Context-Aware Systems and Applications. First International Conference, ICCASA 2012, Ho Chi Minh City, Vietnam, November 26-27, 2012, Revised Selected Papers

Research Article

Handwriting Recognition Using B-Spline Curve

Download
593 downloads
  • @INPROCEEDINGS{10.1007/978-3-642-36642-0_33,
        author={Khoi Nguyen-Tan and Nguyen Nguyen-Hoang},
        title={Handwriting Recognition Using B-Spline Curve},
        proceedings={Context-Aware Systems and Applications. First International Conference, ICCASA 2012, Ho Chi Minh City, Vietnam, November 26-27, 2012, Revised Selected Papers},
        proceedings_a={ICCASA},
        year={2013},
        month={2},
        keywords={B-Spline curve matching curve handwriting Optical character recognition reconstruction},
        doi={10.1007/978-3-642-36642-0_33}
    }
    
  • Khoi Nguyen-Tan
    Nguyen Nguyen-Hoang
    Year: 2013
    Handwriting Recognition Using B-Spline Curve
    ICCASA
    Springer
    DOI: 10.1007/978-3-642-36642-0_33
Khoi Nguyen-Tan1,*, Nguyen Nguyen-Hoang1,*
  • 1: DaNang University of Technology
*Contact email: ntkhoi@dut.udn.vn, hoangnguyenbkit@gmail.com

Abstract

This paper aims at presenting novel approach for curve matching and character recognition such as printed writing, handwriting, signatures, etc. based on B-Spline curve. The advantages of the B-Spline that are continuous curve representation and affine invariant, and the robustness. The recognition process is composed of two main steps: sample training and recognition. The computer must be trained with data from bitmap image file. The next step is pre-processing input data from the binary image and finding its skeleton. The reconstruction of a B-Spline curve representing the sample character is applied to find out the control points. Then the sample B-spline curve of each character is stored in a database. For the test character, it has the same process with the sample character. The matching is done by computing the Euclidean distance between the control points of test curve with those of all sample characters to recognize the character. The experimental results show the performance of the proposed algorithm.