基于支持向量域的快衰落信道混沌预测  

Chaos prediction of fast fading channel using support vector domain

作　　者：任韧 [1] 朱世华 [2] 任大男 [3] 

机构地区：[1]西安交通大学,电子与信息工程学院,陕西,西安,710049,西安交通大学,理学院,陕西,西安,710049 [2]西安交通大学,电子与信息工程学院,陕西,西安,710049 [3]西北大学,数学系,陕西,西安,710069

出　　处：《通信学报》2005年第12期117-125,共9页Journal on Communications

基　　金：中国科学院资助项目，国家科技攻关项目，西安交通大学校科研和教改项目

摘　　要：以多径无线衰落信道存在奇怪吸引子为前提,从支持向量域SVD出发,研究了无线衰落信道的响应特性.根据相空间延迟重构理论建立了快衰落信道的预测模型,使用训练集合作为支持对象元素,通过机器自学习缩小模型泛化误差上界,采用最小二乘支持向量域,实现了非线性高维相空间映射,预测了衰落信道.实验结果表明:支持向量域含支持向量数少,运算速度快,顽健性强,核函数选择灵活,为自适应实时估计和检测提供了一种方法.该方法在小样本和未知信道概率密度下,信道预测值和真实值取得了一致.Because the wireless fading channel existed strange attractors, based on the origin of support vector domain （SVD） concept, SVD predictive model was proposed of multi-path fading channel. According to the chaotic reconstructing theory of takens phase space delay, the chaotic fading channel model was established. The training set was used to be support object elements. Machine self-learning made error least upper bound of generalization model to be minimum. The non-linear higher dimension map was realized by the least squares support vector domain. The future fading channel data was predicted from training data set. The experiment result indicateds that the support Vector domain needs little support vector with fast convergence rate. The system has robustness characteristic and kernel function of flexible choice. It is a method to fit adaptive real-time estimation and detection. With the small sample and unknown probability density, the multi-path predictive series consists with true value series in Doppler fast fading channel.

关 键 词：支持向量域 衰落信道 最小二乘 混沌序列预测 

分 类 号：TN929.5[电子电信—通信与信息系统]