混沌背景下基于神经网络的微弱谐波信号检测

Weak Signal Detection Based on Neural Network under the Background of Chaos

欲得到淹没于强噪声背景中的弱信号,提出一种基于径向基函数网络的检测方案.利用网络最优化结构对重构后的采样序列训练集相空间进行非线性拟合,进而预测验证集.对预测误差作功率谱分析可获得背景噪声有效抑制后的弱谐波信息.仿真检测分别在Lorentz、Duffing和Logistic混沌背景及白噪声条件下进行,最低检测信噪比分别可达-36 dB、-36 dB、-15 dB.

To detect weak signal buried in strong chaos background, a method based on RBF network is brought up. The optimal structure of network is trained by training set, obtained after sampling time series＇ reconstruction of Phase Space. Then the network is to forecast test set. On analyzing the power spectrum of prediction error, in which chaos background is sharply bater the required weak signal can be extracted. Simulations are taken respectively under Loxentz, Duffing and Logistic background adding white noise. The lowest detectable signal-to-noise ratios are -36 dB. -35 dR. and -15 AB.