{"corpus_id":8203975,"paper_sha":"36158edac846d8fbf0dd04a2289055d55b33e5de","doi":"10.1109/TNNLS.2019.2956965","arxiv_id":"1610.04057","pmid":31905151,"pmcid":null,"mag_id":2998192574,"dblp_id":"journals/tnn/LiuHCWY20","acl_id":null,"title":"Stroke Sequence-Dependent Deep Convolutional Neural Network for Online Handwritten Chinese Character Recognition","year":2016,"publication_date":"2016-10-13","venue":"IEEE Transactions on Neural Networks and Learning Systems","journal":{"name":"IEEE Transactions on Neural Networks and Learning Systems","pages":"4637-4648","volume":"31"},"journal_issn":null,"journal_title":null,"publication_types":["JournalArticle"],"pubmed_pub_types":["Journal Article","Research Support, Non-U.S. Gov't"],"s2_fields_of_study":["Medicine","Computer Science"],"reference_count":53,"citation_count":33,"influential_citation_count":1,"is_open_access":true,"arxiv_categories":["cs.CV"],"arxiv_license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","arxiv_journal_ref":null,"mesh_headings":null,"chemicals":null,"comments_corrections":null,"source_flags":5,"s2_open_access_pdf_url":"https://arxiv.org/pdf/1610.04057","s2_open_access_landing_url":"https://www.semanticscholar.org/paper/36158edac846d8fbf0dd04a2289055d55b33e5de","s2_open_access_license":null,"s2_open_access_status":"GREEN","pmc_open_access_pdf_url":null,"pmc_open_access_landing_url":null,"pmc_open_access_license":null,"pmc_open_access_status":null,"unpaywall_open_access_pdf_url":null,"unpaywall_open_access_landing_url":null,"unpaywall_open_access_license":null,"unpaywall_open_access_status":null,"abstract":"We propose a novel model, called stroke sequence-dependent deep convolutional neural network (SSDCNN), which uses the stroke sequence information and eight-directional features of Chinese characters for online handwritten Chinese character recognition (OLHCCR). SSDCNN learns the representation of OLHCCs by incorporating the natural sequence information of the strokes. Furthermore, it naturally incorporates the eight-directional features. First, SSDCNN inputs the stroke sequence and transforms it into stacks of feature maps following the writing order of the strokes. Second, the fixed-length, stroke sequence-dependent representations of OLHCC are derived through convolutional, residual, and max-pooling operations. Third, the stroke sequence-dependent representation is combined with the eight-directional features via a number of fully connected neural network layers. Finally, the Chinese characters are recognized using a softmax classifier. The SSDCNN is trained in two stages: 1) the whole architecture is pretrained using the training data until the performance converges to an acceptable degree. 2) The stroke sequence-dependent representation is combined with the eight-directional features by a fully connected neural network and a softmax layer for further training. The model was experimentally evaluated on the OLHCCR competition tasks of International Conference on Document Analysis and Recognition (ICDAR) 2013. The recognition error was a maximum 58.28% lower in SSDCNN than in a model using the eight-directional features alone (5.13% versus 2.14%). Owing to its high accuracy (97.86%), the proposed SSDCNN reduced the recognition error by approximately 18.0% as compared with that of the winning system in the ICDAR 2013 competition. SSDCNN integrated with an adaptation mechanism, called the SSDCNN+Adapt model, and reached a new state-of-the-art (SOTA) standard with an accuracy of 97.94%. The SSDCNN exploits the stroke sequence information to learn high-quality OLHCC representations. Moreover, the learned representation and the classical eight-directional features complement each other within the SSDCNN architecture.","claims":[{"public_id":"cl_1236985f806558a010eb551c9cf31da7","status":"active","text":"Combining the stroke sequence-dependent representation with eight-directional features yields better recognition than using eight-directional features alone, reducing recognition error by up to 58.28% (5.13% versus 2.14%).","confidence":0.98,"contributors":[{"id":1,"public_id":"12632b8b5f","public_label":"Anonymous (12632b8b5f)","roles":["extraction"],"url":"https://sah.borca.ai/u/12632b8b5f"}],"url":"https://sah.borca.ai/claims/cl_1236985f806558a010eb551c9cf31da7"},{"public_id":"cl_8cc935f87ccacf8c19cbace2d0add314","status":"active","text":"Stroke sequence information can be incorporated into a deep convolutional architecture to learn high-quality online handwritten Chinese character representations.","confidence":0.95,"contributors":[{"id":1,"public_id":"12632b8b5f","public_label":"Anonymous (12632b8b5f)","roles":["extraction"],"url":"https://sah.borca.ai/u/12632b8b5f"}],"url":"https://sah.borca.ai/claims/cl_8cc935f87ccacf8c19cbace2d0add314"},{"public_id":"cl_abf729b577f8dfb3aef3d050151e5afe","status":"active","text":"The adapted SSDCNN+Adapt variant reached 97.94% accuracy and established a new state-of-the-art result.","confidence":0.96,"contributors":[{"id":1,"public_id":"12632b8b5f","public_label":"Anonymous (12632b8b5f)","roles":["extraction"],"url":"https://sah.borca.ai/u/12632b8b5f"}],"url":"https://sah.borca.ai/claims/cl_abf729b577f8dfb3aef3d050151e5afe"},{"public_id":"cl_7f90687acfa369481486c73b6a55df57","status":"active","text":"The model achieved 97.86% accuracy on the ICDAR 2013 OLHCCR competition tasks and reduced recognition error by approximately 18.0% relative to the winning system in that competition.","confidence":0.97,"contributors":[{"id":1,"public_id":"12632b8b5f","public_label":"Anonymous (12632b8b5f)","roles":["extraction"],"url":"https://sah.borca.ai/u/12632b8b5f"}],"url":"https://sah.borca.ai/claims/cl_7f90687acfa369481486c73b6a55df57"}],"concepts":[{"public_id":"co_17c18255e7236d62f232e352e7d4849e","status":"active","name":"convolutional, residual, and max-pooling operations","description":"Neural network operations used to transform input feature maps into fixed-length representations.","types":["neural network operation"],"aliases":[],"contributors":[{"id":1,"public_id":"12632b8b5f","public_label":"Anonymous (12632b8b5f)","roles":["extraction"],"url":"https://sah.borca.ai/u/12632b8b5f"}],"url":"https://sah.borca.ai/concepts/co_17c18255e7236d62f232e352e7d4849e"},{"public_id":"co_1d8421ce8875d1366f9d711ecc75f8e7","status":"active","name":"ICDAR 2013","description":"The International Conference on Document Analysis and Recognition 2013 benchmark or competition setting used for evaluation.","types":["benchmark","conference"],"aliases":[],"contributors":[{"id":1,"public_id":"12632b8b5f","public_label":"Anonymous (12632b8b5f)","roles":["extraction"],"url":"https://sah.borca.ai/u/12632b8b5f"}],"url":"https://sah.borca.ai/concepts/co_1d8421ce8875d1366f9d711ecc75f8e7"},{"public_id":"co_1fa19595232c82fc061626fcafb259d8","status":"active","name":"recognition error","description":"The proportion of characters misrecognized by the system.","types":["metric"],"aliases":[],"contributors":[{"id":1,"public_id":"12632b8b5f","public_label":"Anonymous (12632b8b5f)","roles":["extraction"],"url":"https://sah.borca.ai/u/12632b8b5f"}],"url":"https://sah.borca.ai/concepts/co_1fa19595232c82fc061626fcafb259d8"},{"public_id":"co_4919a27039217b4441f3dfd793c09755","status":"active","name":"SSDCNN+Adapt model","description":"An SSDCNN variant augmented with an adaptation mechanism for improved recognition performance.","types":["model","adapted model"],"aliases":[],"contributors":[{"id":1,"public_id":"12632b8b5f","public_label":"Anonymous (12632b8b5f)","roles":["extraction"],"url":"https://sah.borca.ai/u/12632b8b5f"}],"url":"https://sah.borca.ai/concepts/co_4919a27039217b4441f3dfd793c09755"},{"public_id":"co_618b158767a5567f64953f097ba55630","status":"active","name":"state-of-the-art","description":"The best reported performance level among comparable methods on the evaluated task.","types":["performance status"],"aliases":["SOTA"],"contributors":[{"id":1,"public_id":"12632b8b5f","public_label":"Anonymous (12632b8b5f)","roles":["extraction"],"url":"https://sah.borca.ai/u/12632b8b5f"}],"url":"https://sah.borca.ai/concepts/co_618b158767a5567f64953f097ba55630"},{"public_id":"co_662cfcbd689e79a7aeccfc41e39c8436","status":"active","name":"stroke sequence-dependent deep convolutional neural network","description":"A deep convolutional neural network that encodes online handwritten Chinese characters using stroke order information and directional features.","types":["method","neural network"],"aliases":["SSDCNN"],"contributors":[{"id":1,"public_id":"12632b8b5f","public_label":"Anonymous (12632b8b5f)","roles":["extraction"],"url":"https://sah.borca.ai/u/12632b8b5f"}],"url":"https://sah.borca.ai/concepts/co_662cfcbd689e79a7aeccfc41e39c8436"},{"public_id":"co_740dcf27f5290e5850d20f3295ab6527","status":"active","name":"stroke sequence information","description":"The ordered sequence of strokes produced while writing a Chinese character.","types":["input feature","sequence information"],"aliases":[],"contributors":[{"id":1,"public_id":"12632b8b5f","public_label":"Anonymous (12632b8b5f)","roles":["extraction"],"url":"https://sah.borca.ai/u/12632b8b5f"}],"url":"https://sah.borca.ai/concepts/co_740dcf27f5290e5850d20f3295ab6527"},{"public_id":"co_7aa8059db189b4972b5379601b5ef7ee","status":"active","name":"softmax classifier","description":"A classifier that converts network outputs into class probabilities for character recognition.","types":["classifier"],"aliases":["softmax layer"],"contributors":[{"id":1,"public_id":"12632b8b5f","public_label":"Anonymous (12632b8b5f)","roles":["extraction"],"url":"https://sah.borca.ai/u/12632b8b5f"}],"url":"https://sah.borca.ai/concepts/co_7aa8059db189b4972b5379601b5ef7ee"},{"public_id":"co_8e05508644a1f84fe972c177db7ed2d2","status":"active","name":"OLHCCR competition tasks","description":"The online handwritten Chinese character recognition tasks from the ICDAR 2013 competition.","types":["evaluation task"],"aliases":["ICDAR 2013 OLHCCR competition tasks"],"contributors":[{"id":1,"public_id":"12632b8b5f","public_label":"Anonymous (12632b8b5f)","roles":["extraction"],"url":"https://sah.borca.ai/u/12632b8b5f"}],"url":"https://sah.borca.ai/concepts/co_8e05508644a1f84fe972c177db7ed2d2"},{"public_id":"co_ad817bfd063fe639aa4f30ce1c898195","status":"active","name":"fully connected neural network layers","description":"Dense neural network layers used to combine learned representations with additional features.","types":["neural network layer"],"aliases":["fully connected layers"],"contributors":[{"id":1,"public_id":"12632b8b5f","public_label":"Anonymous (12632b8b5f)","roles":["extraction"],"url":"https://sah.borca.ai/u/12632b8b5f"}],"url":"https://sah.borca.ai/concepts/co_ad817bfd063fe639aa4f30ce1c898195"},{"public_id":"co_cf0df74162a1988963362f12b5279f9a","status":"active","name":"eight-directional features","description":"Directional features representing stroke or contour information in eight orientations.","types":["feature"],"aliases":[],"contributors":[{"id":1,"public_id":"12632b8b5f","public_label":"Anonymous (12632b8b5f)","roles":["extraction"],"url":"https://sah.borca.ai/u/12632b8b5f"}],"url":"https://sah.borca.ai/concepts/co_cf0df74162a1988963362f12b5279f9a"},{"public_id":"co_dfa36b2ac45326866581b8fe9f8ccd98","status":"active","name":"online handwritten Chinese character recognition","description":"The task of recognizing Chinese characters from online handwriting data.","types":["task"],"aliases":["OLHCCR"],"contributors":[{"id":1,"public_id":"12632b8b5f","public_label":"Anonymous (12632b8b5f)","roles":["extraction"],"url":"https://sah.borca.ai/u/12632b8b5f"}],"url":"https://sah.borca.ai/concepts/co_dfa36b2ac45326866581b8fe9f8ccd98"},{"public_id":"co_f2400063d92e40e2c0f76e95ff37a243","status":"active","name":"recognition accuracy","description":"The proportion of characters correctly recognized by the system.","types":["metric"],"aliases":[],"contributors":[{"id":1,"public_id":"12632b8b5f","public_label":"Anonymous (12632b8b5f)","roles":["extraction"],"url":"https://sah.borca.ai/u/12632b8b5f"}],"url":"https://sah.borca.ai/concepts/co_f2400063d92e40e2c0f76e95ff37a243"}],"external_ids":{"DOI":"10.1109/TNNLS.2019.2956965","ArXiv":"1610.04057","PubMed":31905151,"PubMedCentral":null,"MAG":2998192574,"DBLP":"journals/tnn/LiuHCWY20","ACL":null},"open_access":{"is_open_access":true,"pdf_url":"https://arxiv.org/pdf/1610.04057","landing_url":"https://www.semanticscholar.org/paper/36158edac846d8fbf0dd04a2289055d55b33e5de","source":"semantic_scholar","pdf_url_source":"semantic_scholar_open_access_pdf","license":null,"status":"GREEN","reason":null},"reference_availability":{"status":"available","references_indexed":true,"full_text_available":true,"full_text_source":"arxiv","count_basis":"semantic_scholar_metadata","extraction_status":"not_applicable","reason":null},"source":{"provider":"episteme2","base_corpus":"semantic_scholar_dump","freshness_mode":"unknown","basis":["semantic_scholar_metadata","postgres_metadata"],"limits":["paper metadata is based on indexed upstream scholarly datasets","claims and concepts are available only for extracted papers","absence of claims or concepts means no extracted graph data is available in this response"],"status":"available","degraded":false,"degraded_reasons":[],"diagnostics":{"status":"available","degraded":false,"degraded_reasons":[],"metadata_status":"available","graph_status":"available","abstract_status":"available"},"source_flags":5},"paper_id":631774,"paper_uid":"bb3c3713-a7f1-47da-8d3f-b7f9cc64b516","canonical_identity":{"paper_id":631774,"paper_uid":"bb3c3713-a7f1-47da-8d3f-b7f9cc64b516","identity_status":"available","lookup_basis":"semantic_scholar_external_id","compatibility_path":"corpus_id"},"url":"https://sah.borca.ai/papers/8203975"}