Hand written signature is complex behavioral biometric trait and is widely accepted for personal and document authentication. In this paper we propose Off-line Signature Identification Based on Discrete Wavelet Transform (DWT) and Spatial Domain Features (OSIDS) method. The method is tested using genuine and skilled forgery signatures. The signature is preprocessed using edge detection, filtering and morphological operation to convert into single pixel width. The global features are extracted from preprocessed signature. The DWT is applied on original signature to obtain features from four sub bands. The global features are fused with DWT features to derive final set of features. The test signature features are compared with data base signature features vector using correlation technique. It is observed that the values of FAR and EER are low in the case of proposed algorithm compare to existing algorithm. As FAR value is less, that indicates skilled forgery is successfully rejects.
展开▼
