This compilation and all annotations are copyright © Jean Renard Ward, 1996, 1999, 2002, 2005, 2006
Copyright © Fri Sep 5 20:13:50 EDT 2008
Contact: jrward@alum.mit.edu

Mirrors of this page can be found at:

• http://users.erols.com/rwservices/pens/biblio86.html
• http://65.78.28.198:81/pens/biblio86.html
• http://rwservices.no-ip.info:81/pens/biblio86.html

      Main Reference Page      

• [Abe86] () Abe, K., Azumatani, Y., Mukouda, M., and Suzuki, S. "Discrimination of Symbols, Lines, and Characters in Flowchart Recognition", Proceedings of the 8th International Conference on Pattern Recognition, Paris, October 1986, pp 1071-1074.

  • Fundamental problem is separating/segmentation (flowchart) symbols drawn small size from characters
  • Problem of broken lines and curves messes up flowchart/character recognition
  • Sketch recognition

• [Ahmed86] () Ahmed, P. "Computer recognition of totally unconstrained handwritten ZIP codes", Ph.D. Thesis, Department of Computer Science, Concordia University, Montreal, Canada, 1986.

• [AEGTelefunken86] (*p) Richard Oed, Eberhard Mandler, Wolfgang Doster and Juergen Schuerman "Personal correspondence", September 1986.

  • Independent group working on prototype on-line recognition product

• [Amin86] () Amin, A. and Masini, G. "Machine Recognition of Multi-font Printed Arabic Texts", Proceedings of the 8th International Conference on Pattern Recognition, Paris, October 1986, pp 392-395.

  • Arabic: 95% segmentation, 85% recognition on 100 Arabic words
  • Commercial product by Doctor Mark Spikell

• [AnnoGraphics86] () AnnoGraphics "Datapad (TM) Portable Hand-print Character Recognition Data Entry System", AnnoGraphics, Incorporated, 10720 Kelley Driver, Fairfax, VA 22030, 1986, Mark Spikell, founder.

• [AnnoGraphics86a] (*p) Spikell, Mark "Personal correspondence", AnnoGraphics / DataPad, 1986

• [Aoki86] () Aoki, K. and Yamaya, Y. "Recognizer with Learning Mechanism for Hand-written Script English Words", Proceedings of the 8th International Conference on Pattern Recognition, Paris, October 1986, pp 690-692.

  • 85.4% with three writers (but says "real" performance would be much better)

• [Applicon86] . Applicon "Bravo3 User's Guide", Applicon Incorporated, Ann Arbor, Michigan 1986

  • cited in RubinSM84: Ledeen recognizer?

• [Averbuch86] (*p) Averbuch, A., et al "An IBM-PC Based Large-Vocabulary Isolated-Utterance Speech Recognizer", IBM Research Report RC 11663 (Log #53271) 1/28/86, IBM T.J. Watson Research Center, PO Box 218, Yorktown Heights, NY 10598.

  • Speech: a gigantic data-base of 27 million words in correspondence was still not enough for coverage
  • Enrollment: in 27 million words of correspondence, several days of week and months of year missing from 5000 most frequent and had to be added by hand
  • Other authors: Bahl, L., Bakis, R., Brown, P., Cole, A., Dagget, G.
  • Other authors: Das, S., Davies, K., DeGennaro, S., de Sourza, P.
  • Other authors: Epstein, E., Fraleigh, D., Jelinek, F., Katz, S., Lewis, B.
  • Other authors: Mercer, R., Nada, A., Nahamoo, D., Picheny, M., Shichman, G.
  • Other authors: Spinelli, P.

• [Babaguchi86] () Babaguchi, N., Tsukamoto, M., and Aibara, T. "Knowledge Aided Character Segmentation from Handwritten Document Image", Proceedings of the 8th International Conference on Pattern Recognition, Paris, October 1986, pp 573-575.

  • Complete character segmentation (in OCR) not possible only with general knowledge
  • Specific knowledge of characters will help segmentation (i.e. segmentation not general-purpose-method solvable as humans do it?)

• [Bahl86] (*p) Bahl, L.R. "Speech Recognition - An IBM Research Project", unpublished manuscript, December 1986. Author's address: Continuous Speech Recognition Group, IBM T.J. Watson Research Center, Yorktown Heights, NY 10598.

  • Speech: enrollment of 1000 sentences, testing on only 50 sentences

• [Baird86a] (*p) Baird, H.S. "Accurate Skew Estimation and the Top-Down Analysis of Document Images", unpublished manuscript, December 12, 1986, submitted to 1987 IEEE/CS 1st International Conference on Computer Vision, London. Author's address: AT&T Bell Laboratories, 2C-557, 600 Mountain Avenue, Murray Hill, NJ 07974.

  • Automatic rotation/scaling/skew-angle correction using "minimum energy" in OCR pre-processing
  • Published version

• [Baird86] () Baird, H.S., Kahan, S., and Pavlidis, Theo "Components of an Omnifont Page Reader", Proceedings of the 8th International Conference on Pattern Recognition, Paris, October 1986, pp 344-348.

  • Accidental merging/splitting of characters known to be a severe problem in optical character recognition
  • Production of multi-font OCR recognition an engineering technology problem, not science

• [Bao-chang86] () Bao-chang, P. "Floating Mask Method for Extracting Hand-printed Character Features", Proceedings of the 8th International Conference on Pattern Recognition, Paris, October 1986, pp 324-326.

  • Chinese OCR: 99.93% accuracy, 200,000 samples of 51 characters by 90 persons
  • Testing: uses very small subset of Chinese, thereby avoiding the problem of large character set

• [BarkerPG86] (*p) Barker, P.G., Najah, M., and Manji, K.A. "Pictorial communication with Computers", manuscript in preparation, October, 1986. Interactive Systems Research Group, Department of Computer Science, Teesside Polytechnic, County Cleveland, United Kingdom.

• [Bin86] () Bin, L. and Shuxiang, Z. "A New Approach to Recognition of Both Handwritten and Multi-font Printed Chinese Characters", Proceedings of the 8th International Conference on Pattern Recognition, Paris, October 1986, pp 641-643.

  • Model for Chinese education from psychologists (cognitive?), educators, linguists
  • The better a method fits a (cognitive) model, the better the recognition algorithm is

• [Blesser86a] (*p) Blesser, B. "Digitizing Tablet System Having Stylus Tilt Correction", United States Patent 4,577,057, assigned to Pencept, Incorporated, Waltham, Massachusetts, March 18, 1986.

  • Two-coil digitizer tablet stylus/pen for tilt measurement and/or error correction

• [Blesser86b] (*p) Blesser, B. "Digitizing Tablet System Including a Tablet Having a Grid Structure Made of Two Orthogonal Sets of Parallel Uniformly Sized and Spaced U Shaped Loops of Conductive Material", United States Patent 4,582,955, assigned to Pencept, Incorporated, Waltham, Massachusetts, April 15, 1986.

