摘要
针对通信信号检测的研究中存在大规模计算的问题,在实验室算法仿真阶段通过CPU对码元串行解调的蒙特卡罗仿真面临巨大的时间和计算资源开销,无法实现快速仿真验证实验,制约了研究工作的进展。以2FSK通信信号为例,利用过零周期数作为混沌相变判别指标,建立了基于GPU的混沌解调并行蒙特卡罗仿真模型,利用GPU高效的并行能力实现了快速的2FSK通信信号混沌解调的并行蒙特卡罗仿真。实验室对106量级的随机码元仿真显示该并行蒙特卡罗仿真模型可以实现近24倍的加速比,能高效地实现通信信号的混沌解调,为混沌解调算法研究奠定了基础。
The obstacle we face in the process of chaos based demodulation algorithm research using Duffing oscillator is of computational complexity of the numerical solution of stochastic differential equations （SDEs）. In this paper, we combined the modern GPU with this data - intensive problem and designed a parallel Monte - Carlo simulation model for 2FSK demodulation algorithms using Duffing oscillator. In the model, we chose zero - crossing number as the criterion for justification of phase transition. In our experiments, 10^6 binary codes were tested under different SNRs. The results show that the speed of this parallel model conducted on GPU was one order of magnitude faster than that on the CPU platform, and by properly designing the datum size, this model can aehieve a highest speedup ratio of 24. This parallel model can greatly enhance the simulation time and give a common model for the research of the effectiveness and reliability of different demodulation algorithms.
出处
《计算机仿真》
CSCD
北大核心
2014年第6期207-211,共5页
Computer Simulation
基金
国家自然基金"泰山学者建设工程专项经费"(61179018)
关键词
混沌解调
蒙特卡罗仿真
杜芬振子
并行
Chaos based demodulation
Monte Carlo simulation
Duffing oscillator
Parallel
