超声回波特征值的统计学分析方法

Statistic evaluation of ultrasonic echo feature value

介绍了一种基于统计学假设检验理论进行模式识别中的特征评价与选择的方法,该方法理论意义比较清楚,计算简单,能够选出合适的特征.其实现是采用统计学中t检验方法来进行的,即用检验各组样本特征之间差异是否显著,来评价所提取的特征值的优劣;采用启发式方法和判别分析来进一步考察所提取的特征值的分类效果.实现了超声回波特征空间的降维处理与最佳特征子集选择.

The evaluation of feature values was carried out by checking the significance of difference between the features of each group. The heuristics method and discriminant analysis were used to research the classification results of features extracted. Finally, the dimensional reduction and selection of feature space for ultrasonic and the optimum feature subset have been successfully implemented using the method mentioned above.