  • Digitizer tablet geometry to minimize stray signal pick-up

• [Blesser86] () Blesser, B., and Ward, J. "Human Factors Affecting the Problem of Machine Recognition of Hand-printed Text", Vol III Proceedings of the 7th Annual Computer Graphics Conference, National Computer Graphics Association, May 11-15 1986, pp 498-514.

  • Paper shows all the variations of "G"s: actual presentation did "A"s
  • We give many examples of combinatorial explosion of handwritten character variability from base forms

• [Brault86] (*p) Brault, J.J. and Plamondon, R. "Coupling Visual and Dynamic Features to Study Handwritten Signatures", Proceedings of Graphics Interface/Vision Interface 86, Vancouver, B.C. 26-30 May 1986.

  • Shows defective tablet line, but thinks it's just digitizer noise

• [Bono86] (*p) Bono, Peter R. "Support for Character Recognition Devices by the Graphics Standards", presentation to PHIGS graphics-standards committee, May 23, 1986

  • Presentation on behalf of several on-line pen computing companies to the ANSI graphics standards committees

• [Borning86] (*) Borning, Alan "Graphically Defining nNew Building Blocks in ThingLab", Human-Computer Interaction, 1986, Vol 2 No 4 pp 269-295 (abstract only)

  • Animation editor in ThingLab, with object constraints

• [Burr86] () Burr, D.J. "A Neural Network Digit Recognizer", Proceedings of the 1986 International conference on Systems, Man, and Cybernetics, Atlanta, Georgia, 1986, pp 1621-1625.

  • Martin,GL90

• [Buxton86] (*) Buxton, W. "Chunking and Phrasing and the Design of Human-Computer Dialogues", Proceedings of IFIP World Computer Congress, Dublin Ireland, September 1-5, 1986.

  • Gesture input / sketch recognition in graphical user interfaces

• [Buxton86a] (*) Buxton, William, and Myers, Brad A. "A Study in Two-Handed Input", to appear in 1986 Conference on Computers and Human Interaction, CHI'86, Boston April 12-17, 1986

  • digitizer tablet plus other controls, for selection of text

• [Buxton86b] (*p) Buxton, William "There's More to Interaction than Meets the Eye", in User Centered System Design: New Perspectives on Human-Computer Interaction, Norman, Donald A. and Draper, Stephen W, Eds, Lawrence Erlbaum Associates, Hillsdale NJ and London, 1986

• [Carpenter86] (*p) Carpenter, G.A. and Grossberg, S. "A Massively Parallel Architecture for a Self-organizing Neural Pattern Recognition Machine", manuscript in preparation, Authors' address: Center for Adaptive Systems, Department of Mathematics, Boston University, Boston, Massachusetts 02215.

  • Adaptive: give example of "simple" 5x5 characters, but patterns
  • Do not match human patterns (random bits)

• [Casey86] () Casey, R.G. "Text OCR by Solving a Cryptogram", Proceedings of the 8th International Conference on Pattern Recognition, Paris, October 1986, pp 349-351.

  • Adaptive recognition using clear-text and "unknown" type-font text for a cryptogram: but is clear-text always known?

• [Casio86_IF8000] (*) Casio "Casio IF-8000 Calculator", Collection from Pocket Computing Museum, http://cdecas.free.fr/computers/pocket/if8000.htm, 2002

  • 1984 Calculator (this reference says 1986) with touchscreen input, and also zone-based handwriting recognition on a touch-film keyboard/keypad
  • Same reference includes Apple Newton Message Pad 110

• [Chunbiao86] () Chunbiao, G. and Guorong, X. "Automatic Recognition of Printed Chinese Characters by Four Corner Codes", Proceedings of the 8th International Conference on Pattern Recognition, Paris, October 1986, pp 1013-1015.

  • 6763 Kanji/Chinese characters in minimal Chinese standard (3000 common): much more than Japanese standard (1493)
  • Chain codes/segments: Chinese/Kanji characters in only nine kinds of primitive segments

• [CohenM86] () Cohen, M. and Grossberg, S. "Neural dynamics of speech and language coding: developmental programs, perceptual grouping, and competition for short-term memory", Human Neurobiology, 1986, Vol 5, pp 1-22.

  • According to Daniel Bullock's letter from BU in 1986, Cohen did handwriting recognition in 1970.
  • Theoretical adaptive work: "random growth" of human brain

• [Cooper86] () Cooper, Leon N., Elbaum, Charles, Reilly, Douglas L., and Scofield, Christopher L. "Pattern class separation and identification system", United States Patent 4,216,524, assigned to Nestor, Incorporated, Providence, Rhode Island, 1986.

• [Cordella86] () Cordella, L.P. and Marcelli, A. "Normalization and Decomposition of Thin Lines Representing Handprinted Characters", Proceedings of the 8th International Conference on Pattern Recognition, Paris, October 1986, pp 723-725.

  • Decomposing characters into segments (curves, arcs) that are invariant with respect to handwriting variability (compare to chain-code segments)
  • Structural analysis beats the pants off of statistical / adaptive methods
  • Cites Suen/Shillman paper on U-V discrimination
  • Recognition by grammar/syntax of (chain-code) segments

• [Crane86] (*) Crane, H.D., and Ostrem, J.S. "Confusion Grouping of Strokes in Pattern Recognition Method and System", United States Patent 4,573,196, February 25, 1986, assigned to Communication Intelligence Corporation, Menlo Park, California.

  • Confusion matrix for Chinese/Kanji character recognition, using distance and curvature metrics

• [Currit86] () Currit, P.A., Dyer, M., and Mills, H.D. "Certifying the Reliability of Software", IEEE Transactions on Software Engineering, Vol SE-12 No 1, January 1986, pp 3-14.

  • The accepted approach to software development is to specify and design a product -- and then to test selectively with cases perceived to be typical of those requirements. Frequently the result is a product which works well against inputs to be tested but which is unreliable in unexpected circumstances
  • Testability of software: software-engineering regressions tests
  • Statistical testing of software can predict reliability: however, it cannot tell you about the existence, severity, or number of remaining errors
  • Sonic tablet digitizer: 0.1mm accuracy over 1.0 meter square active area (but then says +-2.0mm after all calibration!)
  • Sonic tablet digitizer: local drafts throw off same as 1 degree C temperature
  • Sonic tablet digitizer: calibrate to correct for air speed vs temperature

• [deBruyne86] (*) de Bruyne, P. "Compact Large-Area Graphic Digitizer for Personal Computers", IEEE Computer Graphics and Applications, Vol 6 No 12, December, 1986, pp 49-53.

  • Sonic digitizer tablet: ultrasonic shock wave: wireless/cordless cursor/stylus, using spark gap sonic pen, accuracy/error is 0.1% of distance measured

• [Derouault86] () Derouault, A. and Merialdo, B. "Natural Language Modeling for Phoneme-to-Text Transcription", IEEE Transactions on Pattern Analysis and Machine Intelligence", Vol 8 No 6, November 1986, pp 742-749.

