lundonj@ul.ie (10/24/90)
I am presently working on a NN problem that will form the basis of my thesis concerning the separation of machine ( easier ) and hopefully handwritten ( much harder ) characters from a page of text. This algorithm will separate the words into letters sufficiently enough to enter them to a trained NN for analysis ( backpropagation? ). My question is could anybody out there provide me with their approaches/techniques/ references to similar problems that you may have encountered especially the following 1. segmentation/separation of the characters from the inputted bit-map page. 2. Most suitable preprocessing technique to use to reduce the dimensionality of the NN input if any at all. 3. Is back-propagation or any of its derivatives most suitable for a problem of this kind. 4. Which classical approaches should I compare the final results to. All replies will be acknowledged. If the response is large enough the I will endeavour to concatenate them into a mini report for the net. Thanks in advance, James A. Lundon. -------------------------------------------------------------------------- | James A. Lundon, | | | Postgraduate Student, | "Regret for the things we did | | Dept. of Mathematics, | can be tempered by time. It | | University of Limerick, | is the things that we did not | | Limerick, | do that is inconsolable" | | Ireland. | | | | Sidney J. Harris. | | Phone number (061) 333644 X 2148 | | | E-mail lundonj@ul.ie | | --------------------------------------------------------------------------