{"pk":31690,"title":"Recognizing Handprinted Digit Strings: a Hybrid Connectionist/Procedural Approach","subtitle":null,"abstract":"We describe an alternative approach to hand-\nprinted word recognition using a hybrid of proce-\ndural and connectionist techniques. We utilize two\nconnectionist components: one to concurrently make recognition and segmentation hypotheses, and another to perform refined recognition of segmented\ncharacters. Both networks are governed by a pro-\ncedural controller which incorporates systematic do-\nmain knowledge and procedural algorithms to guide\nrecognition.\nW e employ an approach wherein an image is pro-\ncessed over time by a spatiotemporal connectionist\nnetwork. The scheme offers several attractive fea-\ntures including shift-invariance and retention of lo-\ncal spatial relationships along the dimension being\ntemporalized, a reduction in the number of free pa-\nrameters, and the ability to process arbitrarily long\nimages.\nRecognition results on a set of real-world isolated\nZIP code digits are comparable to the best reported\nto date, with a 9 6 . 0 % recognition rate and a rate of\n9 9 . 0 % w h e n 9.5% of the images are rejected.","language":"eng","license":{"name":"","short_name":"","text":null,"url":""},"keywords":[],"section":"Submitted Presentations","is_remote":true,"remote_url":"https://escholarship.org/uc/item/2zn3s20d","frozenauthors":[{"first_name":"Thomas","middle_name":"","last_name":"Fontaine","name_suffix":"","institution":"University of Pennsylvania","department":""},{"first_name":"Lokendra","middle_name":"","last_name":"Shastri","name_suffix":"","institution":"University of Pennsylvania","department":""}],"date_submitted":null,"date_accepted":null,"date_published":"1993-01-01T13:00:00-05:00","render_galley":null,"galleys":[{"label":"PDF","type":"pdf","path":"https://journalpub.escholarship.org/cognitivesciencesociety/article/31690/galley/22758/download/"}]}