dc.contributor.author | Moodley, D![]() |
|
dc.contributor.author | Ram, V![]() |
|
dc.contributor.editor | Ram, Vevek![]() |
|
dc.date.accessioned | 2018-08-15T10:21:30Z | |
dc.date.available | 2018-08-15T10:21:30Z | |
dc.date.issued | 1996 | |
dc.identifier.citation | Moodley, D. & Ram, V. (1996) Feature extraction in neural networks: an application in handwritten character recognition. Industry Meets Academia: Proceedings of the 1996 National Research and Development Conference, The South African Institute of Computer Scientists and Information Technologists, Interaction Conference Centre, University of Natal, Durban, 26 & 27 September, hosted by The Department of Computer Science and Information Systems, University of Natal, Pietermaritzburg, edited by Vevek Ram, (ISBN 0-620-20568-7). | en |
dc.identifier.isbn | 0-620-20568-7 | |
dc.identifier.uri | http://hdl.handle.net/10500/24639 | |
dc.description.abstract | This paper examines and provides an overview of feature extraction techniques that are currently used in neural networks for image recognition. These include moment invariants, Zemike moments, Fourier descriptor techniques, Gabor and wavelet filters and the Neocognitron. An implementation of an handwritten character recognition system is discussed to illustrate the practical significance of feature extraction. The methods of Zemike moments and the Neocognitron are used for feature extraction and the multilayered perceptron is used as the classifier. The results are also compared to the same application without feature extraction. | en |
dc.language.iso | en | en |
dc.title | Feature extraction in neural networks: an application in handwritten character recognition | en |