A hierarchical approach to fuzzy possibilistic-reasoning for recognising chinese characters

Abstract

The paper presents a system based on a new paradigm of a three-layer hierarchy with fuzzy possibilistic reasoning rules and associative memory (AM) neural networks to recognise radicals in Chinese characters. The system extracts features of the complex topological structure of characters and constitutes a hierarchical representation based on their physical shape and classified according to their logical meanings and pattern structure. Using fuzzy possibilistic reasoning rules, the system applies this hierarchy to identify individual radicals embedded within a character. An associative memory neural network is employed to recognise identified radicals. Comparison of this system with other methods is carried out. Several special cases and application limitations of the system are discussed. Test results show that the system is effective and reasonable.

Keywords:

• Associative-memory neural network
• Chinese character recognition
• fuzzy possibilistic reasoning
• inference rules
• extraction

C.R.Categories:

• F1.1
• I.2.3
• I.5.1
• C.1.3