Online signature identification method based on sample synthesis and sorting learning

Abstract

The invention discloses an online signature identification method based on sample synthesis and sorting learning. The method comprises the following steps: acquiring a signature sequence and preprocessing the signature sequence; establishing a Sigma Lognormal fitting model of the signature sequence; adding disturbance sampling to the fitting parameters to synthesize a real signature and a forged signature; carrying out sorting learning to optimize average precision loss training neural network model parameters; and for the signature sequence to be identified, extracting features by using the trained neural network model, normalizing the features, and calculating an Euclidean distance between feature vectors to judge authenticity. According to the online signature identification method, the problem that forged samples are difficult to obtain in an online signature identification task is solved, the Sigma Lognormal model is used for generating the samples, collection of forged signature data is not depended on any more, sorting learning is carried out to learn correlation and similarity information between the signature samples, and the high-precision online signature identification method is achieved. The invention has the characteristics of high accuracy, good adaptability and the like, and has good practical value.

Description

technical field [0001] The invention relates to the technical fields of deep learning and artificial intelligence, in particular to an online signature verification method based on sample synthesis and ranking learning. Background technique [0002] As an important method of identity authentication, handwritten signatures are widely used in finance, justice, and banking. Building an automatic signature authentication system has good practical application scenarios and practical value. Signature verification is divided into online signature verification and offline signature verification. Compared with offline signature verification, online signature can collect pressure and speed information during the writing process, which can better reflect the user's writing habits. With the popularization of online writing collection equipment, online signature verification tasks are further widely used. [0003] Traditional online signature verification methods often use manually desi...

AI Technical Summary

Problems solved by technology

[0005] In order to deal with the difficulties and challenges existing in the existing online signature verification technology, the present invention makes full use of the SigmaLognormal model to synthesize data to solve the problem that forged samples are difficult to obtain in practical applications, and then learns finer-grained features by optimizing the average precision loss sorting learning method , providing an online signature verification method based on sample synthesis and ranking learning

Images