  • Statistical modelling and grammatical rules combined for context in French natural language

• [DosterW86a] () Doster, W. "Designing a Document Analysis System", tutorial at 8th International Conference on Pattern Recognition, Paris, France, 27-31 October, 1986.

  • Cited in Baird86a

• [DosterW86] () Doster, W. "private letter of May, 1986", AEG Aktiengesellschaft, Forschungsinstitut Ulm, Sedanstrasse 10, D-7900 Ulm, West Germany.

  • Doster also has a DCR (instead of OCR) character data base

• [Firdman86] () Firdman, H.E. "Components of AI Systems", AI Expert, Vol 1 No 1, 1986, pp 81-85.

  • General comments on what an AI system "is"
  • DCR as Artificial intelligence: "domain specific: rules of thumb extracted from experts"

• [Flurry86] (*) Flurry, Gregory A. "Real-time rub-out erase for an electronic handwriting facility", United States Patent 4,633,436 assigned to International Business Machines Corp, Armonk, New York, December 30, 1986.

  • Stylus patent for tablet digitizer: Looks very similar to erase function of Wang Freestyle: real-time rubout erase on back end of stylus.

• [FoleyJD86] () Foley, J. et al "Managing the Design of User-Computer Interfaces", Report available from Computer Graphics Consultants, Incorporated, 616 G Street SW, Washington DC 20024, 1986.

  • Jim Foley's consulting report on user interface subsystems

• [Fu86] () Fu, K.S. "A Step Towards Unification of Syntactic and Statistical Pattern Recognition", IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol 8 No 3, May 1986, pp 398-404. Previously published in Vol 5 No 2, March, 1983.

  • Could also be listed as Fu83: was re-printed in same journal
  • Attributed grammar, control diagram, semantics, statistical pattern recognition, syntactic pattern recognition, syntax-semantics tradeoff
  • Sub-patterns, substructure (primitives) of pattern as features
  • There is yet no systematic method to select appropriate syntactic and semantic complexities (features) for a specific pattern recognition problem
  • Gives examples of distorted hand-written "E"s, and a possible tree/heirarchical representation of the structure for recognition

• [Guanxiong86] () Guanxiong, Z. and Hongyuan, W. "Angle Coding and it's Application in Pattern Recognition", Proceedings of the 8th International Conference on Pattern Recognition, Paris, October 1986, pp 1001-1003.

  • Chain-codes: eight-direction angle codes (Freeman codes)

• [Hase86] () Hase, M., Suzuki, G., and Hisayasu, I. "A Method for Extracting Marked Regions from Document Images", Proceedings of the 8th International Conference on Pattern Recognition, Paris, October 1986, pp 780-782.

  • Separating graphics and text characters from an image

• [Hekmatpour86] () Hekmatpour, S. and Ince, D.C. "Forms as a Language Facility", A.C.M. SIGPLAN Notices, Vol 21 No 9, September 1986, pp 43-48.

  • Quotes Gehani82
  • Proposes forms as so important to deserve own data type in a language

• [Hidai86a] () Hidai, Y., Ooi, K., and Nakamura, Y. "Stroke Re-ordering Algorithm for On-line Hand-written Character Recognition", Proceedings of the 8th International Conference on Pattern Recognition, Paris, October 1986, pp 934-936.

  • BLRT/chain-code: segments at eight discrete slants

• [Hidai86b] () Hidai, Y., Ooi, K., Nakamura, Y., and Kurosawa, Y., "Hand-written characters segmenting system", European Patent 177319, April 9, 1986, assigned to Toshiba.

  • Hand-written characters segmenting system has tablet determining boundary between characters according to stroke duration (inter-stroke timeout) and coordinates data (height from tablet)

• [Holbaek-Hanssen86] () Holbaek-Hanseen, Erik, Braten, Knut, and Taxt, Torfinn "A General Software System for Supervised Statistical Classification of Symbols", Proceedings of ICPR International Conference on Pattern Recognition, Paris, October 27-31, 1986, pp 144-149.

  • Handwriting character recognition using grid lengths of chain code skeletons, boundary tracing, "auxiliary features"

• [Hongo86] () Hongo, Yasuo and Nitta, Yoshio "Pattern Recognition Apparatus", United States Patent 4,628,533, December 9, 1986, assigned to Fuji Electric Company, Limited, Kanagawa, Japan.

  • Character recognition by pixel matching on OCR

• [Hull86a] () Hull, Jonathan J. "Hypothesis generation in a computational model for visual word recognition", IEEE Expert, 1986, pp 63-70.

  • Cited in Sinha88: for context

• [Hull86] () Hull, Jonathan J. "The Use of Global Context in Text Recognition", Proceedings of the 8th International Conference on Pattern Recognition, Paris, October 1986, pp 1218-1220.

  • Context via "nearest neighbor" words from a dictionary
  • Cognitive: OCR recognize words by overall outline of vertical bars, etc., not actual individual letters
  • Cites Shillman on use of human cognitive psychology, but this time it's on word-level recognition

• [Hulls86] () Hulls, L.R. "Polyphase Digitizer", United States Patent 4,570,033, assigned to Numonics Corporation, Lansdale Pennsylvania, February 11, 1986.

  • Phase-shifting tablet digitizer, using three-terminal grid winding

• [Hutchins86] () Hutchins, E., Hollan, J. and Norman, D.A. "Direct Manipulation Interfaces", in User-Centered System Design, D.A. Norman and S. Draper, editors, Erlbaum Associates, Hillsdale, New Jersey, 1986, pp 87-124.

  • Cited in JohnsonJ89

• [ICSA86] () ICSA "CRS (Character Recognition System) software product description", Intelligent Computer Systems and Applications, Inc., Box 1555, Station H, Montreal Quebec H3G 2N5 Canada, 1986.

  • Ching Suen's commercial company to sell OCR character recognition software.

• [Inagaki86] (*) Inagaki, Naoki "Key-In Device", United States Patent 4,578,811, March 25, 1986

  • Zone-based / matrix based character recognition, using the keypad of the Casio 8000 calculator as the input table.
  • Compare: Unistroke/ single stroke character recogition recognition software.

• [Jandrell86] () Jandrell, L.H.M. "Data Input Device with a Circuit Responsive to Stylus Up/Down Position", United States Patent 4,575,580, assigned to Astec International, Limited, Kowloon, Hong Kong.

  • Resistive sheet tablet using a force/pressure-sensitive conductive sheet
  • Gives fix for wild-points at starts and ends of strokes from up/down contact resistance on resistive sheet tablets

• [Jelinek86a] () Jelinek, F. "Markov Source Modeling of Text Generation", unpublished manuscript, December 1986. Author's address: Continuous Speech Recognition Group, IBM T.J. Watson Research Center, Yorktown Heights, NY 10598.

