[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-187991-en":3,"doc-seo-187991-105":30,"detail-sidebar-cat-1-en-105":92},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":11,"category_id":12,"category_name":13,"doc_title":14,"doc_description":15,"doc_content":16,"file_id":17,"file_url":18,"file_type":19,"file_size":20,"view_count":4,"is_deleted":4,"is_public":11,"is_downloadable":11,"audit_status":11,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":15,"update_tm":28,"read_time":29},187991,1099514067415,"Rowan","https://ap-avatar.wpscdn.com/avatar/100002539d78ffe74a7?x-image-process=image/resize,m_fixed,w_180,h_180&k=1779092875211072502",1,158,"General","Connell - Techniques for Online Handwritten Character Recognition","Online handwritten character recognition compares multiple modeling strategies, including primitive decomposition, motor models, elastic matching, stochastic models, neural networks, and structural models. The content evaluates advantages and limitations such as robustness to writing-style variation, dependency on training data, classification speed, and ability to model temporal relations. Reported experiments provide accuracy figures across digits and mixed character sets, alongside throughput and rejection behavior, using reference templates and varied training setups.","| Technique | Advantages | Disadvantages |\n| --- | --- | --- |\n| Primitive decomposition [3,4] | Powerful high-level features | Not very robust to large variations in writing style |\n| Motor models [5}7] | Takes advantage of pen dynamics | May lack robustness when writing style variations are large |\n| Elastic matching [8}10] | Works very well for writer-dependent data writerdependent data Does not require a relatively large amount of training data | Does not generalize well for writerindependent tasks Classi\"cation\u003Cbr>time grows linearly with the number of training examples |\n| Stochastic models [2,13] Neural networks [14}16] | Models temporal relations well\u003Cbr>Classi\"cation time is fast | Requires a large amount of training data\u003Cbr>Does not model temporal relations very well |\n\n\n| Author | Method | Accuracy | Notes |\n| --- | --- | --- | --- |\n| Li and Yeung [8] | Nearest neighbor using elastic matching | Upper and lowercase classi\"er: 87.1% on 780 lowercase, 92.9% on 780 uppercase Digit classi\"er: 96.3% on 300 digits | 0.35 s/char. 180 reference templates selected from 840 examples |\n| Scattolin and Krzyzak [9] | Nearest neighbor using weighted elastic matching | 88.67% on 1650 digits | 33 writers used 293 reference templates |\n| Yaeger et al. [16] | Combined online and o%ine neural network classi\"er | 95-class classi\"er (52 chars, 10 digits, 33 symbols): 86.1% | 45 writers used |\n| Chan and Yeung [17] | Manually designed structural models | 97.4% on 9300 upper and lowercase chars | 150 writers used |\n| Li et al. [19] | Hidden Markov models | 92%(after 3.9% reject) on 3126 digits | Unipen data used |\n| Prevost and Milgram [20] | Combined online and o%ine nearest-neighbor classi\"ers | digits: 98.70% uppercase: 97.81% lowercase: 96.84% | Total of 11,162 chars from Unipen data used |\n\n| (0, 0, 223), ( !2, !1, 223), ( !1, 2, 223), (0, 5, 182), (1, 9, 194), (3, 13, 193), (5, 16, 190), (7, 19, 201), (10, 22, 239), (14, 24, 289), (16, 21, 261),(16, 17, 196),(16, 13, 180),(16, 9, 188),(16, 5, 189),(15, 1, 182),(15, !2, 187),(14, !5, 178),(13, !9, 172),(13, ! 13, 187),(12, ! 17, 193),(11, !21, 189), (9, !25, 183), (8, !28, 186), (6, !32, 196), (4, !35, 190), (1, !38, 185), ( !1, !41, 207), ( !4, !43, 211),( !8, !44, 218), ( ! 12, !45, 254), ( !15, !42, 255), ( ! 16, !38, 240), ( !15, !34, 237), ( !12, !31, 232), ( !9, !30, 220),( !5, !30, 211), ( !1, !31, 196), (2, !33, 172), (6, !34, 172), (9, !35, 179), (13, !36, 172), (17, !37, 159), (21, !37, 155),(25, !36, 155),(27, !35, 155)\u003Cbr>(a) |\n| --- |\n| (0, 0, 192), (2, 2, 192), (6, 3, 192), (10, 4, 187), (14, 5, 201), (18, 5, 216), (22, 4, 232), (25, 1, 251), (25, !1, 232), (24, !5, 200), (22, !9, 199),(20, !12, 195),(17, !15, 191),(14, ! 18, 186),(11, !20, 180),(8, !23, 186),(5, !25, 264),(2, !27, 343),(3, !23, 286),(7, !21, 207),(10, !19, 195), (14, !18, 203), (18, ! 17, 208), (22, !18, 203), (26, !19, 205), (29, !21, 211), (32, !24, 201), (34, !28, 192),(36, !31, 202), (37, !35, 212), (37, !39, 203), (36, !43, 196), (35, !47, 195), (33, !50, 186), (31, !54, 195), (29, !57, 205),(26, !59, 196), (22, !62, 200), (19, !63, 202), (15, !64, 202), (11, !64, 201), (7, !63, 196), (3, !62, 199), (0, !60, 213),( !3, !58, 253),( !4, !54, 259),( !1, !51, 221),(1, !49, 199),(5, !47, 199),(5, !47, 199)\u003Cbr>(b) |\n| (0, 0, 61), ( !3, ! 1, 61), (0, 0, 61), (3, 3, 193), (7, 4, 186), (11, 6, 185), (14, 7, 189), (18, 8, 188),(22, 9, 209),(26, 9, 243),(29, 7, 255), (30, 3, 230),(29, 0, 204), (28, !4, 204), (26, !7, 196), (23, ! 10, 180), (21, ! 13, 191), (18, ! 16, 195), (15, ! 18, 184), (11, !21, 189), (8, !23, 260),(5, !24, 330), (6, !21, 276), (9, ! 18, 207), (12, !16, 199), (16, !14, 209), (20, ! 13, 215), (24, ! 14, 213), (27, ! 