计算机科学

计算机科学 ›› 2026, Vol. 53 ›› Issue (8): 156-164.doi: 10.11896/jsjkx.250600092

• 计算机图形学 & 多媒体 • 上一篇    下一篇

基于多尺度方向注意力Transformer的笔迹性别识别方法

蔡毅1, 王晓宾1, 陈蕊丽1, 许金锋2   

  1. 1 中国人民公安大学侦查学院 北京 100038
    2 中国人民公安大学警体战训学院 北京 100038
  • 收稿日期:2025-06-13 修回日期:2025-09-22 出版日期:2026-08-15 发布日期:2026-08-17
  • 通讯作者: 王晓宾(xiaobin08d016@126.com)
  • 作者简介:(463575639@qq.com)
  • 基金资助:
    智能警务四川省重点实验室开放课题(ZNJW2023KFMS007);中国人民公安大学刑事科学技术双一流创新研究专项(2023SYL06)

Handwriting Gender Recognition Method Based on Multi-scale Directional Attention Transformer

CAI Yi1, WANG Xiaobin1, CHEN Ruili1, XU Jinfeng2   

  1. 1 School of Investigation, People’s Public Security University of China, Beijing 100038, China
    2 School of Police Training, People’s Public Security University of China, Beijing 100038, China
  • Received:2025-06-13 Revised:2025-09-22 Published:2026-08-15 Online:2026-08-17
  • About author:CAI Yi,born in 2001,master.His main research interest is document verification.
    WANG Xiaobin,born in 1989,Ph.D,associate professor.His main research interest is document verification.
  • Supported by:
    Key Laboratory of Intelligent Policing of Sichuan Province Project of China(ZNJW2023KFMS007) and Special Project Research of Double First-Class Innovation in Criminal Science and Technology of People’s Public Security University of China(2023SYL06).

PDF (PC)

99

摘要： 笔迹性别识别作为生物特征识别的重要分支,在身份验证和犯罪侦查等领域具有广泛的应用前景。传统方法主要依赖手工特征提取,存在特征表达不全面、跨语言适配性差等局限。为解决这些问题,提出了一种基于多尺度方向注意力的Transformer笔迹性别识别网络(MSDAttFormer)。该方法创新性地将Transformer的全局建模能力与笔迹特有的方向性特征相结合,通过多尺度特征融合模块整合不同层次的笔迹特征,引入笔迹方向注意力模块精确捕捉性别相关的方向性差异,采用改进的Transformer编码器建立特征间的长距离依赖关系。构建了包含1 620个书写者的中文笔迹性别识别数据集(CHAP),并在CHAP和希伯来语数据集(HHD)上进行了全面实验验证。结果表明,MSDAttFormer在CHAP数据集上的准确率达到84.46%,F1-Score达到83.97%,ROC AUC达到90.25%,在HHD数据集上的准确率达到85.71%,显著优于现有方法,验证了所提方法的有效性和跨语言泛化能力。

关键词: 笔迹性别识别, Transformer, 多尺度特征融合, 方向注意力, 深度学习

Abstract: Handwriting gender recognition,as an important branch of biometric recognition,has broad application prospects in identity verification and criminal investigation.Traditional methods mainly rely on manual feature extraction,which has limitations such as incomplete feature representation and poor cross-language adaptability.To address these problems,a multi-scale directional attention Transformer for handwriting gender recognition(MSDAttFormer) is proposed.This method innovatively combines the global modeling capability of Transformer with the directional features unique to handwriting,integrates handwriting features at different levels through a multi-scale feature fusion module,introduces a handwriting directional attention module to accurately capture gender-related directional differences,and employs an improved Transformer encoder to establish long-range dependencies between features.A Chinese handwriting gender recognition dataset(CHAP) containing 1 620 writers is constructed,and comprehensive experimental validation is conducted on both CHAP and Hebrew handwriting dataset(HHD).The results show that MSDAttFormer achieves an accuracy of 84.46%,F1-Score of 83.97%,and ROC AUC of 90.25% on the CHAP dataset,and an accuracy of 85.71% on the HHD dataset,significantly outperforming existing methods and validating the effectiveness and cross-language generalization capability of the proposed method.

Key words: Handwriting gender recognition, Transformer, Multi-scale feature fusion, Directional attention, Deep learning

中图分类号: 

