Recognition of Humboldt's Handwriting in Complex Surroundings

TitleRecognition of Humboldt's Handwriting in Complex Surroundings
Publication TypeConference Paper
Year of Publication2010
AuthorsSteinke, KH
EditorGehrke, M.
Tertiary AuthorsDzido, R.
Conference Name12th International Conference on Frontiers in Handwriting Recognition (ICFHR)
Date Published11/2010
Conference LocationKolkata
KeywordsLegendre polynomials, text localization, writer recognition
AbstractIn the Botanic Museum of Berlin in Germany exist about 3.5 million dried plants on paper sheets. They were collected in the last two hundred years by many collectors among those was the famous researcher “Alexander von Humboldt”. because e every collector left his handwriting on the sheet the question came up if it is possible to find out automatically which sheet belongs to Humboldt’s collection. To solve the problem many challenging sub problems have to be solved before. First text regions have to be localized and extracted. Then printing is to be distinguished from handwriting. For the latter an off-line writer recognition procedure had to be developed in order to determine the writer of the handwriting. Often two or more writers can be found on one sheet. To make sure that the suspect writer is correct an interactive method is applied which transforms the static character into a dynamic form. Different mathematical procedures are used such as the reproduction of the write line of individual characters by Legendre polynomials. All methods were proved on the international IAMdatabase [3]. 93 writers with at least 5 samples were chosen from the IAM-database. By combining multiple characters (up to thirteen), the recognition rate rises considerably and reaches 98.7%. A global statistical approach using the whole handwritten text results in a similar recognition rate of more than 98%. By combining the methods by a ranking method a recognition rate of 99.5% is achieved.
URLhttp://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=5693621
DOI10.1109/ICFHR.2010.91