Study On The Distributed Collaborative Signal Parameter Estimation And Blind Identification Systems

Current wireless communication systems widely apply the digital modulation communication. Due to the rapid improvement of communication technologies and the growth of communication needs, various types of digital modulation methods are increasing. Wireless communication transmission environment is more and more complex. It is more difficult to receive the target signal in complex communication environment and demodulate the useful information from variety of modulation signals, especially in non-cooperation communication. Communication modulation mode blind recognition technology, which is a vital area of implementation whether in military of civilian, consisting of accurately estimating the target signal parameters and identify the digital modulation without any priori information.This thesis studied the distributed digital modulation blind recognition technology, based on the single node digital modulation blind recognition system, which contains the process of distributed system in detail, the parameter estimation algorithm based on different fusion rules, and the blind recognition algorithm for distributed system in decision level, feature level and element level. The system is proposed to address the issue of blind recognition for digital modulated signal in complex communication environment. We proposed simulations to analyze the performance of system The major work and contributions of this thesis are summarized as follows.Firstly, this thesis introduces the system structure design of blind recognition for single-node digital modulation signal including the estimate algorithm of bandwidth, frequency offset, symbol rate and SNR in preprocessing module and modified constellation identification reconstruction algorithm in recognition module. The system recognition performance has been simulated under different frequency offset, different roll-off factor of shaping filter and different number of symbols, which verified the effectiveness of parameter estimation algorithm and blind I recognition algorithm. Then the range of application is been analyzed.Secondly, based on the blind recognition for single-node digital modulation signal, this thesis designed the blind recognition system scheme of multi-node distributed digital modulation signal on the feature level and decision level. Multi-node parameter values integration scheme is utilized to obtain more accurate parameter estimates, then this thesis designed a distributed system parameter estimation process based on fusion criterions and compared the estimation performances of average fusion and weight fusion. Based on different fusion targets, feature-level fusion based on constellation matching likelihood and decision level fusion based on the recognition result have been designed. According to these two schemes, different fusion rules have been designed including simple and efficient statistical decision criteria and weight fusion with the introduction of SNR estimates. Distributed Blind recognition simulation has been implemented under different fusion criteria and different nodes. Results show that the performance of feature-level fusion is 2 ~ 3d B better than the decision level fusion.Finally, this thesis expatiates distributed blind recognition system design of element level digital modulation signal including the one based on sampling-level data fusion and the one based on symbol-level data fusion. For symbol-level distributed system, the processing of node and fusion terminal, interactive data analysis and three different data fusion algorithms have been implement. This thesis simulated the performance of two element-level distributed systems with different fusion algorithm, results show that sampling-level system has a better performance than the symbol-level system. For 16 QAM system, under the same criterions, sampling-level system is 1~5dB better than the symbol-level system. Symbol-level data fusion distributed blind recognition system has a 2~4dB improvement than feature-level blind recognition system.