  • TP391
[1] HUBER R A,HEADRICK A M.Handwriting identification[M].Boca Raton,FL,USA:CRC Press,1999.
[2] GOODENOUGH F L.Sex differences in judging the sex ofhandwriting[J].The Journal of Social Psychology,1945,22(1):61-68.
[3] HAMID S,LOEWENTHAL K M.Inferring gender from handwriting in Urdu and English[J].The Journal of Social Psychology,1996,136(6):778-782.
[4] UPADHYAY S,SINGH J,SHUKLA S K.Determination of sex through handwriting characteristics[J].International Journal of Current Research and Review,2017,9(13):11-18.
[5] TOPALOGLU M,EKMEKCI S.Gender detection andidentif-ying one’s handwriting with handwriting analysis[J].Expert Systems with Applications,2017,79:236-243.
[6] HE K,ZHANG X,REN S,et al.Deep residual learning for image recognition[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR).Las Vegas,NV,USA:IEEE,2016:770-778.
[7] BOUADJENEK N,NEMMOUR H,CHIBANI Y.Robust soft-biometrics prediction from off-line handwriting analysis[J].Applied Soft Computing,2016,46:980-990.
[8] BOUADJENEK N,NEMMOUR H,CHIBANI Y.Local descriptors to improve off-line handwriting-based gender prediction[C]//2014 6th International Conference of Soft Computing and Pattern Recognition(SoCPaR).Tunis,Tunisia:IEEE,2014:43-47.
[9] BOUADJENEK N,NEMMOUR H,CHIBANI Y.Fuzzy inte-gral for combining SVM-based handwritten soft-biometrics prediction[C]//2016 12th IAPR Workshop on Document Analysis Systems(DAS).Santorini,Greece:IEEE,2016:311-316.
[10] BOUADJENEK N,NEMMOUR H,CHIBANI Y.Age,genderand handedness prediction from handwriting using gradient features[C]//2015 13th International Conference on Document Analysis and Recognition(ICDAR).Tunis,Tunisia:IEEE,2015:1116-1120.
[11] GATTAL A,DJEDDI C,SIDDIQI I,et al.Gender classification from offline multi-script handwriting images using oriented Basic Image Features(oBIFs)[J].Expert Systems with Applications,2018,99:155-167.
[12] AHMED M,RASOOL A G,AFZAL H,et al.Improving handwriting based gender classification using ensemble classifiers[J].Expert Systems with Applications,2017,85:158-168.
[13] MAKEN P,GUPTA A.A method for automatic classification of gender based on text- independent handwriting[J].Multimedia Tools and Applications,2021,80(16):24573-24602.
[14] MORERA Á,SÁNCHEZ Á,VÉLEZ J F,et al.Gender and handedness prediction from offline handwriting using convolutional neural networks[J].Complexity,2018,2018(1):Article No.3891624.
[15] RAHMANIAN M,SHAYEGAN M A.Handwriting-based gender and handedness classification using convolutional neural networks[J].Multimedia Tools and Applications,2021,80(28-29):35341-35364.
[16] XUE G,LIU S,GONG D,et al.ATP-DenseNet:a hybrid deep learning-based gender identification of handwriting[J].Neural Computing and Applications,2021,33(10):4611-4622.
[17] DOSOVITSKIY A,BEYER L,KOLESNIKOV A,et al.Animage is worth 16x16 words:Transformers for image recognition at scale[C]//9th International Conference on Learning Representations(ICLR).Virtual Event:OpenReview.net,2021.
[18] KHAN S,NASEER M,HAYAT M,et al.Transformers in vision:A survey[J].ACM Computing Surveys,2022,54(10s):1-41.
[19] LI Y,LUO C,JIN L,et al.HTR-VT:Handwritten text recognition with vision transformer[J].Pattern Recognition,2025,158:110967.
[20] CASCIANELLI S,PIPPI V,MAARAND M,et al.The LAM dataset:A novel benchmark for line-level handwritten text recognition[C]//2022 26th International Conference on Pattern Recognition(ICPR).IEEE,2022:1506-1513.
[21] LI M,LV T,CHEN J,et al.TrOCR:Transformer-based optical character recognition with pre-trained models[C]//Proceedings of the 37th AAAI Conference on Artificial Intelligence(AAAI).Washington,DC,USA:AAAI Press,2023:13094-13102.
[22] BHUNIA A K,KHAN S,CHOLAKKAL H,et al.Handwriting transformers[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision(ICCV).2021:1086-1094.
[23] PARRES D,PAREDES R.Fine-tuning vision encoder-decodertransformers for handwriting text recognition on historical documents[C]//Document Analysis and Recognition-ICDAR 2023.Springer,2023:253-268.
[24] LIU Y,ZHANG Y,WANG Y,et al.A survey of visual transformers[J].IEEE Transactions on Neural Networks and Learning Systems,2023,35(7):8726-8746.
[25] YANG Y,TAM F,GRAHAM S J,et al.Men and women differ in the neural basis of handwriting[J].Human Brain Mapping,2020,41(10):2642-2655.
[26] VAN DER WEEL F R,VAN DER MEER A L H.Handwriting but not typewriting leads to widespread brain connectivity:a high-density EEG study with implications for the classroom[J].Frontiers in Psychology,2024,14:1219945.
[27] TAN M,LE Q.EfficientNet:rethinking model scaling forconvo-lutional neural networks[C]//International Conference on Machine Learning.2019:6105-6114.
[28] SIMONYAN K,ZISSERMAN A.Very deep convolutional networks for large-scale image recognition[C]//3rd International Conference on Learning Representations(ICLR).San Diego,CA,USA:OpenReview.net,2015.
[29] CHOLLET F.Xception:deep learning with depthwise separable convolutions[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.2017:1251-1258.
[30] SZEGEDY C,LIU W,JIA Y Q.Going deeper with convolutions[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.2015:1-9.
[31] BALAT M,MOHAMED Y,HEAKL A,et al.Arabic handwritten text for person biometric identification:a deep learning approach[J].arXiv:2406.00409,2024.
[32] HUANG G,LIU Z,VAN DER MAATEN L,et al.Densely con-nected convolutional networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.2017:4700-4708.
[33] RABAEV I,ALKORAN I,WATTAD O,et al.Automatic gender and age classification from offline handwriting with bilinear ResNet[J].Sensors,2022,22(24):Article No.9650.
[34] RABAEV I,KURAR BARAKAT B,CHURKIN A,et al.The HHD dataset[C]//2020 17th International Conference on Frontiers in Handwriting Recognition(ICFHR).Dortmund,Germany:IEEE,2020:228-233.
[35] RABAEV I,LITVAK M,ASULIN S,et al.Automatic gender classification from handwritten images:a case study[C]//Computer Analysis of Images and Patterns(CAIP 2021).Virtual Event:Springer,2021:329-343.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
渝ICP备16013121号-4
地址：重庆市两江新区洪湖西路18号    邮编：401121  
电话：023-63500828（编辑部） 023-67039612（会议/专题） 023-67039623（财务/发票）
E-mail：jsjkx12@163.com
本系统由北京玛格泰克科技发展有限公司设计开发