Hybrid Recognition for One Stroke Style Cursive Handwriting Characters

On-line handwriting recognition has continued to persist as a popular research field while pen computing applications are widely used in recent years. This paper proposes a novel hybrid system for one stroke style cursive handwriting character recognition. In the system, user can use fingertip to write various kinds of virtual characters (represented by trajectory of fingertip) such as alpha-numeric characters and Chinese characters through a digital camera based user interface. Without pen-up and pen-down information, the virtual characters are written in one stroke. An on-line and an off-line recognition method for such kind of cursive characters are proposed. A hybrid approach of these two methods is proposed to combine the advantages of both of them. Benefit from the integration, the recognition accuracy was increased from 80.6% (off-line classifier) and 83.4% (on-line classifier) to 90.9% (integrated) for stroke order free one stroke cursive handwriting Chinese characters.

• Type: Conference
• Year: 2005
• Where: ICDAR
• Authors: Teng Long, Lianwen Jin
• Keywords: Cursive Handwriting, Document Analysis, ICDAR 2005, On-Line Handwriting Recognition, Virtual Characters