基于小波神经网络的调制信号识别方法

为了克服神经网络识别类别较多时构建网络复杂、训练速度低的缺点,提出了一种小波变换和阵列式RBF网络结合的方法实现无线通信信号调制类别检测.利用小波变换对常用3种模拟信号和6种数字信号进行多层分解和特征提取,然后利用特征参数通过阵列式RBF网络进行信号调制类别检测.仿真结果表明,小波分析和阵列式神经网络相结合的设计,使无线通信信号调制类型的检测系统在信噪比为-10 dB达到平均辨识率90％以上的性能,同时提高了多类别情况下的检测率.

A recognition method on modulation signal based on wavelet network

In order to overcome single neural network some weakness that it is difficult to expand,modify and maintain the neural network with large categories,a recognition method based on wavelet and RBF network array is used to divide wireless communication modulation signals.The feature extractions of three kinds of analog signals and six digital signals are picked up by a wavelet decomposition method and then feature extractions are classified by RBF neural network array.The simulation results show that the design combined with the wavelet analysis and neural network array makes the modulation category detection system of the wireless communication signal to achieve about 90% performance when SNR is-10 dB,and the detection rate in the multiple categories is improved.