User-independent online signature recognition method and device based on RNN model

Abstract

The invention provides a user-independent online signature recognition method and device based on an RNN model. The method comprises the steps that a registration signature and a to-be-authenticated signature are split according to strokes to obtain multiple parts of stroke data, then the stroke data of the registration signature and the stroke data of the to-be-authenticated signature are connected together, and the registration signature is a signature registered by a user in advance; extracting the connected stroke data by using a stroke RNN model to obtain stroke features corresponding tothe stroke data; inputting the stroke features into a signature RNN model to extract overall signature features, and calculating values S1 and S2 corresponding to similarity and difference of the overall signature features; and performing operation according to the values S1 and S2 corresponding to the overall signature feature similarity and the difference degree, finally obtaining the similaritybetween the registered signature and the signature to be authenticated, and outputting an authentication result. According to the method, fewer features and fewer training samples are used, the precision is improved, and the training speed and the running speed are higher.

Description

technical field [0001] The present invention relates to the technical field of signature detection, in particular to a user-independent online signature verification method and device based on an RNN model. Background technique [0002] Handwritten signatures are divided into online signatures (online signatures) and offline signatures (offline signatures). Offline signatures are static images, and online signatures are serialized data with writing tracks, which require special handwriting equipment. Some handwriting equipment is not only It can record the movement trajectory of writing, and can also record the movement trajectory of the stylus when it is lifted without touching the screen. This is called the pen-up trajectory. At present, the characteristics of the pen-lift trajectory are very useful in some fields, but most handwriting devices, especially those that can only be input through mobile phones and pad devices, cannot record the pen-lift trajectory, resulting in...

AI Technical Summary

Problems solved by technology

[0005] The present invention provides a user-independent online signature verification method and device based on an RNN model to at least solve the problem of poor user-independent handwritten signature verification in the existing field of online signature verification

Images