Quantized Compressive Spectrum Sensing (QuaCoSS)

点击翻译按钮获取中文摘要

Spectrum sensing aims at detecting non-occupied frequency bands in the radio spectrum in order to enable unlicensed (secondary) users to opportunistically communicate over these bands. The goal is to increase communication rates for the secondary users without creating interference for licensed (primary) users. However, this demands to monitor a very large bandwidth so that sampling at the Nyquist rate may result in unacceptably large amount of samples. It is well-known that some of the test statistics used in spectrum sensing feature sparsity, i.e., the frequency spectrum has been shown to be sparsely occupied and the cyclic spectra of man-made signals are sparse. By exploiting this sparsity, one may use compressed sensing techniques instead of traditional Shannon-Nyquist sampling in order to significantly reduce the number of samples, while still ensuring reliable detection of non-occupied bands. In practice, samples have to be quantized before exchanging these.This project aims to explore the effect of quantization in compressed sensing on spectrum sensing. We will explore both the extreme case of one-bit quantization, where only the sign of a measurement is retained, as well as multi-bit quantization schemes. A particular focus is put on structured random measurements such as the random partial Fourier matrix, which are highly relevant in practical applications. While initial theoretical results on quantized compressed sensing are available for Gaussian random measurement matrices, structured random matrices remain completely unexplored in this context up to now. Moreover, we plan to investigate two open fundamental problems in the practical application of quantized compressive spectrum sensing. Firstly, given a fixed bit-budget for communication we will investigate the tradeoff concerning the quantization resolution and the number of measurements taken by secondary users in order to achieve optimal detection performance. Secondly, we will analyze the tradeoff between the occupancy decision frequency and the bit-budget available per decision in order to minimize wasted transmission opportunities.The research group of Mathar will dedicate its efforts to the development, implementation and simulation of quantized compressed sensing algorithms and their application in the spectrum sensing context. The focus of the research group of Rauhut and Dirksen will be on the theoretical analysis of quantized compressed sensing with the aim of deriving rigorous error guarantees and sharp bounds on the required number of measurements. The interaction of the two groups is expected to be crucial for achieving significant progress on the understanding and practical implementation of quantized compressive spectrum sensing.