Research On The Modulation Classification And Parameter Estimation For Single-channel Time-frequency Overlapped Signal And System Implementation

Along with the acceleration of social and military informalizaition process and the rapiddevelopment of radio communication technology, the electromagnetic environment is becomingincreasingly dense and complicated, thus time-frequency overlapped-signal exist widely. In thecontext of signal reconnaissance, spectrum monitoring, electronic countermeasures and othernon-cooperative communication fields, the single-channel receiver is the most common mode inthe application of shortwave, ultra-short wave and satellite signal reception. It is significant tostudy how to clarify the related parameters and identify modulation mode rapidly and effectively,with the purpose of providing essential premise for the subsequent signal demodulation andinformation recovery. The subject takes cyclostationary analysis theory and other tools to solvethe third-party reception problems of single-channel time-frequency overlapped-signal, in orderto enhance the ability of acquiring information under the complex electromagnetic environment,as well as to expand the breadth and depth of original research on signal analysis andelectromagnetic spectrum monitoring. In this paper, the main contents and innovation are asfollows:1. Under the non-cooperative condition, this dissertation establishes the mathematicalmodel of single-channel mixed-signal processing and gives the theoretical basis of theintroduction of cyclostationary tools. Based on the analysis of baseband transmission model andcharacteristic of single-channel received signals, the paper analyses the cyclostationary domainfeature of the single-carrier signal as FSK, PSK, QAM and the multi-carrier signals asOFDM, OQAM, etc. The expressions of cyclic cumulants (CC) and spectral correlationfunction (SCF) are derived and verified by the results of simulation experiment, which sustainthe theoretical analysis.2. Several key issues of single-channel time-frequency overlapped-signal parameterestimation are studied.Aiming at the background noise suppression and the multi peaksextraction problems in the parameter estimation, a preprocessing and spectral line extractionalgorithm based on mathematical morphology is proposed, which improves the detectionperformance of multi spectral lines in the condition of short data, low SNR and the lack of priorknowledge. In view of the contradiction between computational complexity and spectralresolution of the existing spectrum correlation estimation algorithm, a type of MTM-TSMalgorithm for cyclic spectrum estimation is put forward, which reduces the estimation varianceand the amount of data under the premise of ensuring the effective resolution. Taking fulladvantages of SCF profile features, an improved cyclic domain method for parameter estimationof single-channel time-frequency overlapped-signal is carried out, realizing the accurateestimation of the number of signal sources, the carrier frequency and symbol rate, with lowcomputational complexity and strong anti-noise performance. The effectiveness and robustnessof the proposed algorithm are illustrated with experimental results.3. The particle filtering,cyclostationary test and support vector machine theories are applied to separate and classify the single-channel time-frequency overlapped-signal,which expand theclassification set, improve the recognition performance and strengthen the signal monitoringability under blind receiving condition. A single-channel separation algorithm based on particlefiltering and support vector machine is proposed, which can reduce the processing time withoutany performance loss compared with the existing algorithms. On this basis, with cyclicfrequency features of cyclic cumulants, a modulation recognition algorithm based oncyclostationaryy test and SVM classifier is proposed. Without preprocessing tasks as estimationof parameters and synchronization, the method effectively accomplishes discrimination ofcommon digital modulated signals such as FSK、SCLD、OQAM、OFDM, etc. The simulationsand comparison confirm that the method can widen the recognition set and promote the signalblind classification performance under non-cooperative reception conditions.4. The structural design of a comprehensive multi-signal analysis system and theconstruction of functional modules are preliminary implemented, and the relevant theories aswell as algorithms are also tested on a general signal processing platform. Engineeringimplementation and experimental testing results validate that the system is pragmatic and theabove algorithms for multi-signal modulation classification and parameter estimation lead to anefficient performance of practice signal analysis scenarios.