  • Speech: probabilistic language models for vocabulary context
  • Cites need for billions of samples to get statistically reasonable vocabulary probabilities

• [Jelinek86] () Jelinek, F. "Self-Organized Language Modeling for Speech Recognition", unpublished manuscript, December 1986. Author's address: Continuous Speech Recognition Group, IBM T.J. Watson Research Center, Yorktown Heights, NY 10598.

  • Speech recognition: has large data-bases of business correspondence for vocabulary

• [Jorgensen86] () Jorgensen, C. and Matheus, C. "Catching Knowledge in Neural Nets", AI Expert, December 1986, pp 31-38.

  • General overview of neural networks, perceptrons, adelines, etc.
  • "credit assignment problem" in perceptrons and adaptive recognition
  • Neural nets making a comeback: future still not proven

• [Kable86] () Kable, R.G. "Electrographic Apparatus", United States Patent 4,600,807, assigned to Scriptel Corporation, Columbus, Ohio, July 15, 1986.

  • Scriptel patent on resistive-sheet, electrostatic/capacitive-coupling digitizer tablet. Mentions correction matrix for entire surface of a digitizer tablet

• [KankaanpaaA86] (*) Kankaanpaa, Arto "An Advanced display System with Natural Interactivity", Proc. of Eurographics 86,European Computer Graphics Conference and Exhibition, Lisbon, 25..29 August, 1986, pp 186..193

  • Text editing with single-stroke (mostly) gestures on integrated touchscreen digitizer and LCD display, suggests electroluminesent or plasma display.

• [Kao86] () Kao, H.S.R., Van Galen, G.P., and Hoosain, R. "Graphonomics: Contemporary Research in Handwriting", North-Holland Publishers, Amsterdam-New York-Oxford-Toronto, 1986, Number 37 of "Advances in Psychology", Stelmach, G.E. and Vroon, P.A., editors.

• [Kasturi86] () Kasturi, R., Shih, C., and Fletcher, L.A. "An Approach for Automatic Recognition of Graphics", Proceedings of the 8th International Conference on Pattern Recognition, Paris, October 1986, pp 877-879.

  • Separating graphics and text characters from an image

• [Kay86] (*) Kay, Alan "A Talk by Alan Kay: summary of video interview Alan Kay talks about Sketchpad, Grail, the Dynabook", http://www.newmediareader.com/cd_samples/Kay/; The History of the Personal Workstation, 27 May 1986

  • Grail system with freehand input of letterforms, boxes, lines, corrections to previous drawings, Sketchpad, Dynabook. Video includes Grail system showing handwriting recognition (with lightpen) and scratchout gesture. Screenshots from video in file.

• [Kondo86] () Kondo, S. and Attachoo, B. "Model of Handwriting Process and its Analysis", Proceedings of the 8th International Conference on Pattern Recognition, Paris, October 1986, pp 562-565.

  • Features: distortions of end points on strokes do not affect human classification of handwriting, of middles of strokes does affect recognition
  • Stroke structure (order) very important in handwriting process, and therefore should be important in character recognition
  • Big problem in handwriting recognition is finding/extracting features not subject to handwriting distortion
  • Since set of all characters is not clear cut, any definition based on "invariant" features of a class is likely to be contradictory
  • Simulate handwritten characters: problem is finding out what base forms are that you should vary on

• [Kurosawa86] () Kurosawa, Y. and Asada, H. "Attributed String Matching with Statistical Constraints for Character Recognition", Proceedings of the 8th International Conference on Pattern Recognition, Paris, October 1986, pp 1063-1067.

  • Chain code: use primitive segments and to string matching
  • Exhaustive depth-first search of syntactic string for primitive segments
  • Katakana and alphanumeric characters with BLRT-like primitive segments
  • Combinatorial explosion of production rules
  • Combination of primitive-segment syntactic matching and statistical constraints
  • BLRT-chain codes give multiple-matches (3 or so), so second, ad-hoc comparison analysis for each possible pair of characters

• [Lam86] () Lam, S.W. and Baird, H.S. "Performance Testing of Mixed-font, Variable-size Character Recognizers", Computing Science Technical Report No. 126, AT&T Bell Laboratories, Murray Hill, New Jersey, November 6, 1986.

  • Gives statistics for size of test set to justify recognition performance claims of a given percent with a justifiable confidence level
  • To claim 99.5% recognition accuracy with 95% confidence, need 1500 samples
  • To claim recognition accuracy improved from 99% to 99.5% with 95% confidence, need 5000 samples
  • Refers to clustering and partitioning algorithms for feature selection for Bayesian statistics
  • OCR static recognition system was tested without ligatures, without several "special" characters
  • 99.5%+ performance requires very large, carefully-designed tests
  • Test performance of 99.5% was with common cases omitted, and some confusions forgiven
  • For OCR, scanner digitizer performance is critical
  • OCR scanner/digitizer: has short term fluctuations in sensitivity (re-scanning? was there internal averaging?)
  • Testing/training done only with high-quality data, not low-quality
  • For OCR digitizer scanners, scanning performance is not uniform or repeatable page-to-page
  • OCR tested and trained only on data of good quality, but question was raised
  • Resolution: 24 pixels not enough data accuracy for OCR
  • OCR at 200 dots/inch resolution started to fail at 12 point type size
  • Refers to tools to collect automatically a very large (168,000) character test set data base

• [Laube86] () Laube, M. "Audiographic Terminal", ITT Electrical Communication, Vol 60 No 1, 1986, pp 45-50.

  • Says maximum frequency of handwriting is 15 Hz, therefore 40 points/second digitizing rate (don't believe it!)
  • Describes resistive-sheet tablet for handwriting
  • Described force/pressure-sensitive tablet (!!)
  • Handwriting/voice/display terminal, like Interspec

• [Leedham86] (*p) Leedham, C.G. and Downton, A.C. "On-line recognition of Pitman's handwritten shorthand -- an evaluation of potential", International Journal of Man-Machine Studies, Vol 24, 1986, pp 375-393.

  • 9 degree variation in angle of nominal vertical and horizontal strokes
  • Short and long strokes overlap in length
  • Textbook writing (shorthand) is NOT the most recognizable to humans
  • Human recognition: textbook shorthand is 93.5%, student is 97.5%, expert is 48%
  • For small writing, hooks and loops are as big as many of the strokes
  • Unconstrained shorthand input
  • Intended straight strokes consistently tend to be curved
  • Has gall to say "recognition would be better if writing were neater"
  • Ignored strokes if too close to preceeding stroke
  • Stroke length variability is 20% in length
  • Says strokes were "lost" (thrown out) if not classified as "legal"
  • Did not deal with overdrawn/intentionally retraced strokes
  • Says recognition would get better with user practice (human factors/user interface)
  • "no amount of post-processing (context?) will fix segmentation and stroke classification errors"
  • Test protocol intentionally slow to get better quality data
  • Test protocol defeated linguistic context, resulting in users having to concentrate more
  • Refers to need for better (non-simplified) recognition algorithms
  • Cites ergonomic problems of tablet styli
  • Cites "writer's unfamiliarity with the system and capabilities" as performance factor

• [Lettera86] () Lettera, C., Masera, L., Paoli, C. and Proinelli, R. "Use of a Dictionary in Conjunction with a Handwritten Text Recognizer", Proc. ICPR(86), pp. 699-701. 1986

  • Cited by Marlin Eller, Microsoft Pen Computing group

• [Litvin86] () Litvin, Y. "Private letter of August 4, 1986", Author's address: Skylight Software, 2 Charles Street, Bedford, Massachusetts 01730.

