Research On Distributed Modulation Classification Technologies Of Communication Signals In Wireless Sensor Network

Due to the rapid extension of the wireless communication, the electromagnetic environmentbecomes increasingly complex. As one key technology of the electromagnetic spectrummonitoring, the modulation classification technology faces greater difficulties and challenges.The existing researches have been limited to single receiver situation where the classificationperformance has largely depended on channel quality and the signal strength. With the recentincrease in popularity of sensor networks, distributed network structure of spectrum monitoringhas become the main direction of development. Aiming at the demand of electromagneticspectrum monitoring, combining with distributed structure of electromagnetic sensor network,modulation classification technologies are researched and implemented. The main work of thisthesis is summarized as follows:1. Aiming at the limited ability of electromagnetic spectrum monitoring node on computingand storage, the spectral correlation based modulation classification algorithms is studied. Afteranalyzing the spectral correlation of different signals, a group of spectral-correlation features areextracted. According to these features, a binary tree classifier is designed. Simulation resultsshow that the average classification ratio can reach to95%under the condition of5dB SNR.2. Combining with the distributed network structure of electromagnetic sensor network, thedistributed modulation classification algorithms are studied. On the basis of distributedmodulation classification framework, the algorithms are divided into signals fusion, featuresfusion and decisions fusion. According to the model and analysis of distributed modulationclassification, the criterion of distributed modulation classification based on features fusion isproposed. And then the performance comparation of different algorithms for distributedclassification is presented. Simulation results show that distributed modulation classification canimprove the performance compared to single receiver.3. For engineering application, the modulation classification algorithm based on SNRweighed voting is studied. In reducing the network load, the algorithm holds a good performancecomparing to the algorithm of features fusion. And then aiming at the performance of modulationclassification not increasing with the number of the nodes, the preparing conditions of combiningmodulation classification based on weighted voting algorithm are analyzed. Simulation resultsdemonstrate the conditions having certain reference significance.4. Distributed modulation classification implementation and testing. The modulationclassification algorithms based on spectral-correlation features are realized on DSP of the signalprocessing unit. According to the wireless sensor network platform, distributed classificationscheme is proposed, the system testing is presented using the combining-classification-controlsoftware. The testing results confirms that the features can realize the identification of signalsamong PSK, QAM, ASK, FSK and QAM. And comparing with single receiver, combiningclassification can improve the performance.