Research On Multi-node Cooperative Modulation Classification Based On Decision Fusion

Cooperative automatic modulation classification(CAMC)plays a very important role in military and civilian fields,such as spectrum monitoring,cognitive radio,and rescue.Among various CAMC methods,the decision-level CAMC method requires minimal network overhead and is suitable for complex large-scale networks.However,the existing decision-level CAMC methods still face challenges.First,in the existing decision-level CAMC methods,in addition to sending it local decision to the fusion center,some other information also need to be sent to fusion center to produce the global decision,which makes it is impossible to keep the minimal network overhead.Second,the existing decision-level methods are all oriented to static wireless sensor networks,which leads to relying on a fixed fusion center and is not suitable for dynamic wireless sensor networks.This thesis mainly aims at the above two challenges,firstly studies the existing classic decision-level CAMC methods,and a novel credit-based CAMC method is proposed.Then,a temporal discounted weight based CAMC method is further proposed.The main contributions are summarized as follows.1.The existing classic CAMC methods are studied,then we explore the architecture principles of general decision-level CAMC methods.2.A credit-based CAMC method is proposed,in this method,the local sensors only need to send its local decision to the fusion center without sending other additional information,which achieves the minimal communication overhead.3.On the basis of the proposed credit-based CAMC method,the concept of temporary fusion center is further proposed in this thesis,and a temporal discounted weight based CAMC method is proposed by introducing cumulative state and temporal discount factor.In the temporal discounted weight based CAMC method,there is no longer a fixed fusion center,but a temporary fusion center is dynamically selected in each sensing interval.Meanwhile,the temporal discount factor enables it to flexibly cope with the dynamic changes of the network,which means it is still effective in the dynamic wireless sensor network.4.Simulation results demonstrate the effectiveness and superiority of the two proposed CAMC methods.The effect of number of local sensors,interference signals,and dynamic changes on the two proposed CAMC methods is evaluated via simulation under different conditions.The computational complexity of the two proposed CAMC methods is investigated in detail.The simulation results verify that our proposed new decision-level CAMC methods are superior to the single-node AMC method,and the accuracies produced by our proposed new CAMC methods are about 70% higher than that of the existing decisionlevel CAMC method.