用模式识别方法分析心血管病的测试数据

An Analysis of Test Data of Coronary Thrombosis by Pattern Recognition

本文对心血管病诊断中血液流变学的测试样本,用KL展开进行特征压缩,并用Fisher准则与逐次校正相结合的方法,对测试样本及特征压缩后的样本进行分类分析,获得较好的效果。

The technique of analyzing the test data of coronary thrombosis with pattern recognition is described. The feature contraction of the testing sample of hemor-heology in coronary thrombosis diagnosis is carried out by KL transform. By combining the Fisher Rule and successive correction method, a classifying analysis is performed on the testing sample and on the sample after feature contraction. Withthe sample feature and Fisher Rule L=(uA- uB)2/(σB2-σB2), the coefficient Ci of thelinear discriminant function u=∑N Cixi and the critical value uC = (MA uA +MBu B )/(MA+MB)are determined.This is used as the initial value and a linear discriminant function is obtained by successive correction. The method C and the leaving- one -out method are then used to estimate the upper limit and lower limit of error probability of the classi-ficator. Test results are found to be satisfactory.