  • DCR not deterministic: there is no formal criteria for correctness (not true: he misunderstood "deterministic")

• [LuH86] (*) Lu, H.E. and Wang, P.S.P. "A Comment on 'A Fast Parallel Algorithm for Thinning Digitial Patterns'", Communications of the A.C.M., Vol 29 No 3, pp 239-242.

  • Skeletonization/line-thinning, special notes on endpoints NOT to be deleted: compare with smoothing/pre-processing algorithms for digitizer tablet data by Pencept

• [Ma86] () Ma, Y.L., Jang, S.Y., and Ma, C. "Pattern Recognition by Circular Layer Code Approach", Proceedings of the 8th International Conference on Pattern Recognition, Paris, October 1986, pp 783-785.

  • Chinese OCR, 97.5% on 4 characters tested with 20 samples each (!!!)

• [Maamari86] (*) Maamari, F. and Plamondon, R. "Extraction of the Analog Pentip Position, Velocity and Acceleration Signals from a Digitizer", Neuronal and Motor Aspects of Handwriting, North-Holland Publishers, J.S.R. Kau, Editor, 1985.

  • Analog filter on digitizer output gives "improved accuracy" (!!??!!)
  • Low-pass filtering on tablet digitizer lets you get "better" time resolution (?)

• [MacNelly86] (*) MacNelly, Jeff " "Shoe" comic: "I wish I sat behind someone with decent handwriting"", "Shoe", November 21, 1986. Taken from Boston Globe newspaper, Boston, USA, Page 102.

  • Student in a school class during a test:
  • Fill in the blanks: "To the victor belong the *snails*."
  • That can't be right ...
  • I wish I sat behind someone with decent handwriting.
  • (Humorous illustration of the problems of imperfect handwriting recognition, and of the difficulty of human recognition of handwriting)

• [MaedaY86] () Maeda, Y., Yoda, F., Matsuura, K., and Nambu, H. "Character Segmentation in Japanese Hand-written Document Images", Proceedings of the 8th International Conference on Pattern Recognition, Paris, October 1986, pp 768-772.

• [Maier86] () Maier, M. "Separating Characters in Scripted Documents", Proceedings of the 8th International Conference on Pattern Recognition, Paris, October 1986, pp 1056-1058.

  • AI artificial intelligence was seen as panacea
  • Cited in Wolf86

• [Makkuni87] () Makkuni, R "A gestural representation of the process of composing Chinese temples", IEEE Computer Graphics and Applications, Vol 7 No 12, December 1987, pp 45-61.

• [Mandler86] () Mandler, E. "conversation of September, 1986", AEG Aktiengesellschaft, Forschungsinstitut Ulm, Sedanstrasse 10, D-7900 Ulm, West Germany.

  • One-stroke "T" more common in Europe

• [Mangione86b] (*) Mangione, P.A. "SSI's Phonetic Engine (TM)", Speech Technology Magazine, Mar/April 1986.

  • Speech Systems phonetic feature speech recognition

• [Mangione86c] () Mangione, P.A. "Phonetic Recognition - The Basic Building Block", Speech Technology Magazine, April/May 1986.

  • Speech Systems phonetic feature speech recognition

• [Mangione86] () Mangione, P.A. "What about the user?", Proceedings of Speech Tech '86, April 28-30 1986, New York, New York, pp 154-156.

  • Marketing V.P. did paper on Speech Systems' phonetic speech input product

• [Mantas86a] () Mantas, J. "An overview of character recognition methodologies", Pattern Recognition, Vol 19, 1986, pp 425-430.

  • Cited in Baptista88

• [Mantas86] () Mantas, J. "A Fuzzy Decision Operator in Recognising Greek Hand-drawn Characters", Proceedings of the 8th International Conference on Pattern Recognition, Paris, October 1986, pp 753-755.

  • Pattern recognition is not just pattern classification, but also design of recognition systems
  • Statistical versus syntactic/linguistic pattern recognition (does not like statistical)
  • Statistical/Bayesian recognition cannot be done for handwriting because a priori probabilities cannot be calculated for every feature

• [Meisel86] (*p) Meisel, W.S. "Implications of Large Vocabulary Recognition", Proceedings of Speech Tech '86, April 28-30, 1986, New York, New York.

  • Speech Systems speech recognition using phonetics
  • (sloppy cursive): co-articulated speech, vs isolated, connected, continuous
  • Speech: storing multiple-word (multiple-char) segments as special "words"
  • (BLRTs/variability): phoneme detection reduces variability to a set of rules at phoneme level
  • Ambiguity: groups of words (chars) vs single words: "there are four" vs "therefore"
  • Dictionary look-up/search: speed and accuracy of recognition unrelated to
  • Size of dictionary, but only to complexity of syntax
  • Adaptive vs user-independent: controversy over "speaker-independence"
  • Adaptive: must begin at a usable level even so, otherwise bad human-factors (frustration)
  • Adaptive: correction for content, not for error (user wants a different word)
  • Adaptive: separate enrollment session a bad user-interface idea for "normal" speakers
  • Penalty for speaker-independence too high (frustration)
  • Variability: phoneme speech system can extrapolate/predict words after adaptive samples
  • Trade-off between accuracy and constraints on user in speech:
  • What if user uses bad English?

• [Myers86] (*) Myers, Brad A. and Buxton, William "Creating Highly-Interactive and Graphical User Interfaces by Demonstration", Proceedings of SIGGRAPH '86, pp 249 ff.

  • Direct-manipulation

• [Morasso86] (*p) Morasso, P. "Understanding Cursive Script as a Trajectory Formation Paradigm", in "Graphonomics: Contemporary Research in Handwriting", Elsevier Science Publishers B.V. (North-Holland), 1986, pp 137-167.

  • Hand gestures and handwriting: study of movements
  • Measures curvature profile, not just velocity profile, of handwriting motion
  • Claims no handwriting information at frequency above 10 Hz, shows peak at 5 Hz and D.C. (but his samples are mostly loop motion, not fast block printing)
  • Shows plots of low-pass frequency filtering on handwriting data at 10 Hz and below
  • Intends to collect lots of allograph handwriting samples to build a future recognizer for cursive script

• [Morita86] (*) Morita, Toshiaki; Horii, Masahiro; Tasaka, Shigeru; Hirose, Hitoshi "Handwriting Character Recognition Device", United States Patent 4,607,386 assigned to Sharp Kabushiki Kaisha, Osaka, Japan, August 19, 1986

  • Chinese handwritten character recognition without regard to stroke order

• [Murase86] (*) Murase, H., and Wakahara, T. "Online Hand-Sketched Figure Recognition", Pattern Recognition, Vol 19 No 2, pp 147-160, 1986.

