摘要
AR模型、混合高斯模型分别可以很好地拟合样本的功率谱和概率密度。AR模型参数估计可以使用最大似然估计法(MLE);而在各高斯分量概率密度互不重叠的条件下,使用动态簇算法(DC)则可快速而精确地估计出混合高斯模型参数。使用MLE/DC参数估计方法,并在两种方法间建立一定耦合,即可对有色非高斯数据进行准确的功率谱和概率密度建模,进而实现数据的预白与高斯化。
The PSD and PDF of sample series can be well fit with the AR model and Gaussian mixture model respectively. Parameters ofAR PSD Model can be obtained by maximum likelihood estimate (MLE). And if there are no overlaps between each Gaussian component, parameters of Gaussian mixture PDF model can be exact estimated quickly with the dynamic cluster algorithm (DC). With MLE/DC and some coupling between them, the PSD and PDF parameters of colored non-Gaussian data can be estimated correctly. So processing of prewhitening and gaussianizing can be realized.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2006年第11期3196-3199,共4页
Journal of System Simulation