additive noise has a Gaussian distribution. However, while this is a good model for thermal noise, various studies have shown that the noise experienced in most radio channels, due to a variety of man-made and natural electromagnetic sources, is non-Gaussian and exhibits impulsive characteristics. Unfortunately, conventional signal processing algorithms developed for Gaussian noise conditions are known to perform poorly in the presence of non-Gaussian noise. For this reason, the main goal of this dissertation is to develop statistical signal processing algorithms for the detection and modulation classification of signals in radio channels where the additive noise is non-Gaussian. One of the major challenges involved in the design of these algorithms is that they are expected to operate with limited or no prior knowledge of the signal of interest, the fading experienced by the signal, and the distribution of the noise added in the channel. Therefore, this dissertation develops new techniques for estimating the parameters that characterize the additive non-Gaussian noise process, as well as the fading process, in the presence of unknown signals. These novel estimators are an integral contribution of this dissertation. The signal detection and modulation classification problems considered here are treated as hypothesis testing problems. Using a composite hypothesis testing procedure, the unknown fading and noise process parameters are first estimated and then used in a likelihood ratio test to detect the presence or identify the modulation scheme of a signal of interest. The proposed algorithms, which are developed for different non-Gaussian noise models, are shown to outperform conventional algorithms which assume Gaussian noise conditions and also algorithms based on other impulsive noise mitigation techniques. This dissertation has three major contributions. First, in environments where the noise can be modeled using a Gaussian mixture distribution, a new expectation-maximization algorithm based technique is developed for estimating the unknown fading and noise distribution parameters. Using these estimates, a hybrid likelihood ratio test is used for modulation classification. Second, a five-stage scheme for signal detection in symmetric α stable noise environments, based on a class of robust filters called the matched myriad filters, is presented. New algorithms for estimating the noise distribution parameters are also developed. Third, a modulation classifier is proposed for environments in which the noise can be modeled as a time-correlated non-Gaussian random process. The proposed classifier involves the use of a whitening filter followed by likelihood-based classification. A new H_â filter-based technique for estimating the whitening filter coefficients is presented.

This paper describes the implementation of a scanning signal detector and automatic modulation classification system. The classification technique is a completely blind method, with no prior knowledge of the signal's center frequency, bandwidth, or symbol rate. An energy detector forms the initial approximations of the signal parameters. The energy detector used in the wideband sweep is reused to obtain fine estimates of the center frequency and bandwidth of the signal. The subsequent steps reduce the effect of frequency offset and sample timing error, resulting in a constellation of the modulation of interest. The cumulant of the constellation is compared to a set of known ideal cumulant values, forming the classification estimate. The algorithm uses two platforms that together provide high speed parallel processing and flexible run-time operation. High-rate spectral scanning using an energy detector is run in parallel with a variable down sampling path; both are highly pipelined structures, which allows for high data throughput. A pair of processing cores is used to record spectral usage and signal characteristics as well as perform the actual classification. The resulting classification system can accurately identify modulations below 5 dB of signal-to-noise ratio (SNR) for some cases of the phase shift keying family of modulations but requires a much higher SNR to accurately classify higher-order modulations. These estimates tend toward classifying all signals as binary phase shift keying because of limits of the noise power estimation part of the cumulant normalization process. Other effects due to frequency offset and synchronization timing are discussed.

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.

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.