Modulation Classification In Nonideal Environments

Modulation classification (MC) is an intermediate step between signal detection and demodulation, which is a technique to automatically classify the modulation type of a received signal. It is a branch of non-cooperative communication techniques that exploits several classical communications disciplines including signal detection, parameter estimation and channel identification. It is a challenging problem, especially in a non-cooperative environment, as no prior information on the incoming signals is available. The received signals of MC are often concealed in noise including Gaussian white noise and impulse noise, and are often in multipath fading environments. Usually, MC is modeled as a multiple hypothesis test problem, which makes a choice from a set of candidate modulations.MC is a rapidly evolving signal analysis area and has received more and more scientific attention. It plays an important role in both military and commercial applications. It is widely used for spectrum surveillance and management, interference identification, military threat evaluation, electronic counter-measures, source identification and many others.In this dissertation, we mainly discuss the MC problem in nonideal environments, especially in slow flat fading channels. We also deal with the MC problem of signals with unknown parameters including carrier offset and phase offset caused by non-cooperative communications, signal amplitude, noise power and inter-symbols interference.1. We derive the framework of modulation classification for signals propagated over slowly fading channels based on average likelihood rate test (ALRT) methods. The framework can be applied to both linear and nonlinear modulation. We then apply the framework to FSK and QAM signals, and provide their likelihood functions and classifier model respectively. The performance of the FSK signal classifier is analyzed using the receiver operating characteristic. The FSK classifier is then extended to an asynchronous classifier and an approximate classifier. The QAM signal classifier based on the framework is compared to another amplitude-probability-functions based classifier. They are proved to possess the same performance.2. We propose a higher order cumulant (HOS) based classifier for linear modulated signals propagated over slowly fading channels. The 4th order cumulants of the received signals are chosen as the classification statistics. The theoretical values of the 4th order cumulants for various constellations of interest are derived for different kinds of fading types. And the means and variances of the estimated cumulants of the incoming signals are analyzed. The estimated cumulants of the incoming signals are proved to be asymptotically unbiased and consistent with theoretical analysises and simulations. Finally, the proposed HOS based classifier is validated via extensive simulations.3. For linear modulationed signals propagated over frequency selective fading channels, Markov chain Monte Carlo (MCMC) method is applied to solve the modulation classification problem with multiple unknown parameters including carrier offset and phase offset caused by non-cooperative communications, signal amplitude, noise power and inter-symbols interference. The MCMC method based classifier is a general likelihood rate test classifier, where the unknown parameters are estimated using the samples generated by MCMC steps. The posterior conditional distributions of the unknown parameters and transmitted symbols are derived from the prior distributions of the received signals.4. To study the modulation classification problem for linearly modulationed signals traveled over slowly fading channels, a new algorithm based on equal gain combining is proposed. The model of modulation classification based on equal gain diversity is given in which equal gain combining is used to combine the signals received by multiple antennas. The normalized fourth-order cumulants of the signal at the output of the combiner are estimated, which can be used as modulation classification features. The theoretical values of the normalized fourth-order cumulants of the signal at the output of the combiner for various constellations are derived as the classification criterions. Computational complexity of the developed algorithm is order N, where N is the number of the complex baseband data samples. Theoretical arguments are verified via extensive simulations and comparisons with existing approaches.