An accurate handwritten digits recognition system based on DWT and FCT

Abstract EN

In this paper an accurate Indian handwritten digits recognition system is proposed.

The system used three proposed method for extracting the most effecting features to represent the characteristic of each digit.

Discrete Wavelet Transform (DWT) at level one and Fast Cosine Transform (FCT) is used for features extraction from the thinned image.

Besides that, the system used a standard database which is ADBase database for evaluation.

The extracted features were classified with K-Nearest Neighbor (KNN) classifier based on cityblock distance function and the experimental results show that the proposed system achieved 98.2% recognition rate.