Low Complexity Automatic Modulation Classification Based on Order-Statistics Order
Abstract: In this paper, we propose three automatic modulation classification classifiers based on order-statistics statistics and reduced order order-statistics, statistics, where the order-statistics order are the random variables sorted by ascending order and the reduced orderorder statistics represent a subset of the original order order-statistics. statistics. Specifically, the linear support vector machine classifier applies the linear combination of the orderorder statistics of the received signals, while the approximate maximum likelihood and the backpropagation neural n networks etworks (BPNNs) classifier resort to the reduced order-statistics statistics to decrease the computational complexity. Moreover, BPNN is applicable for modulation classification both in known and unknown channel scenarios. It is shown that in the known channel scenar scenario, io, the proposed classifiers provide a good tradeoff between performance and computational complexity, while in the unknown channel scenario, the proposed BPNN classifier outperforms the expectation maximization classifier in terms of both classification performance erformance and computational complexity. Simulations results are provided to evaluate the proposed classifiers.