有效降低计算量的粒子滤波多用户检测新方法

Multi-user detection method for effectively reducing computational complexity based on particle filter

粒子滤波算法中,建议分布的选择直接决定着该算法的估计精度和计算量。针对传统粒子滤波多用户检测方法计算量大的问题,提出了采用最大似然估计算法来确定粒子的建议分布。通过最大似然估计对均匀分布的粒子进行优选,使优选后的粒子分布更接近后验概率密度。因此,只需少量粒子就能实现较高的估计精度,从而降低计算量。同时,采用了多级检测的形式,通过逐级干扰对消来去除干扰用户对检测性能的影响。通过仿真分析证明了改进后的粒子滤波检测方法检测性能优于传统粒子滤波检测器,有效地解决了传统方法中计算量大、检测效率低的问题,同时具有较好的抗远近效应能力。

The choice of proposal distribution of particle filter algorithm directly determines the estimation accuracy and computational complexity. According to the computational complexity of traditional particle filter for multi-user detection, a novel particle filter is proposed which the proposal distribution is determined by maximum likelihood estimation. The uniform distributed particles are re-selected based on maximum likelihood estimation so that the particles distribution of the optimized selection is closer to poster probability density. Therefore, a small amount of particles is needed to achieve higher accuracy, while the computational complexity is reduced. Multistage detector is adapted to eliminate interference usest effects through successive interference cancellation. Simulation results prove that the detection performance of improved detector is superior to traditional ones. The problem of computational complexity and lower detection efficiency in traditional detector is effectively solved, and the improved method performances better in near-far resistance.