在MIMO频率选择性衰落信道下,使用级联训练序列是解决频偏估计中相位模糊问题的方法之一.级联训练序列由若干个短序列和若干个长序列组成.短序列是为了获得更大的估计范围而长序列是为了获得更高的估计精度.本文利用克拉美罗下界(Cramer-Rao Lower Bound,CRLB)作为衡量标准分析了级联训练序列的频偏估计性能,推导了级联训练序列的均方误差公式,得到了级联训练序列的性能渐进等于长序列的估计性能,并且证明了事实上最优的级联训练序列结构并不是目前广泛采用的级联结构,而是将整个序列全部分割成短序列.最后通过仿真验证了理论分析的结果。
Using concatenated training sequences for frequency-offset estimation is one of the solutions to the problem of phase ambiguity over MIMO frequency-selective fading channels. A concatenated sequence consists of short sequences for coarse estimation and long sequences for fine estimation. In this paper,we use the Cramer-Rao Lower Bound(CRLB)as a metric to analyze the performance of concatenated training sequences. It is shown that the performance of a concatenated training sequence is asymptotically equal to that of the long sequences. Furthermore, we show that the optimal concatenated training structure that minimizes the threshold and the CRLB is just dividing the complete sequence into the shortest blocks without long blocks. Finally, numerical results validate our analysis.