连续特征空间离散化及类条件概率分布估计  

Discretization of Continues Feature Space and Estimation for Class-Conditional-Probability Distribution

作　　者：周德全[1] 刘国岁[2] 吴黎光[2] 苏卫民[2] 

机构地区：[1]空军第一航空学院 [2]南京理工大学电光学院

出　　处：《信号处理》1998年第A12期114-117,共4页Journal of Signal Processing

基　　金：兵器工业总公司"九五"预研项目!34.6.1

摘　　要：贝叶斯决策理论在模式识别、信号检测中得到了非常广泛的应用。但这一理论的应用是以己知类条件概率密度为前提。本文提出了类条件概率密度随机变量（特征）空间离散化及类条件概率分布估计方法。给出了以此为基础构成的贝叶斯分类算法。最后，将此方法用于雷达目标的识别。The Bayesian decision theory is widely used in pattem recognition and signal detection. Only when Class-conditional -probability density is known, the theory can be used. A discrete method for stochastic variable (features) space of class-cond itional-probability density and estimation method for class-conditional -probability distribution is proposed. Bayesian classification algorithm based on the method is given. Finally, the mcthods are applied to recognize radar targets.

关 键 词：模式识别 条件概率 离散化 贝叶斯分类 

分 类 号：O235[理学—运筹学与控制论] TP391.4[理学—数学]