Hierarchical classification of modulation signals

Abstract

This thesis addresses the problem of classifying both analog and digital modulation signals using different kinds of classifiers. The classification of modulation signals has both civilian and military applications. A total of 31 statistical signal features are extracted and used to classify 11 modulation signals plus white noise in a hierarchical fashion. The modulation signals include carrier wave (CW), AM, FM, SSB, FSK2, FSK4, PSK2, PSK4, OOK, QAM16, and QAM32. A hierarchy of classifiers is introduced, and the genetic algorithm is employed to obtain the most effective set of features at each level of the hierarchy. Based on the selected features at each level, Bayesian and neural network classifiers are designed to separate the most distinct subclasses at that level. The thesis also discusses the real-time implementation of the developed classification system on a high performance DSP processor, namely TMS320C6701. Various steps taken to optimize the classification algorithm on the DSP processor, and the real-time performance issues are presented. The classification results and number of operations on the DSP indicate the effectiveness of the introduced hierarchical classification strategy in terms of both performance and processing time.