Research On Off-line Chinese Signature Automatic Verification Method Based On BP Neural Network

As a mean of identification, the handwritten signatures are widely used in the real world, however, which also has the risks of being easily forged. Besides, the artificial identification of fake signatures is always short of efficiency and accuracy. To solve this issue, although a large number of researchers, from the 1960's to now, have carried out fruitful research works and made some progress in the aspect of using computer technology to automatically identify Off-line Chinese handwritten signature, they are still facing some problems and a certain insufficient. Therefore, it is necessary to carry out further researches on the automatic identification methods of Off-line Chinese handwriting signature.To support the operation of automatically identifying offline Chinese signatures, this paper in the use of mature technologies existed to prior handle the signature images, which includes the binarization method for image based on OTSU, the smoothing method for original image based on median filtering, the eliminating blank area method for image based on scanning image pixels and the method of adjusting lean image via Hough transform to detect approximate linear existed in the signature.Considering some inadequacies of the existing signature identification method based on skeleton segment information, this paper not only continues following the traditional skeleton segment to design skeleton segment trace feature, but also extracts outline information of Chinese signature and segments it to obtain profile segment motion feature of Chinese signature. Then, the feature vectors of automatic identification of Chinese signature were constructed properly.Analyzing the differences between multiple signature feature vectors of Chinese characters in the same group, this paper obtains a number of proper identify parameters for automatic identification. At the same time, the signature segments of skeleton and contour were matched to construct classifier of BP neural network, so that the classifier can complete Chinese automatic identification work of offline signature. At last,this paper proposes Off-line handwritten Chinese signature verification algorithm based on BP neural network.This paper designs training experiments and identify experiments results based on the data collected, of which the result shows that the proposed method is effective.Meanwhile,comparing with other existing Off-line Chinese signature verification method, the method proposed in this paper has a better accuracy and an acceptable rate of rejecting error relatively.