混沌背景中微弱谐波信号检测的SVM方法

为了提高混沌背景下的微弱谐波信号检测能力，提出了一种提取混沌背景中微弱谐波信号的支持向量机（support vector machines，SVM）方法。该方法的突出特点是针对小样本或嵌入维数未知的情况，建立混沌噪声的一步预测模型，抑制噪声对混沌背景信号预测的影响，起到预滤波作用，然后从预测误差中提取微弱谐波信号。实验结果表明，该方法具有比传统RBF神经网络预测方法更强的稳健性和泛化性，在信噪比（SNR）为-47．931dB时仍可检测出强混沌中的微弱谐波信号。

Detection of weak harmonic signal embedded in chaotic noise using SVM

In order to detect weak harmonic signal embedded in noisy chaotic background, the method of support vector machine (SVM) is proposed, which is used to extract weak harmonic signal under chaotic background. The outstanding feature of the method is building one-step prediction model of the chaotic noise (including white or colored noise) for small sample or unknown embedded dimension situation. The model can suppress the influence of noise on the prediction of chaotic background signal, and provide pre-filtering function, then it extracts weak harmonic signal from predicted error. Experimental result shows that comparing with conventional RBF neural network prediction method, SVM method has stronger robustness and generalization ability. The method can detect weak harmonic signal when signal-noise-radio (SNR) is as low as -47.931 dB.