{"title":"书写者验证任务中字形复杂度的相关性","authors":"A. Bensefia, Chawki Djeddi","doi":"10.1109/IRI49571.2020.00016","DOIUrl":null,"url":null,"abstract":"Recognizing and identifying people, based on their physical and behavioral characteristics, have always had a wide range of applications, inciting researchers to propose dedicated human recognition systems for each human characteristic. These systems operate according to two different modes: identification mode, where the task is to assign one of the preregistered identities in the system to the human’s sample read as input. The second mode is the verification (authentication), is a decision task stating if a human’s sample read as input belongs really to the claimed identity. Handwriting has emerged as one of these behavioral features that attracted a lot of interests during the last decade. Many writer identification systems have been developed comparing to writer verification (authentication) systems. In this paper we propose an original approach based on the usage of the shape complexity to authenticate writers’ identities. To this end, a local feature (grapheme) is considered, where the graphemes are generated automatically with a dedicated segmentation module. The Fourier Elliptic Transform was used to measure the shape complexity of the resulting graphemes. Only the top complex graphemes (K-Graphemes) were used to measure the similarity between a pair of handwritten samples. The approach was evaluated with 3 sets of 50 different writers of the BFL dataset, where we obtained a performance of almost 80% of good acceptance at 8% error rate. These results validate completely the relevance of the shape complexity in writer recognition tasks.","PeriodicalId":93159,"journal":{"name":"2020 IEEE 21st International Conference on Information Reuse and Integration for Data Science : IRI 2020 : proceedings : virtual conference, 11-13 August 2020. IEEE International Conference on Information Reuse and Integration (21st : 2...","volume":"36 1","pages":"53-58"},"PeriodicalIF":0.0000,"publicationDate":"2020-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"2","resultStr":"{\"title\":\"Relevance of Grapheme’s Shape Complexity in Writer Verification Task\",\"authors\":\"A. Bensefia, Chawki Djeddi\",\"doi\":\"10.1109/IRI49571.2020.00016\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Recognizing and identifying people, based on their physical and behavioral characteristics, have always had a wide range of applications, inciting researchers to propose dedicated human recognition systems for each human characteristic. These systems operate according to two different modes: identification mode, where the task is to assign one of the preregistered identities in the system to the human’s sample read as input. The second mode is the verification (authentication), is a decision task stating if a human’s sample read as input belongs really to the claimed identity. Handwriting has emerged as one of these behavioral features that attracted a lot of interests during the last decade. Many writer identification systems have been developed comparing to writer verification (authentication) systems. In this paper we propose an original approach based on the usage of the shape complexity to authenticate writers’ identities. To this end, a local feature (grapheme) is considered, where the graphemes are generated automatically with a dedicated segmentation module. The Fourier Elliptic Transform was used to measure the shape complexity of the resulting graphemes. Only the top complex graphemes (K-Graphemes) were used to measure the similarity between a pair of handwritten samples. The approach was evaluated with 3 sets of 50 different writers of the BFL dataset, where we obtained a performance of almost 80% of good acceptance at 8% error rate. These results validate completely the relevance of the shape complexity in writer recognition tasks.\",\"PeriodicalId\":93159,\"journal\":{\"name\":\"2020 IEEE 21st International Conference on Information Reuse and Integration for Data Science : IRI 2020 : proceedings : virtual conference, 11-13 August 2020. IEEE International Conference on Information Reuse and Integration (21st : 2...\",\"volume\":\"36 1\",\"pages\":\"53-58\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2020-08-01\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"2\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2020 IEEE 21st International Conference on Information Reuse and Integration for Data Science : IRI 2020 : proceedings : virtual conference, 11-13 August 2020. IEEE International Conference on Information Reuse and Integration (21st : 2...\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/IRI49571.2020.00016\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2020 IEEE 21st International Conference on Information Reuse and Integration for Data Science : IRI 2020 : proceedings : virtual conference, 11-13 August 2020. IEEE International Conference on Information Reuse and Integration (21st : 2...","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/IRI49571.2020.00016","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Relevance of Grapheme’s Shape Complexity in Writer Verification Task
Recognizing and identifying people, based on their physical and behavioral characteristics, have always had a wide range of applications, inciting researchers to propose dedicated human recognition systems for each human characteristic. These systems operate according to two different modes: identification mode, where the task is to assign one of the preregistered identities in the system to the human’s sample read as input. The second mode is the verification (authentication), is a decision task stating if a human’s sample read as input belongs really to the claimed identity. Handwriting has emerged as one of these behavioral features that attracted a lot of interests during the last decade. Many writer identification systems have been developed comparing to writer verification (authentication) systems. In this paper we propose an original approach based on the usage of the shape complexity to authenticate writers’ identities. To this end, a local feature (grapheme) is considered, where the graphemes are generated automatically with a dedicated segmentation module. The Fourier Elliptic Transform was used to measure the shape complexity of the resulting graphemes. Only the top complex graphemes (K-Graphemes) were used to measure the similarity between a pair of handwritten samples. The approach was evaluated with 3 sets of 50 different writers of the BFL dataset, where we obtained a performance of almost 80% of good acceptance at 8% error rate. These results validate completely the relevance of the shape complexity in writer recognition tasks.