  • Sketch context: must be valid connections to other flowchart symbols
  • 97.2% accuracy: test protocol was only 20 people, nine symbols, four times each
  • Mentions lots of user-interface work on on-line sketch recognition
  • For "standardized" sketch symbols, high variability (no regularity) in stroke order
  • Says stroke order variable, but end/connection points are always the same
  • Refers to stroke connection variability for "sketched" symbols

• [Nagy86b] () Nagy, G.R., Seth, S., Einspahr, K., and Meyer, T. "Efficient Algorithms to Decode Substitution Ciphers with Application to OCR", Proceedings of the 8th International Conference on Pattern Recognition, Paris, October 1986, pp 352-355.

  • Adaptive recognition using clear-text and "unknown" type-font text for a cryptogram: but is clear-text always known?

• [Nagy86] (*p) Nagy, G., Seth, S., and Stoddard, S.D. "Document Analysis with an Expert System", in "Pattern Recognition in Practice", Gelsema, E.S. and Kanal, L.N., editors, Elsevier Science Publishers B.V. (North-Holland), 1986.

  • Tiling/windows applied for format of written/printed documents

• [Nakagawa86] () Nakagawa, M., Aizawa, T., Komoda, C., Ideda, Y., and Takahashi, N. "Syntactic Pattern Recognition with Stochastic Dissimilarity in Japanese On-line Input Systems (JOLIS)-1/1.5", Proceedings of the 8th International Conference on Pattern Recognition, Paris, October 1986, pp 1059-1061.

  • Chain-code: Japanese recognition by classifying strokes into primitive strokes
  • Chain-code: eight stroke directions/segments
  • Chain-codes: extension to elastic matching

• [Nakagawa86a] (*p) Nakagawa, Masaki, Manabe, Toshihiko, Aoki, Katsuo, Ikeda, Yuji, and Takahashi, Nobumasa "On-line Handwritten Character Recognition as a Japanese Input Methods", Proceedings of ???, pp 191-196

  • on-line handwriting recognition system for Japanes, JOLIS-1: chain codes (8-directional), 29 primitive strokes types by dictionary look-up

• [Nestor86] (*) Nestor, Inc. "Nestor Press clipping file", Nestor, Inc.

• [OedR86a] () Oed, R. "telephone conversation of July 17, 1986", AEG Aktiengesellschaft, Forschungsinstitut Ulm, Sedanstrasse 10, D-7900 Ulm, West Germany.

  • You get different writing forms after the first 3-4 lines out of 20

• [OedR86b] () Oed, R. "conversation of September, 1986", AEG Aktiengesellschaft, Forschungsinstitut Ulm, Sedanstrasse 10, D-7900 Ulm, West Germany.

  • Gesture/symbols are modified characters: slash-M, "DF"

• [Ogozalek86] () Ogozalek, V.Z. and Van Praag, J. "Comparison of elderly and younger users on keyboard and voice input computer-based composition tasks", in Human Factors in Computing Systems, Proceedings of CHI 83 Conference, 1983, Boston.

  • Cited in Martin89
  • Martin89 cites as saying speech recognition and typing just as fast a user-interface for typing documents

• [OlsonB86] (*) Olson, Bruce T. "Is your Handwriting Legible Enough for the Computer?", Journel of Law enforcment Report writing, Vol 3, No 1, December 1986

  • Pencept PenPad for capturing police reports using handwriting recognition

• [Orita86] () Orita, M., Kanasaki, M., Toda, Y., Mishima, T., Suzuki, M., Onuma, C., and Takatoo, M. "Image extraction using density distribution characteristics", European Patent 198481, October 22, 1986, assigned to Hitachi, Limited, Tokyo, Japan.

  • OCR scanner preprocessing to extract character images from background clutter

• [Pavlidis86a] () Pavlidis, Theo "A vectorizer and feature extractor for document recognition", Computer Vision, Graphics, Image Processing, Vol 35, 1986.

  • Cited in Kahan87

• [Pavlidis86] () Pavlidis, Theo "A Critical Survey of Image Analysis Methods", Proceedings of the 8th International Conference on Pattern Recognition, Paris, October 1986, pp 502-511.

  • Image processing / recognition design requires many methods, not just one theoretical basis
  • A digitizer tablet good enough for writing input

• [PCMagazine86] (*) PC Magazine "Best of 1985 - Pencept Penpad", PC Magazine, January 14, 1986, Ziff-Davis Publishing Company

  • Review of best PC products of 1985 -- describes Pencept PenPad with PenCad software, use of "pen macros" / letter gesture commands

• [Pencept86] () Pencept "Penpad Penpad 300 (TM) Digitizing Tablet Product Information", Pencept, Incorporated, 39 Green Street, Waltham, Massachusetts 02154, 1986.

• [Pencept86a] (*p) Pencept "PENWARE (tm) Pendraw (tm) II User's Guide", Pencept, Incorporated, 39 Green Street, Waltham, Massachusetts 02154, 1986.

• [Pencept86b] (*p) Pencept "Penpad 300 User's Guide", Pencept, Incorporated, 39 Green Street, Waltham, Massachusetts 02154, 1986.

• [Pencept86c] (*p) Pencept "Penpad 320 User's Guide", Pencept, Incorporated, 39 Green Street, Waltham, Massachusetts 02154, 1986.

  • shows single-stroke and multiple-stroke characters shapes in one set. Includes system disk with mouse driver to adapt tablet input (including command templates, character recognition) to mouse/keyboard input.

• [Pencept86d] (*p) Pencept "Writing Samples", Pencept, Incorporated, 39 Green Street, Waltham, Massachusetts 02154, 1986.

  • Collections of unconstrained, natural writing samples taken from hotel receipts, university bulletin board notes, bills of sale, and cross-word puzzles from the magazines in waiting rooms and airlines from various North American and European cities, showing true variability effects

• [Pencept86e] (*p) Pencept "Software Toolkit for the Penpad 310 and Penpad 320", Pencept, Incorporated, 39 Green Street, Waltham, Massachusetts 02154, 1986.

• [Peterson86] () Peterson, J.L. "A note on undetected typing errors", Communications of the A.C.M., Vol 29, 1986, pp 633-637.

  • Cited in Sinha88: on spelling context correction

• [Pickering86] () Pickering, J. "Touch-sensitive screens: the technologies and their applications", International Journal of Man-Machine Studies, Vol 25, 1986, pp 249-269.

  • Mack89 on tablet technology:

• [Plamondon86a] (*p) Plamondon, R. and Baron, R. "On-line Recognition of Handprint Schematic Pseudocode for Automatic Fortran Code Generation", Proceedings of the 8th International Conference on Pattern Recognition, October 27-31, 1986, Paris, pp 741-744.

