摘要
多谐波正弦拟合算法常用最小二乘法将多谐波正弦信号采样数据拟合到多谐波正弦函数模型中.而在实际测量过程中,当相关噪声存在时最小二乘法的拟合精度会下降,为减少相关噪声对多谐波正弦拟合算法拟合精度的影响,在四参数正弦拟合算法的基础上提出了一种基于加权最小二乘的多谐波正弦拟合算法.并结合遗传算法避免了四参数正弦拟合算法对初始估计频率的依赖,提高了算法的鲁棒性.将算法在CPU/GPU平台上进行了实现,提高了算法的执行效率.在CPU/GPU平台下对受相关噪声污染的多谐波正弦信号进行了分析.实验结果表明,相比基于遗传算法的多谐波正弦拟合算法,所提算法的谐波幅值估计精度提高了1个数量级,算法执行时间缩短了近96%.
Multi-harmonic sine fitting algorithm provided a least squares method for fitting the sampling data of the multi-harmonic sinusoidal signals to the multi-harmonic sine function model. However when correlated noise exists in test process, least squares fitting accuracy would considerably reduce. Based on four parameter sine fitting algorithm（FSFA）, a multi-harmonic sine fitting algorithm based the weighted least squares was presented. Combined with genetic algorithm, proposed algorithm avoided dependence on the initial estimated frequency and improved the robustness of the algorithm. The algorithm was realized in the CPU/GPU platform, and the efficiency of the algorithm was improved. In the CPU/GPU platform, sinusoidal signals disturbed by harmonics and noise were analyzed. Experimental results show that harmonic parameter estimation accuracy is improved by one order of magnitude and the processing time is shortened by over 960%
出处
《中北大学学报（自然科学版）》
CAS
北大核心
2013年第6期636-641,共6页
Journal of North University of China(Natural Science Edition)
基金
国家自然科学基金资助项目(51275492)
国际科技合作基金资助项目(2010DFB10480)
关键词
CPU
GPU
多谐波正弦拟合
遗传算法
并行处理
参数估计
CPU/GPU
multi estimation harmonic sine fitting
genetic using the recommended algorithm. algorithm
parallel processing
parameter
