论文部分内容阅读
这篇论文致力于基于神经网络的通信信号调制分类器设计研究.论文提出了一种结构优化的神经网络分类器,它采用6个特征参数,可以对2FSK、4FSK、8FSK、2PSK、4PSK、8PSK、OQPSK、16QAM、CPFSK、MSK共10种调制类型实现正确分类识别。结构优化的目标在于寻找最小神经网络结构,从而使该结构确实适合由训练数据所定义的实际函数.优化的结构使神经网络分类器获得了十分出色的推广性能和分类性能.
This dissertation is devoted to the design and research of communication signal modulation classifier based on neural network.A structural optimization neural network classifier is proposed in this paper, which uses 6characteristic parameters and can detect 2FSK, 4FSK, 8FSK, 2PSK, 4PSK, 8PSK , OQPSK, 16QAM, CPFSK, MSK a total of 10 kinds of modulation types to achieve the correct classification and identification. The goal of structural optimization is to find the minimum neural network structure so that the structure really fits the actual function defined by the training data.The optimized structure makes the neural network classifier get very good promotion performance and classification performance.