This paper proposes a method to generate a hierarchical knowledge base oriented to pattern recognition based on example attribute learning. The primary goal of this study is to extend the recognition process from the simple low level of the sample's memory to high levels of their conceptual memory so that the PR process can be brought about on different conceptual levels. The authors combine the traditional AI method with the modern artificial neural network method to make concepts obtained from training samples have very strong descriptive power for objects to be recognized. Algorithms for constructing attribute classes and knowledge bases are given which have been applied in a case study of handwritten character recognition. The test results show that the system proposed can acquire a high recognition rate when it has learned enough training samples.
展开▼
