基于局域波分解的非参数概率密度估计

Nonparametric probability density estimation based on local wave decomposition

局域波分解在提取信号趋势方面具有异乎寻常的效果.根据局域波缓变趋势提取算法,在小波概率密度估计思路的基础上,结合密度估计的直方图法,建立了局域波概率密度估计新方法.此方法能有效去除样本数据直方图中的高频成分,获得低频趋势,即概率密度.在混合高斯概率密度估计中的应用表明,对于无断点的密度函数,其具有计算简单、精度较高的优点.

Local wave decomposition is good at accurate trend extracting.Therefore,based on the algorithm of slow-varying trend extracting of local wave method,as well as wavelet probability density estimation method and histogram method,a new probability density estimation method is presented.The proposed method can get rid of high frequency components of histogram and obtain the low frequency trend,i.e.probability density.The application of this method to Gaussian mixture model density estimation proves the advantages of the approach for non-breakpoint density function estimation.And it is easier in computation and more accurate.