An Enhanced Approach to Memetic Algorithm Used for Character Recognition

Character recognition is a best case to apply logics from Memetic Algorithms (MA) for image processing. In cases, like finger print matching, cent percent accuracy is expected but the character recognition on other hand can auto correct some errors. Time of processing is not the first criteria in figure print analysis but accuracy is a must, whereas while extracting characters from image speed of processing becomes more important parameter. This aspect of character recognition provides wide scope of implementing MA. The typing on QWERTY keyboard is the best example of brain using MA and dividing the character search in two parts with 13 characters for left hand and 13 for right. We never need to cross hands for typing next character as the design of keyboard is ensures that in most of the cases consecutive characters appear in specific sequence and brain keeps itself already prepared to hit next key but waits for confirmation. As we move dipper into string the search starts reducing which further enhances the predictive capacity of brain for expected next character. This can be seen as local search and cultural evolution which is performed by brain with each successive character in a string in [1, 2]. Hence character recognition is more of string recognition than treating every character in isolation. This paper explains the above theory with results and also presents an enhanced MA for character recognition.