Research On Signal Identification Techniques For Spectrum Sharing

As a limited natural resource,the wireless spectrum has tremendous economic and social values.With the advent of the information and communication industry,the level of informatization has been greatly improved and changed,which leads to higher de-mands on the wireless spectrum.The current spectrum allocation scheme benefits inter-ference management,while limits the flexibility of spectrum access.In 2002,the Fed-eral Communications Commission has stated that "In many bands,spectrum access is a more significant problem than physical scarcity of spectrum,in large part due to lega-cy command-and-control regulation that limits the ability of potential spectrum users to obtain such access".As a technique to effectively enhance the spectrum occupancy,spectrum sharing is critical for the sustainable development of the future mobile commu-nication systems.In order to successfully receive and proceed the signals of users in the spectrum sharing system,and to ensure the transmission security,the signal identification techniques should be intensively investigated.Consider the issues that the user parameters and identities are unknown in the spec-trum sharing system,the target of improving spectrum occupancy,and the requirements of transmission efficiency and security,this dissertation studies signal identification tech-niques including spectrum sensing,modulation classification and specific emitter iden-tification by using algorithm design and optimization,modeling and feature extraction,and performance evaluation.The main contributions of the dissertation include:1)For spectrum sensing,considering the non-cooperative relationship among users in the spectrum sharing system,and the fact that the unlicensed user receiver cannot obtain a priori knowledge of the licensed user,the dissertation proposes a spectrum sensing algorithm via hypothesis testing theory,which needs no a priori knowledge of the licensed user,is applicable to various channel conditions,and has advantages of low complexity and high detection performance.Consider the non-time-slotted system,where the licensed and unlicensed users are asynchronized,the full-duplex sensing is adopted to sense spectrums and transmit signals simultaneously,and the optimal throughput and frame length of the unlicensed user are analyzed.2)For modulation classification,the classification problems for the constellation-based modulation and continuous phase modulation are investigated under Gaussian chan-nels.For the constellation-based modulation format,the dissertation proposes a hy-pothesis testing-based modulation classifier,which designs proper features according to different modulation formats and channel conditions,numerical results demon-strate that the algorithm is robust against non-ideal channels,such as the flat-fading channels,etc.,and significantly improves the classification performance.For the continuous phase modulation,by analyzing the memorable property,the dissertation proposes a modulation algorithm based on the hidden Markov model,which can es-timate the hidden variables and unknown parameters.The proposed algorithm solves the classification problem of continuous phase modulation in flat-fading channels.3)For modulation classification,considering the frequency selective fading channels,the dissertation proposes a likelihood-based modulation classifier for single user sce-narios.By properly designing the "complete data" of the expectation-maximization algorithm,the original maximization problem is decomposed into separate ones,so that the maximum likelihood estimates of the unknown channel coefficients are ob-tained.In the single user case,the Cramer-Rao lower bounds of the unknowns and the performance upper bounds are derived.The single user scenario is then extended to the multiuser case,by re-designing the "complete data",the multiuser modulation classification problem in frequency selective fading channels is solved.4)For specific emitter identification,the nonlinearity of the power amplifier is modeled as the specific feature of the transmitter/emitter.Consider a traditional single hop system under various non-ideal channels,it is apparent that the non-ideal channels cause distortions of the specific feature.Furthermore,the single hop case is extended to the multihop one,the amplify-and-forward procedure of signals from the source introduces features of intermediate devices,which contaminate the features of the source.To solve the specific emitter identification problem in such two cases,the dissertation proposes three algorithms based on the adaptive time-frequency analysis theory.Numerical results show that the proposed algorithms significantly enhance the identification performance,and are robust against the aforementioned scenarios.