Constructing attribute classes by example learning: the research of attribute-based knowledge-style pattern recognition
Lijia Zhou, Hongxun Song, Shukui Zhao · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 1991
ABSTRACTO!:6ervinS and stu4ving the pr of man's recogridon of patten We find that man's thinkiig xdvid more ofter ,go into the deep leveis of coireprtai memeiy thin remain at physiological level of sample memory. This paper propces a method to generatehierarthical knowledge e oriented to pattern rognidon ced on example attribute lealrarig The pnmary goal of this study is to extend the rognition prc from the simplelow leve] of sample's memory to high levels of their concepn1 memory so that PR pz can be brought about on thiferent corceptual levels. In oz ppzoach we combine the tradonal Al method with modem Artificial Neural Network method so to make coirepts obtained from trainizg samples hive very strong derziptive power for objts to be recognized Adgorithms for cortrudng attribute ciasac and knowledge base are given in thepaper which hve been applied in a ca.ce stz4r of hendwritren charxter rognidon The testresults show that the system pzopced can qAre high rognition rate when it hes Ieaznt