A METHOD OF EXTRACTING FEATURES FROM LARGE SCALE PATTERNS
Ping Wang · Journal of Tianjin University of Science and Technology · 2003
The statistic and analysis of high-dimension data is the important difficulty in the present application and theory of statistics. According to the characteristic of patterns data and expert's experience, clustering and selecting typical patterns and calculating central fields are performed in sequence. Then using the distances between every sample and the central fields as features of the sample, 40 pieces of features are extracted from hundreds of data successfully. The validity of the method through which features are extracted is proved by the results of statistic analysis for the features. At the same time, the method to integrate features based on K-L transform is brought forward. which help us to fulfill the test to the difference of multi-dimension features between pattern classes.