Title :
Combined orthogonal vector and pseudo-inverse approach for robust pattern recognition
Author :
Chao, Daniel Y. ; Wang, David T.
Author_Institution :
Dept. of Comput. & Inf. Sci., New Jersey Inst. of Technol., Newark, NJ, USA
Abstract :
A neural-based orthogonal vector with a pseudo-inverse approach for pattern recognition is presented. A method for generating N orthogonal vectors for an N-neuron network is also presented. This approach converges the input to the corresponding orthogonal vector representing the prototype vector. Moreover, it can restore an image to the original image and thus has error recovery capability. Patterns with 20% error were recognized, using English capital latters as an example. The robustness of this approach is obvious
Keywords :
convergence; error correction; image reconstruction; inverse problems; neural nets; pattern recognition; vectors; English capital latters; convergence; error recovery capability; image restoration; neural-based orthogonal vector; pseudo-inverse approach; robust pattern recognition; Associative memory; Chaos; Hamming distance; Image restoration; Information science; Neural networks; Neurons; Pattern recognition; Prototypes; Robustness;
Conference_Titel :
Neuroinformatics and Neurocomputers, 1992., RNNS/IEEE Symposium on
Conference_Location :
Rostov-on-Don
Print_ISBN :
0-7803-0809-3
DOI :
10.1109/RNNS.1992.268629
