Wavelet Based Approach for Off Line Handwritten Signatures Verification

Though a lot of research has been undertaken in the area of handwritten signatures verification, the recognition rates still need to be improved. The low recognition rates are largely attributed to the fact thatthere are intra personal variations in an individual’ssignature. In this paper a wavelet–based off–line signatureverification system is proposed. The proposed system extracts the low frequency components and also the high frequency components which are also known as approximation coefficients and detail coefficients respectively in wavelet terminology. The high frequency components represent the fast changing parts of the signal and the low frequency components represent the less varying or smooth parts of the signal. The handwritten signature images written on the papers are scanned into the machine and stored in jpeg format. The signatures images are binarized and bounding rectangles are put covering only the signature area. The bounded signatures are normalized using the Bicubic interpolation method and are thinned. Since the signatures of the same person though not identical, but definitely exhibit some form of consistency or stability, hence regularity analysis has been done using a regular wavelet like Daubechies wavelet transform on the pre processed handwritten signature images. The decomposition is done for six levels on the signature images which are in jpeg format and the principal component analysis is done. The principal components are chosen according to the ‘kais’ rule. The principal component vector is used to train the standard K-NN classifier and classification is done. The results of the proposed method are quite satisfactory in case of random forgeries. Certainly for the skilled forgeries, the method needs to be improved.