Novel Cooperative Automatic Modulation Classification by Credit-based Consensus Fusion

A novel cooperative automatic modulation classification (AMC) scheme by use of the credit-based consensus fusion is introduced in this paper. We propose a two-stage cooperative AMC approach to be deployed by a wireless sensor network consisting of a fusion center (FC). The unknown modulation scheme of a target signal is first identified by each individual sensor using graph-based automatic modulation classification. Then all local (individual) decisions are combined by the fusion center to produce the ultimate result according to the associated individual credits which are pertinent to the respective historical classification accuracies. Compared to the existing AMC schemes, which can be implemented only on a single sensing device (receiver), our proposed new cooperative AMC method can increase the overall classification accuracy as the cooperation and credit-based consensus fusion employed in our new method can greatly mitigate the adverse effect of channel distortion and noise impairment often encountered in the existing single-sensing-device modulation classifiers. Meanwhile, the extra communication overhead required by our new method is quite limited and therefore the incurred communication cost is minimum. Monte Carlo simulation results have justified the effectiveness and superiority of our proposed new cooperative AMC method.