OCR-Mirror Image Reflection Approach: Document Back Side Character Recognition by Using Neural Fuzzy Hybrid System

摘要

OCR is technical approach to analyze the handwritten text and turn it into a structure which the process the system more effectively for searching, re-storing, retrieval and indexing purpose. Many innovations are discovered in the field of OCR, but still many challengers are waiting for the solutions such as the recognition of document-backside-layered characters along with front side layer characters. The recognition of paper-backside-layered handwritten cum printed text is very difficult than the recognition of paper-frontside-layered characters. The fuzzy logic system process the data with help of bunch set of primary-based knowledge and delivers it into high successive recognition rate by using fuzzy classification functions, inference rules with decision-making procedures. Neural networks are compatible and best area to solve the pattern cum text recognition tasks, the learning based process start to implement from the basic imprecise data and algorithmic steps are formed with help of neuron-based learning process cum observations, but it is not effective to accomplish the user expected requirements for making the decisions. The successful combinational platform of neural fuzzy based closed loop system is proposing many technical ideas with effective solutions to solve the significant problems in OCR. This research methodology has proposed the mirror image reflection approach for recognition of document backside layer characters by using neural fuzzy hybrid system: it categorize the single page into the two sub layers, the first sub layer contains the paper front side text and the second sub layer contains paper back side text with corresponding pictures, characters, numerals and figures. Here the front sub layer text is completely bypassed with backside layer. The document-backside-layer input characters converted into neural-inputs, transformed it into fuzzy-sets, then fuzzy-rules applied based on the knowledge in fuzzification, related output responses are generated and the pre-cum-post-processing technique has applied to backside layer characters. Our ancestors used the palm leaves as the language-script-storage-tool and enclosed valuable information about the astrology, geographical, engineering, architecture, scientific, astronomy and so on. This hidden invention provides the technical ideas and alternate salvation for engineering and medical challenges. But the major problem is extracting the information from it without open, turn the pages. Difficult to open bunched of palm leaves from a bundle due to running out of their natural life. This paper is proposing ideological approach cum methodology with charitable path to solve the problem of extracting the document backside characters.