15, 210),(30, !18, 197), (33, !21, 201), (35, !24, 205), (36, !28, 196), (37, !32, 203), (37, !36, 203), (36, !40, 196), (35, !44, 195),(33, !47, 192), (31, !51, 201), (28, !54, 201), (25, !56, 196), (22, !58, 199), (18, !59, 196), (14, !60, 201), (10, !60, 201),(6, !59, 188),","cbCaibccS1VAQulj","https://ap.wps.com/l/cbCaibccS1VAQulj","pdf",480924,14,"English","en",105,"# Techniques Overview\n## Primitive decomposition\n## Motor models\n## Elastic matching\n## Stochastic models and neural networks\n# Comparative Results\n## Author-method accuracy table\n## Digit class and template selection\n## Training set throughput and recognition rate","[{\"question\":\"Which recognition techniques are discussed and what core advantages do they claim?\",\"answer\":\"The document covers primitive decomposition, motor models, elastic matching, stochastic models, neural networks, and manually designed structural models. It highlights high-level feature power, pen-dynamics exploitation, strong writer-dependent performance, fast classification, and good temporal modeling depending on the method.\"},{\"question\":\"What main disadvantages or limitations are reported for these techniques?\",\"answer\":\"Limitations include weak robustness under large writing-style variation, reduced generalization for writer-independent tasks, linear growth of classification time with training example count, and the need for large training datasets for neural approaches.\"},{\"question\":\"How do training setup and template selection affect recognition rate and throughput?\",\"answer\":\"The results compare full training sets versus manually selected, auto-selected, and edited templates, showing changes in recognition accuracy and processing speed. The edited setup reports high throughput and the best recognition rate among the listed training configurations.\"}]","Connell - Techniques for Online Handwritten Character Recognition | PDF",1788386657,5,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":14,"keywords":34,"description":15,"schema_data":35,"social_meta":87,"head_meta":89,"extra_data":91,"updated_unix":28},"connell-techniques-for-online-handwritten-character-recognition","",{"@graph":36,"@context":86},[37,54,69],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":11},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/template/","Template",2,{"item":49,"name":13,"@type":43,"position":50},"https://docshare.wps.com/template/general/",3,{"item":52,"name":14,"@type":43,"position":53},"https://docshare.wps.com/template/connell-techniques-for-online-handwritten-character-recognition/187991/",4,{"url":52,"name":14,"@type":55,"author":56,"headline":14,"publisher":58,"fileFormat":61,"inLanguage":23,"description":15,"dateModified":62,"datePublished":63,"encodingFormat":61,"isAccessibleForFree":64,"interactionStatistic":65},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-09-03","2026-09-02",true,{"@type":66,"interactionType":67,"userInteractionCount":11},"InteractionCounter",{"@type":68},"ViewAction",{"@type":70,"mainEntity":71},"FAQPage",[72,78,82],{"name":73,"@type":74,"acceptedAnswer":75},"Which recognition techniques are discussed and what core advantages do they claim?","Question",{"text":76,"@type":77},"The document covers primitive decomposition, motor models, elastic matching, stochastic models, neural networks, and manually designed structural models. It highlights high-level feature power, pen-dynamics exploitation, strong writer-dependent performance, fast classification, and good temporal modeling depending on the method.","Answer",{"name":79,"@type":74,"acceptedAnswer":80},"What main disadvantages or limitations are reported for these techniques?",{"text":81,"@type":77},"Limitations include weak robustness under large writing-style variation, reduced generalization for writer-independent tasks, linear growth of classification time with training example count, and the need for large training datasets for neural approaches.",{"name":83,"@type":74,"acceptedAnswer":84},"How do training setup and template selection affect recognition rate and throughput?",{"text":85,"@type":77},"The results compare full training sets versus manually selected, auto-selected, and edited templates, showing changes in recognition accuracy and processing speed. The edited setup reports high throughput and the best recognition rate among the listed training configurations.","https://schema.org",{"og:url":52,"og:type":88,"og:title":14,"og:site_name":59,"og:description":15},"article",{"robots":90,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":93},[94,99,104,108,113,118,123,128,132],{"id":95,"doc_module":11,"doc_module_name":46,"category_name":96,"show_sort_weight":97,"slug":98},11,"Presentations",90,"presentations",{"id":100,"doc_module":11,"doc_module_name":46,"category_name":101,"show_sort_weight":102,"slug":103},12,"Resumes",80,"resumes",{"id":21,"doc_module":11,"doc_module_name":46,"category_name":105,"show_sort_weight":106,"slug":107},"Invoices",70,"invoices",{"id":109,"doc_module":11,"doc_module_name":46,"category_name":110,"show_sort_weight":111,"slug":112},15,"Posters",60,"posters",{"id":114,"doc_module":11,"doc_module_name":46,"category_name":115,"show_sort_weight":116,"slug":117},16,"Social Media",50,"social-media",{"id":119,"doc_module":11,"doc_module_name":46,"category_name":120,"show_sort_weight":121,"slug":122},17,"Forms",40,"forms",{"id":124,"doc_module":11,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},18,"Letters",30,"letters",{"id":129,"doc_module":11,"doc_module_name":46,"category_name":130,"show_sort_weight":29,"slug":131},21,"Paper Templates","papers-templates",{"id":12,"doc_module":11,"doc_module_name":46,"category_name":13,"show_sort_weight":4,"slug":133},"general-158"]