In this paper, we mainly describe a new approach of Handwritten Chinese Character Recognition (HCCR), which is based on eigen-character extraction. The procedure of the eigen-character extraction method is explained including initialization, eigen character extraction (or eigen spaces generation) and character recognition. Two different methods are presented to do eigen character recognition respectively. Besides, k Nearest Neighbor (kNN) is implemented to improve the recognition rate of the new approach. In the end, a comparison is made between the eigen-character extraction approach and other existing approaches through simulation based experiments. The results show that our approach has a satisfying rate and could be further improved if combined with some other methods such as elastic matching and wavelet methods.
Handwritten chinese character recognition using eigenspace decomposition
Kuruoglu E E
2012
Abstract
In this paper, we mainly describe a new approach of Handwritten Chinese Character Recognition (HCCR), which is based on eigen-character extraction. The procedure of the eigen-character extraction method is explained including initialization, eigen character extraction (or eigen spaces generation) and character recognition. Two different methods are presented to do eigen character recognition respectively. Besides, k Nearest Neighbor (kNN) is implemented to improve the recognition rate of the new approach. In the end, a comparison is made between the eigen-character extraction approach and other existing approaches through simulation based experiments. The results show that our approach has a satisfying rate and could be further improved if combined with some other methods such as elastic matching and wavelet methods.File | Dimensione | Formato | |
---|---|---|---|
prod_213000-doc_48784.pdf
accesso aperto
Descrizione: Handwritten chinese character recognition using eigenspace decomposition
Dimensione
1.49 MB
Formato
Adobe PDF
|
1.49 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.