Signature feature recognition method and system and based on mask pre-training model, equipment and storage medium

Abstract

The invention discloses a handwritten data signature recognition method based on a mask pre-training model, and relates to the technical field of data recognition processing. The method comprises steps of extracting various writing feature information generated by writing on terminal equipment; preprocessing the plurality of feature sequences to obtain a plurality of feature data sequences with fixed lengths; covering and shielding partial feature values in the feature data sequence; converting a plurality of feature data sequences of which partial feature values are covered into feature vectors of corresponding features, superposing all the feature vectors and carrying out layer normalization and coding, extracting a feature sequence corresponding to the covered part, and confirming a mean square error function by combining a predefined target value; and updating the network weight of the prediction model through a mean square error function, and constructing the prediction model to identify the handwriting features. The invention can be widely applied to electronic signature recognition.

Description

technical field [0001] The invention relates to a handwritten signature data recognition technology, in particular to a handwritten online signature data recognition technology based on a pre-training model. Background technique [0002] Since the advent of the digital age, handwritten online signatures have been used as electronic data deposits in government, business, and financial neighborhoods, enhancing the authenticity of electronic data. Online signature data has the characteristics of real-time authentication. As long as the user signs his name, the system can recognize it. However, different models need to be trained for different tasks, which increases costs and reduces production efficiency. This paper proposes a mask-based pre-training model that enables transfer learning for similar tasks, reducing industrial costs and improving production efficiency. [0003] The mask pre-training model of the handwritten signature comes from the mask-based natural language pr...

AI Technical Summary

Problems solved by technology

[0005] The present invention proposes an online handwritten signature recognition based on mask pre-training in view of the shortcomings of the prior art, such as large amount of model calculation, poor mobility, single consideration of writing features, poor versatility and lack of pre-training technology in the electronic signature recognition method method and system

Images