摘要
首先介绍从信号幅度、相位、频率及功率谱等特性中提取的三种特征参数 ,应用这三种参数采用人工神经网络对模拟调制信号进行了识别。在基于网络输出最小均方误差和最佳正确识别率的原则下 ,提出识别CW、AM、SSB、FM信号的隐含层的神经元个数为 10个时较佳 ;得出输出层采用log -sigmoid函数为激励函数时 ,可以达到较佳的识别效果的结论。
Three features parameters derived from the instantaneous amplitude,instantaneous phase,instantaneous frequency and power spectrum of analogue modulated signals are presented in this article.Artificial neural networks (ANNs) are used to recognize analogue modulation including CW?AM?SSB?FM with the features. Based on the maximum probability of correct decisions and an minimum sum squared error (SSE), the more optimum node number in the hidden layer is presented in this paper, and it is found that when the log-sigmoid function is used as the activation function in the output layer, recognition performance is better.
出处
《杭州电子工业学院学报》
2002年第3期34-36,共3页
Journal of Hangzhou Institute of Electronic Engineering