  • Graphical program editing user-interface using DCR

• [Plamondon86] (*p) Plamondon, R. and Baron, R. "A dedicated microcomputer for handwritten interaction with a software tool: system prototyping", Journal of Microcomputer Applications, Vol 9, 1986, pp 51-60.

  • On-line character recognition for graphical software tool
  • User-interface goal: must interface to application without changing application
  • Too much feedback distracts the user from the writing task

• [Postl86] () Postl, W. "Detection of Linear Oblique Structures and Skew Scan in Digitizer Documents", Proceedings of 8th International Conference on Pattern Recognition, Paris, France October 1986, pp 687-689.

  • Cited in Biard86a
  • What a nice thing our recognition macros are for a user-interface

• [RadioElectronics86] (*p) Radio Electronics "Equipment Report - Pencept Penpad 320", Radio Electronics, January 1987, pp 22 ff

• [Rhyne86] () Rhyne, J.R. and Wolf, C.G. "Gestural Interfaces for Information Processing Applications", IBM Research Report RC 12179 (Log #54544) 9/2/86, 1986, IBM T.J. Watson Research Center, PO Box 218, Yorktown Heights, NY 10598.

  • Description of IBM's early informal study on what editing marks subjects use on spreadsheets
  • Report on IBM's prototype gesture/handwriting Lotus interface
  • Early description of IBM's electronic ink hardware tablet and display
  • Refers to "gesture language" for the qualitatively similar set of symbol marks and operations subjects use to edit spreadsheets
  • Refers to gesture language, combining gesture (pointing) and menu selection, gesture (command) and handwriting (data)
  • Accuracy of recognition has great effect on user interface (need to confirm, not just correct)
  • Recognition errors/accuracy: user assumes his or her writing was correct, so errors and means to correct must be very obvious
  • For a small gesture symbol set, intra-user variability was found to be small
  • For a small gesture symbol set, inter-user variability was usually three or so "qualitatively similar" (ambiguous/confusable) symbols
  • IBM had no answer on reconciling small display character size with larger handwritten character size for electronic ink
  • User interface for electronic ink: enlarge area of interest, let user write handwritten characters normal size on enlarged image
  • Optical/visual parallax, also ranging parallax (pen tilt)
  • Pen tilt error of 0.06" tablet digitizer accuracy led to pointing errors and slow user interaction rate effects on user interface
  • Handwriting substantially worse on slick glass surface than on paper for electronic ink, much lower recognition accuracy
  • User interface problems for handwriting are big: no obvious ideas on how to solve

• [Rothfjell86] () Rothfjell, Rolf E. "Method and Device for Signature Verification", United States Patent 4,581,482, April 8, 1986, assigned to Esselte Security Systems AB, Stockholm, Sweden.

  • Signature verification using a constrained writing form to get alignment points for segmentation (!)

• [Sabourin86] () Sabourin, R. and Plamondon, R. "Preprocessing of Handwritten Signatures from Image Gradient Analysis", Proceedings of the 8th International Conference on Pattern Recognition, October 27-31, 1986, Paris, pp 576-579.

  • Static (OCR) signature verification
  • Signature verification: test protocol included no forgery attempts

• [Scheifler86] () Scheifler, Robert W. and Gettys, Jim "The X Window System", Internal Report, Laboratory for Computer Science, Massachusetts Institute of Technology, Cambridge, Massachusetts, October 1986.

  • Description of the X-Window graphical user interface system

• [Schomaker86] () Schomaker, L.R.B. and Thomassen, A.J.W.M. "On the Use and Limitation of Averaging Handwriting Signals", in in "Graphonomics: Contemporary Research in Handwriting", Elsevier Science Publishers B.V. (North-Holland), 1986, pp 225-238.

  • Low-pass average filter on data, but with anchored endpoints where velocity is zero
  • Describes anchoring inflection (zero velocity) points on handwritten characters as similar to dynamic time warping for speech recognition
  • Time-normalization of cursive handwriting
  • Force/pressure sensitive pen used in handwriting research
  • Capacitive coupling, resistive-sheet digitizer tablet

• [Scriptel86] () Scriptel "SPC-Series Technical Report", Scriptel Corporation, 4145 Arlingate Plaza, Columbus, OH 43228, May, 1986.

• [Searby86] () Searby, A.D. and Bowman, D.W. "Stylus Devices Responsive to Pressure Changes for Use in Videographic and Like Apparatus", United States Patent 4,580,007, April 1, 1986, assigned to Quantel, Limited, Surrey, England.

  • Force/pressure-sensitive digitizer stylus with improved life and performance

• [Shridhar86] () Shridhar, Am. and Badreldin, A. "Recognition of Isolated and Simply Connected Handwritten Numerals", Pattern Recognition, Vol 19 No 1, pp 1-12, 1986.

  • Refers to stroke connection within and between characters

• [Signify86] () Signify Inc. "Sign-On product information", 9005 Red Branch Road, Columbia Maryland 21045, 1986 (see also ATI).

  • Signature verification product: now called ASI/Autosig Systems Inc..

• [Srinivasan86] () Srinivasan, S., Palaniswamy, K., and Natarajan, A.E. "Machine Recognition of the Indian Language Characters Using a Tree-structure Based On Primitives", Proceedings of the 8th International Conference on Pattern Recognition, Paris, October 1986, pp 726-728.

• [Suen86] () Suen, C.Y. "Human Recognition of Handprinted Characters and Distance Measurements", in "Graphonomics: Contemporary Research in Handwriting", Elsevier Science Publishers B.V. (North-Holland), 1986, pp 213-224.

  • Needed 29x39 resolution to get decent handwriting recognition
  • Simple (not serious) recognizer algorithm using pixel template matching, then entropy, information content, and nearest neighbor distances: did not correlate to human recognition
  • Human use multiple distance measurements (multiple features) to recognize characters
  • Says entropy, hamming distance, linear and cross correlation, centre of gravity, Mahalonobis distance of template matching were all poor distance functions

• [Suydam86] () Suydam, W.E. "Approaches to Software Testing Embroiled in Debate", Computer Design, Vol 25 No 21, November 15, 1986, pp 49-55.

  • Long, informal review article on software testing and reliability
  • David Parnas: Only actual use will bring out the mistakes (not testing)
  • Parnas: need for regression testing
  • Parnas: problem of classical testing is unweighted likelihood of failure
  • Parnas: testing must consider seriousness of different types of failure
  • Parnas: testing must differentiate between important errors and insignificant errors
  • Parnas: (Titanic effect) software design implements a mental model of real world: must test the model
  • AI testing: people think it's different from software testing: it is not
  • AI testing: when mechanical engineering was a new as AI, had not invented the right angle yet
  • Dijkstra: testing shows presence of bugs, not absence
  • Testing: must distinguish between active failure (real bug) vs passive (design omission)
  • Testing: only field testing finds "real" errors
  • Testing: unanticipated data can seriously screw a program

• [Tagushi86] () Tagushi, Y., and Yamanami, T. "Position Detecting Apparatus", United States Patent 4,617,515, October 14, 1986, assigned to Wacom Company, Limited, Japan.

