Spectrum sensing for cognitive radio and radar systems

Abstract

The use of the radio frequency spectrum is increasing at a rapid rate. Reliable and efficient operation in a crowded radio spectrum requires innovative solutions and techniques. Future wireless communication and radar systems should be aware of their surrounding radio environment in order to have the ability to adapt their operation to the effective situation. Spectrum sensing techniques such as detection, waveform recognition, and specific emitter identification are key sources of information for characterizing the surrounding radio environment and extracting valuable information, and consequently adjusting transceiver parameters for facilitating flexible, efficient, and reliable operation. In this thesis, spectrum sensing algorithms for cognitive radios and radar intercept receivers are proposed. Single-user and collaborative cyclostationarity-based detection algorithms are proposed: Multicycle detectors and robust nonparametric spatial sign cyclic correlation based fixed sample size and sequential detectors are proposed. Asymptotic distributions of the test statistics under the null hypothesis are established. A censoring scheme in which only informative test statistics are transmitted to the fusion center is proposed for collaborative detection. The proposed detectors and methods have the following benefits: employing cyclostationarity enables distinction among different systems, collaboration mitigates the effects of shadowing and multipath fading, using multiple strong cyclic frequencies improves the performance, robust detection provides reliable performance in heavy-tailed non-Gaussian noise, sequential detection reduces the average detection time, and censoring improves energy efficiency. In addition, a radar waveform recognition system for classifying common pulse compression waveforms is developed. The proposed supervised classification system classifies an intercepted radar pulse to one of eight different classes based on the pulse compression waveform: linear frequency modulation, Costas frequency codes, binary codes, as well as Frank, P1, P2, P3, and P4 polyphase codes. A robust M-estimation based method for radar emitter identification is proposed as well. A common modulation profile from a group of intercepted pulses is estimated and used for identifying the radar emitter. The M-estimation based approach provides robustness against preprocessing errors and deviations from the assumed noise model.