Improving the operability of personal health record system by dynamic dictionary configuration for OCR

Abstract

Information is managed by computers, which enabled us to handle, maintain, and distribute huge amount of data. However, data transfer from the source to the destination of information is not always achieved in idealway. One good example in Japan is personal health related information, such as the report of health check-up led by government. Paper-printed report is considered to be a reasonable way for distributing the report especially before the era when the internet becomes one of the infrastructures of the society. However, the Japanese government recognized the need of individual-based health maintenance for extending healthy life expectancy. In accordance with this context, PHR or personal health record systems are developed and operated for experimental proof. One of the difficulties revealed is low operability of data migration from printed data. One of the promising solutions for that issue is to apply OCR for printed document. In this paper, we propose a method of dictionary optimization to improve the accuracy of OCR by adaptable switching of dictionary parameter. Experimental results showed that the recognition accuracy improved for both Japanese letters and figures.