Seyed Aliakbar Mousavi ¹, Majid Meghdadi ², Zahra Hanifeloo ², Putra Sumari ¹, Muhammad Rafie Mohd Arshad ¹

Affiliations
1 School of Computer Science, Universiti Sains Malaysia (USM), MY
2 Computer Department, Faculty of Engineering, University of Zanjan, IR

Abstract

This paper provides a fast method to recognize a variety of Persian banknotes at different scales. In this technique, the PCA, LDA and sparse representation methods are utilized at feature extraction stage and follows with MLP neural networks, LVQ and SOM in classification. Finally, the application of sparse matrix representation method and combination of both SOM and LVQ neural networks would lead to the best efficiency with precision of 91.15% in recognition of Persian banknotes particularly the worn ones.

Keywords

Banknote Recognition, LVQ, Neural Network, PCA, Sparse Representation.

Full Text