基于导频的OFDM信道估计小波核SVM算法

Based on pilot wavelet kernel recursive least square support vector machine OFDM channel estimation algorithm

针对传统多径衰落下的OFDM导频信道估计性能低下,地板效应的缺陷,提出了基于导频的小波递归最小二乘支持向量机(WRLS-SVM)时变信道频率估计算法。首先讨论了小波成为核函数的条件,构造了小波核。然后将根据结构风险递归二乘最小化准则回归估计支持向量机原理,把导频训练序列映射到高维空间,并在高维空间采用结构小波核函数,达到了将低维空间的非线性估计转化为高维空间的线性估计的目的。仿真结果表明,在快衰落信道条件下,小波递归最小二乘支持向量机导频信道估计方法可以获得较好误码性能。该方法优于传统的信道插值方法。

Aiming at the lower performance and floor effect of the traditional orthogonal frequency division multiplexing channel estimation, the OFDM channel estimation algorithm based on wavelet recursive least square support vector machine （WRLS SVM） is proposed. Firstly the condition that a wavelet function becomes kernel functions is discussed and the wavelet kernel function is constructed. The mapping the pilot training from the lower dimension space to high dimension space, and in the high dimension space, the wavelet kernel function is adopted, according to recursion least square criteria, the lower linear inseparable problem is convert to the separable problem. The simulation shows that under the fast fading channel circumstance, the WRLS-SVM channel estimation can acquire the high error performance. The method is better than the traditional method.