摘要
针对最大似然时延估计算法的峰值搜索计算复杂度较高且容易陷入局部收敛，造成估计误差较大的问题。提出了一种利用蒙特卡罗的最大似然时延估计（MCML）算法。首先利用信道频域响应估计矢量建立似然函数；然后把时延估计问题转化为求解随机变量的期望问题，将采用指数化似然函数构造的标准化概率密度函数趋近于冲激函数，使得随机变量的方差趋近于零；最后采用蒙特卡罗方法对随机变量进行抽样，从而利用抽样的均值估计出时延。较之传统方法，蒙特卡罗方法避免了网格搜索，降低了计算复杂度，保证了全局收敛性和估计精度。仿真结果表明：在信噪比0～25dB的条件下，MCML算法均能始终逼近克拉美罗界；当信噪比为25dB时，MCML算法的时延估计分布范围缩小为马尔科夫链蒙特卡罗算法的34％。
A maximum likelihood time delay estimation algorithm using Monte Carlo method （MCML） is proposed to solve the problems that the maximum likelihood delay algorithm has high computational complexity due to peak searching and is easy to fall into local convergence. A likelihood function is constructed by using the channel response estimation vector in frequency domain. Then the MCML translates the time delay estimation into the expectation of a random variable, and a standardization probability density function is built from the index likelihood function to approximate an impulse function, and to make the variance of the random variable approach zero. Finally, the random variable is sampled using the Monte Carlo method, and the time delay is estimated from sampling mean. Compared with the traditional methods, the MCML avoids the grid search, reduces the computational complexity, and ensures the global convergence and estimation accuracy. Simulation results show that the MCML is always close to Cramer-Rao bound, and the time delay estimation range of the MCML is 34% of the MCMC＇s range when the signal to noise ratio is from 0 dB to 25 dB.
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
2015年第8期24-30,共7页
Journal of Xi'an Jiaotong University
基金
国家高技术研究发展计划资助项目(2012AA01A502
2012AA01A505)
关键词
最大似然
蒙特卡罗
时延估计
克拉美罗界
maximum likelihood
Monte Carlo
time delay estimation
Cramer-Rao bound
