A Robust Method for Recognizing Accents in Vietnamese Handwriting Characters

Handwriting character recognition is one of the most common research topics. Many approaches have applied to English characters and achieve high accuracy. However, the complexities in the language of each country are not same. Vietnamese handwriting character recognition is facing many problems, most of them come from the accent. This paper focuses on accent recognition, especially when there is a connection between two accents a common problem which affects the identification result. Our approach starts with separating accent from character using the connected-component labeling method. The obtained accent then is checked if it is single or multiple (the combination of many accents). The recognition is performed using support vector machines with the single accent, or hidden Markov model if the accent is multiple. Proposed solution has been tested and obtained high accuracy. Keywords—Vietnamese handwriing character, accent, corner detector, branch separating, invariant moment, hidden Markov models (HMMs)