基于小波神经网络的数字信号调制识别

Digital modulation recognition based on wavelet nerual network

本文采用熵特征的提取方法,大大减小了特征值的计算量,增强了小波神经网络(WNN)识别的有效性。同时采用改进的算法训练小波神经网络,有效的避免算法陷入局部最小值,克服了传统BP网络的固有缺点,并提高了小波神经网络的训练速度。结果表明,该系统能快速有效的识别出数字信号的调制类型,具有较高的识别精度。

In this paper,tincture extraction method based on entropy is used to decrease computational times and strengthen the effectiveness of classification.At the same time,improved arithmetic is used to train WNN ,which effectively avoide getting into partial minimum values.It conquers the inherent flaw of conditional BP net,and improves training speed as well as ..The test shows that the system can recognize types of digital modulation signals.Application of this method to modulation recognition of practial signals shows satisfactory performance.