  • Cordless stylus digitizer tablet with magnets under the grid of wire loops

• [TanakaT86] () Tanaka, Toshinori, and Kobayashi, Shunsuke "Entry of Data and Command for an LCD by Direct Touch: An Integrated LCD Panel", SID 86 Digest of Technical Papers, Society for Information Display, 1986, pp 318-320.

  • Touch-entry-device integrated tablet/display using crystal deformation from touch force/pressure in LCD pixel cells: pixels degrade, but refreshing fixes them

• [TanzawaS86] (*) Tanzawa, Setsu "Recording pen", Japanese Patent Application Number S60-88861

  • Light-sensing optical stylus digitizer for handwriting input, using X and Y lines in different colors. Compare with Anoto.

• [Tappert86] (*) Tappert, C.C., Fox, A.S., Kim, J., Levy, S.E., and Zimmerman, L.L. "Handwriting recognition on transparent tablet over Flat Display", Society for Information Display Digest of Technical Papers, Vol XVII, pp 308-312, May 1986.

  • Cited in Rhyne86: Scriptel Digitizer
  • Rhyne86 cites this on visual parallax for electronic ink: 0.06" lateral shift causes subject's discomfort. Tappert describes also slippery surface. Cites internal Pencept report by Phillips and Prentice.

• [Teulings86a] () Teulings, H., Thomassen, A.J.W.M., van Galen, G.P. "Invariants in Handwriting: The Information Contained in a Motor Program", in "Graphonomics: Contemporary Research in Handwriting", Elsevier Science Publishers B.V. (North-Holland), 1986, pp 305-315.

  • Spatial characteristics are more invariant in handwriting than time or force characteristics (? for signature verification?)
  • But because handwriting is "efficient", time characteristics in handwriting are invariant on trained sequences, like signature verification

• [Teulings86] () Teulings, H., Mullins, P.A., and Stelmach, G.E. "The Elementary Units of Programming Handwriting", in "Graphonomics: Contemporary Research in Handwriting", Elsevier Science Publishers B.V. (North-Holland), 1986, pp 21-32.

  • Handwriting data low-pass filtered from 16 Hz to 48 Hz frequency
  • There is no one, single grapheme unit of handwriting: units depend on what the subject is writing

• [Tognazzini86] () Tognazzini, Bruce "You Had To Be There", report on User Interface Reference Model workshop at SIGCHI '86 conference, SIGCHI Bulletin, Vol 18 No 2, October 1986, pp 21-24.

  • Uses example of "graphical ambiguity/context" in recognition when discussing user-interface issues

• [Treisman86] () Treisman, A. "Feature and Objects in Visual Processing", Scientific American, November 1986, pp 114B-125.

  • Visual recognition consists of low-level features (automatic), then identification
  • Low-level perception is lines, spots, edges, locations, orientation
  • Properties/features conjoined only at higher level of human recognition
  • (context): "distractors" delay human recognition

• [VanGalen86] () Van Galen, G.P., Meulenbroek, R.G.J., Hylkema, H. "On the Simultaneous Processing of Words, Letters and Strokes in Handwriting: Evidence for a Mixed Linear and Parallel Model", in "Graphonomics: Contemporary Research in Handwriting", Elsevier Science Publishers B.V. (North-Holland), 1986, pp 5-20.

  • Parsed characters in script handwriting into character segments at velocity zeros (sharp corners)
  • Gives maximum velocity of handwriting in cursive script as 11 cm/sec
  • Alternated turning direction vs continuous loops has big impact on handwriting velocity (of course, silly!)
  • Beginnings of words written at slower speed

• [Wakahara86] () Wakahara, Toru "personal correspondence", July 28, 1986.

• [Ward86] (*) Ward, J. "Method and Apparatus for Removing Noise at the Ends of a Stroke Caused by Retracing", United States Patent 4,608,658, August 26, 1986, assigned to Pencept, Incorporated, Waltham, Massachusetts.

  • Re-trace elimination to pre-process handwritten character shapes before recognition

• [Watanabe86] () Watanabe, Y. "letter to J.Ward", Author's address: Terebijon Gakkaishi, Hachinohe Institute of Technology, Department of Engineering, Hachinohe, Japan, 25 November, 1986.

  • Mentions co-work between Watanabe and Jiro Gyoba
  • Interchange of looping and cusping is seen frequently in writing Japanese
  • Kanji characters written to OCR standard are LESS human-recognizable than "sloppy" writing
  • At 300-500 msec/stroke, writing time for Kanji Japanese slower than Roman
  • Constraints: Japanese have characters they can read, but do not know how to write
  • Japanese writers do not conform to formal writing constraints in practice

• [Websters86] (*) Merriam-Webster Inc "Moon Type", Webster's Third New International Dictionary of the English Language Unabridged, Copyright 1986, ISBN 0-97779-201-1

  • Show Moon Type, a simplified single-stroke / Unistroke alphabet used as embossed printing for the blind, popular before Braille became dominant

• [Wolf86] (*p) Wolf, C.G. "Can People Use Gesture Commands?", SIGCHI Bulletin, Vol 18 No 2, pp 73-74.

  • Gesture input - user interface. IBM simple usability study (18 subjects) for "move" gesture consisting of an arrow to indicate what gets moved to where, with different gestures (lasso, bracket, underline) to indicate selection.

• [WuL86] () Wu, L. and Weng, F. "Chain Code for a Line Segment and Formal Language", Proceedings of the 8th International Conference on Pattern Recognition, Paris, October 1986, pp 1124-1127.

  • Some properties of chain codes (straight line theorem) very hard to test

• [YamamotoK86a] () Yamamoto, K., Yamada, H., Saito, T. and Sakaga, I. "Recognition of Handprinted Characters in the First Level of JIS Chinese Characters", Proceedings of the 8th International Conference on Pattern Recognition, Paris, October 1986, pp 570-572.

  • Relaxation method with 98.6% success on Kanji/Chinese characters
  • Testing with ETL-8 and ETL-9 "standard" Chinese writing samples in Japan
  • ETL-9 standard Kanji Chinese samples is 600000 samples of 3036 characters

• [YamamotoK86b] (*p) Yamamoto, K. and Saito, T. "Pattern Reading System", United States Patent 4,566,124, January 21, 1986, assigned to Agency of Industrial Science and Technology, Ministry of International Trade and Industry, Tokyo, Japan.

  • Handwriting character recognition by tracing contours, local extrema/outermost points, chain codes
  • Template matching on direction chain codes, polygonal approximation to pattern for OCR handwriting recognition

• [Yau86] () Yau, S. S. and Tsai, J.J.-P. "A Survey of Software Design Techniques", IEEE Transactions on Software Engineering, Vol SE-12 No 6, June 1986, pp 713-721.

  • General survey article on software engineering

      Main Reference Page