Optimized Distributed Automatic Modulation Classification in Wireless Sensor Networks Using Information Theoretic Measures

Automatic modulation classification of digital signals is essential for intelligent communication systems. This paper addresses the distributed classification of digital amplitude-phase modulated signals in a system of multiple sensors, which observe the unknown signal corrupted with the additive white Gaussian noise. The sensors are connected to a fusion center through block-fading orthogonal multiple access channels. We introduce a new method, where: 1) for classification, an information-theoretic similarity measure known as correntropy is exploited by each local sensor; 2) for transmitting local decisions, an estimation of a priori probabilities is used by local sensors; and 3) for optimizing power allocation to each sensor, the Bhattacharyya distance is employed as the objective function by the fusion center. The proposed scheme improves the classification accuracy through exploiting channel diversity, which, in turn, enhances the overall performance. Simulation results validate the theoretically claimed improvement in the performance.