基于神经网络实现多种数字信号调制方式的自动识别

A．K．Nandi与E．E．Azzouz提出基于决策论的信号调制样式自动识别方法具有简单易行，适合在线分析的优点．本文在该方法的基础上，针对该方法的数字信号调制样式识别进行深入仿真实现，并提出了基于该方法利用正交最小二乘法(OLS)的RBF神经网络实现数字信号调制样式自动识别的方法，最后对计算进行了改进，提高了该方法的识别能力，对SNR为8～20dB的信号识别得到了较好的结果．本文识别的数字信号为2ASK、4ASK、2PSK、4PSK(QPSK)、2FSK、4FSK与16QAM七种信号．

Automatic recognition of the multi-digital signals modulation based on the neural network

The algorithm for automatic recognition of digital signal modulation on the basis of policy_making theory developed by A.K.Nandi and E.E. Azzouz has the advantages of being simple to operate and fit for analysis online. Based on the algorithm above,this paper addressed an algorithm based on OLS RBF neural network. The simulation was made for the modulation recognition and an improved algorithm was proposed to enhance the performance of the method.The better results for signals with Gaussian noise were obtained,whose SNR are low to 8dB.Seven kinds of digital modulation signals were discussed,including 2ASK,4ASK,2PSK,4PSK(QPSK),2FSK,4FSK and 16QAM.