In this dissertation, we develop a novel cognitive radio (CR) architecture, referred to as the Radiobot , whose goals go beyond dynamic spectrum access (DSA) to achieve the main features of cognition, notably, self-learning and self-reconfiguration. The proposed CR architecture is based on a sequence of signal processing and machine learning techniques that enable the Radiobot to sense a wide frequency band and act autonomously by learning from past experience. To achieve its goals, the proposed CR is equipped with the following functionalities: 1) Wideband spectrum sensing, 2) non-parametric signal classification, 3) unsupervised learning and reasoning and 4) decentralized decision-making. To this end, we implement a blind spectrum sensing method based on joint energy/cyclostationary detection. Optimal wideband energy detector is designed based on the Neyman-Pearson (NP) criterion which maximizes the detection probability of primary signals, subject to a certain false alarm rate. Cyclostationary detection is proposed as a means of extracting the underlying cyclic properties of the detected signals in order to identify the types of signals in each frequency band. Once the signal features are extracted, a Bayesian non-parametric classifier based on the Dirichlet process is applied to determine the different types of wireless systems in the surrounding radio frequency (RF) environment. In this dissertation, we extend the Dirichlet process mixture model (DPMM)-based classifier to allow for a mixture of Gaussian and non-Gaussian vector observation models, compared to existing DPMM's with scalar Gaussian observation models. We also develop a sequential DPMM classifier that can be implemented at a low processing cost, making it suitable for real-time operation. Upon identifying the RF activities in the surrounding environment, the Radiobot uses machine learning techniques for decision-making. Thus, we propose a reinforcement learning (RL) algorithm that enables the Radiobot to learn by interacting with its environment. The learning process is formulated in a decentralized partially observable Markov decision process (DEC-POMDP) framework and is shown to lead to a near-optimal policy with little knowledge about the environment. As a result, using its sensing and learning capabilities, the Radiobot can switch among multiple modes of operation to adapt to a dynamic RF environment.
Cognitive radio, Spectrum sensing, Signal classification, Machine learning, Dirichlet process, Reinforcement learning, Cyclostationary detection
Level of Degree
Electrical and Computer Engineering
First Committee Member (Chair)
Second Committee Member
Third Committee Member
Bkassiny, Mario. "Wideband Spectrum Sensing and Signal Classification for Autonomous Self-Learning Cognitive Radios." (2014). http://digitalrepository.unm.edu/ece_etds/35