优化的神经网络分类器在自动调制识别中的应用

摘要

对截获到的信号调制类型进行自动识别是非协作通信中的关键技术之一，在民用及军用领域中有着重要的应用前景。其中，分类器的设计对调制类型识别结果和效率起到了决定性的作用。各种方法中，采用 BP神经网络构造的分类器能获得较好的识别效果，但是传统 BP神经网络存在收敛速度慢、容易陷入局部最小值、网络对初始值敏感等问题。论文采用遗传算法优化 BP神经网络的权值和阀值，可以避免神经网络利用梯度下降法陷入局部最小值的缺陷，从而提高BP网络的学习能力。论文对高斯白噪声信道中6种常用的数字通信信号进行了判定识别。仿真结果表明，用遗传算法优化的神经网络分类器能有效地提高调制信号识别率。%Detected signal’s automatic modulation classification (AMC) is one of the key technologies in non-cooperative communication. It has extensive application prospects in civilian and military fields. The design of classifier plays a decisive role in recognition results and the efficiency of modulated schemes. The classifier based on back-propagation (BP) neural network is better than the existing methods of AMC. However, there are some entrapment at a local optimum, slow convergence rate and sensitive to initial values for the traditional BP neural network. This paper introduces genetic algorithm to optimize the weight and threshold values of the BP neural network. The method can overcome the weakness of gradient-based BP algorithm which is easy to fall into local minimum. And moreover this method can enhance the learning ability of BP neural network. The paper iden-tifies six digital signals in an additive white Gaussian noise channel. The simulation and experimental results show that the optimized neural network classifier designed with genetic algorithm can perform a higher recognition rate of modulation type.