Research And Implementation Of Online Handwritten Signature Authentication Technology

With the development of Internet technology and hardware devices,in recent years,researchers attach great importance to the research of identity authentication technology,especially biometric technology.As a kind of biological behavior characteristics,handwritten signature has the distinguishing features of each user's signature habits,methods and data characteristics.Signature data also has unique advantages:strong user friendliness,easy access and so on.Among them,the online handwritten signature data collected the dynamic characteristics of the signer,which is more difficult to forge.There are some problems in the current signature authentication algorithms,such as:the feature extraction depends on the writer,and the number of reference signatures is large.At the same time,the traditional online signature authentication system in the past is often implemented by template matching method,which is relatively few in the field of deep learning.This paper studies the online signature authentication technology in deep learning,which has a certain supplement and improvement.The main contents of this paper are as follows:(1)Preprocessing and functional feature extraction algorithmThis paper constructs a Chinese online handwritten signature data set,selects the appropriate preprocessing technology for the online signature data,expands the original signature data set features to obtain the initial set of functional features,and then uses genetic algorithm to select features in the initial set,analyzes its performance and obtains good performance.(2)A writer independent online handwritten signature authentication model based on neural network is constructedBy analyzing the characteristics of online handwritten signature data and signature authentication task,the self encoder based on bigru and attention mechanism are used to extract the global feature vector,and then the global feature vector is input into triplet network,and a model of autoencoder+attention+triplet is constructed to complete online handwritten signature authentication.3)A prototype system of online signature authentication is designed and implementedUsing the proposed signature authentication model as the core of back-end algorithm,an online signature authentication system is designed and implemented on the user interface to verify its effectiveness.(4)The proposed model is analyzed and optimizedThe proposed autoencoder triplet model is tested on the online handwritten signature data set collected in this paper,and compared and analyzed with other methods.At the same time,the influence of the scale of the reference set and the different values of the sampling interval on the experimental results is studied,which has a certain value for the follow-up work.