Writer Recognition In Handwritten Documents Based On Deep Learning

Handwriting identification,as an effective verification method for identity recognition,has attracted wide attention from the industry,and has made a series of breakthroughs.Although the research on handwriting identification has achieved fruitful results,these studies are mostly based on online handwriting identification,and the research on offline handwriting recognition needs to be further strengthened.Because offline handwriting is more versatile and easy to obtain in material acquisition,offline handwriting identification has a wider range of application scenarios than online handwriting identification.With the vigorous development of the field of big data applications,technical research in the direction of machine learning has made great progress,which provides a new technical solution for solving problems in computer vision and images.To this end,based on deep learning algorithms,this paper conducts in-depth research on offline handwriting identification,which uses convolutional neural networks to identify offline handwriting to improve recognition accuracy,and proves the effectiveness of this method through experiments.The main contents of this article are as follows:This paper proposes a deep learning-based handwriting handwriting identification method.For the first time,the Beta Elliptic model is used in Chinese handwriting identification to cut and classify Chinese characters.First,based on the stroke characteristics of the Beta Elliptic model and elementary perception code,the text handwriting is pre-processed and segmented into a series of Beta strokes.The classification is based on the curvature information of the Beta stroke segment and the membership of the elementary perception code,so as to achieve handwriting Text segmentation and pre-classification.Local features of offline handwriting were extracted based on convolutional neural network and principal component analysis was performed to remove redundant data and reduce the dimension.Then,the Fisher Vectors algorithm is used to encode the local features,and based on this,global features are generated for offline handwriting identification.Finally,Euclidean distance is selected as the similarity measurement method to determine the writer of the sample to be detected.In this paper,the method proposed in this paper is simulated and verified through a large number of experiments.Validation of related theories on Chinese data sets shows that the results are satisfactory.In addition,the handwriting identification is compared on two different foreign language datasets,and the experimental results are convincing.Through experimental analysis,the method proposed in this paper has a certain effect on handwriting identification,and has stronger adaptability and robustness than traditional methods.