OFDM信道估计算法研究

【作者】 彭娟；

【导师】 束锋；

【作者基本信息】 南京理工大学 ， 通信与信息系统， 2010， 硕士

【摘要】 正交频分复用(OFDM, Orthogonal frequency division multiplexing)具有高的频谱利用率和良好的抗多径干扰能力,是未来无线通信的核心技术之一。信道估计是OFDM的研究热点之一。依据归一化帧频偏(NFDS, Normalized FrameDoppler),其定义为帧长度与信道相关时间之比,本文将信道划分为三类：慢时变(NFDS≤0.1),中等时变(0.1<NFDS≤1),快时变(NFDS>1)。主要工作如下：(一)在慢时变典型城市信道,重点研究了线性内插、Sinc内插、高斯内插和线性最小均方误差(LMMSE, Linear minimum mean square error)算法。针对LMMSE存在统计适配的问题,提出基于门限的自适应LMMSE算法,通过仿真发现：当插入导频较密时,Sinc内插和高斯内插的性能逼近LMMSE,考虑它们的复杂度非常低,因此非常具有吸引力；当统计相关函数同实际信道存在适配时,自适应LMMSE算法的均方误差相对于LMMSE有4dB的信噪比提升。(二)在三种典型时变城市信道,重点研究了二维LMMSE、单维LMMSE、Sinc函数内插和基于多项式函数基展开算法。为了降低复杂度,我们提出加权二次样条函数信道估计算法。仿真表明：在所有三种信道条件下,因为充分挖掘了信道的统计特性,二维LMMSE始终是性能最优的信道估计器,同时也是复杂度最高的估计器；基于多项式函数基展开和二次样条函数基展开具有复杂度低和性能优的特点,因此也具有较大吸引力；特别指出的是,当0.3≤NFDS≤0.8时,二次样条函数基展开算法实现了复杂度和性能的较好折衷。更多还原

【Abstract】 Orthogonal frequency division multiplexing (OFDM) has the following advantages: high spectrum efficiency and a good ability to combat muti-path interference. Thus, it is considered as one of the key techniques of future wireless communications.Channel estimation is one of the hot research topics of OFDM. According to normalized frame Doppler spread (NFDS=frame time length/coherent time), wireless channels are divided into three categories: slow time-varying channel(NFDS≤0.1), medium time-varying channel(0.1<NFDS≤1) and fast time-varying channel(NFDS>1). The main works are as follows:a) In slow time-varying typical Urban (TU) channel, we focus mainly on investigating linear interpolation (LI), Sinc interpolation (SI), Gaussian interpolation (GI) and linear minimum mean square error (LMMSE). The problem of the LMMSE algorithm is statistics mismatch, in order to solve it, an adaptive LMMSE algorithm based on threshold is proposed. Simulation shows: in the case of a dense pilot grid, the performance of SI and GI is close to that of LMMSE, considering their low complexity, they are very attractive; this adaptive algorithm achieves a 4dB signal-to-noise ratio (SNR) gain on mean square error (MSE) performance over the original LMMSE when there is mismatch between statistical correlation and exact correlation.b) In three time-varying TU channels, we focus mainly on the study of two-dimensional LMMSE (2D LMMSE), one-dimensional time-frequency LMMSE (1D LMMSE), sinc interpolation (SI), and polynomial basis expansion based model algorithm. To reduce complexity of channel estimation, a weighted second-order spline (WSOS) basis algorithm is proposed. Simulation results show: due to its exploiting the statistical property of time-varying channels, the 2D LMMSE perform best in three time-varying channels, however, its computational amount is very high compared with polynomial, WSOS and 1D LMMSE; due to low complexity and good performance, polynomial and WSOS are attractive for time-varying channel estimation; particularly, the proposed WSOS can strike a good balance between complexity and performance for 0.3≤NFDS≤0.8.更多还原