Portal field news

Portal field news

in ,

📱 | 2 screen folding "5G" smartphone "Surface Duo 2" Microsoft


2 screen folding "5G" smartphone "Surface Duo 2" Microsoft

If you write the contents roughly
It is compatible with "Surface Slim Pen 2" and allows handwriting input with "Microsoft 365" and Android apps.

Microsoft Japan is a smartphone "Surface Duo ..." compatible with the 5th generation mobile communication system "5G". → Continue reading

 J-CAST News

Wikipedia related words

If there is no explanation, there is no corresponding item on Wikipedia.

Microsoft 365

Handwriting recognition

Handwriting recognition(British: Handwriting recognition) Is a computer function that receives recognizable handwritten input.As an input sourcepaperHandwritten document,写真,Touch panelThere are devices such as.It is divided into "offline handwriting recognition", in which a document written in advance "offline" is optically scanned and input, and "online handwriting recognition", in which handwriting is input "online".

Handwriting recognition is basicallyOptical character recognitionIt is built on the technology of.However, in the case of online handwritten character recognition, it is also necessary to consider the stroke order and the direction in which the line is drawn, and to search for the most suitable character or word.

Offline handwriting recognition

Offline handwriting recognition optical scans a document image written on paper and automatically processes the image with text data (something).Character codeConvert to data consisting of).The data obtained in this way can be regarded as a static representation of handwritten characters.Handwritten characters are habitual and diverse from person to person, so they are normal.Optical character recognitionMore difficult than (OCR). OCR targets printed documentsIntelligent character recognition(ICR) targets handwritten documents that are not so-called cursive.There is currently no OCR / ICR that can recognize any handwritten character.

This technology has been successfully used in businesses that process many handwritten documents, such as insurance companies.The quality of recognition can be significantly improved by structuring the document.

Limitation of target area

By limiting the target area, the recognition rate of the handwritten character recognition system is often improved.For exampleZIP code(Zip Code), The types of characters are limited to numbers from 0 to 9.Therefore, the possibility of specifying a character can be considerably limited.

There are mainly the following techniques.

  • Limit the types of characters used.
  • Adopt a special format so that a specific character type is written in a specific position.

Character extraction

In offline character recognition, documents written in the past may be scanned and input.Here, it is necessary to separate and extract individual characters from the scanned image.There are tools for this,[1], It is hard to say that the problem of this process has been completely solved.A common problem is when two letters are written together and therefore extracted as a single letter.This is a major problem that makes the next character recognition process difficult.Various algorithms have been devised to solve this.

Character recognition

After extraction, an image of each character is input to the recognition engine to identify what the corresponding character code is.There are several recognition techniques.

neural network

neural networkThe recognition engine by is first learned by the training image group.It then uses a trained network to recognize characters.Neural networks are learning the attributes that distinguish each training image and look for similar attributes in the image being recognized.Neural networks are easy to set up, but the recognition rate is low if only training images with different tendencies are given.

Feature extraction

Feature extraction works in a similar way to neural networks, but it requires programmers to determine what they consider important and set them manually.

The attributes referred to here are as follows.

  • Aspect ratio
  • Top half pixel ratio
  • Right half pixel ratio
  • Number of line segments written with one stroke
  • Average distance from the center of the image
  • Symmetry around the Y axis
  • Symmetry around the X axis

This technique has the advantage that attributes can be controlled during the recognition process.However, features are not learned automatically, so it is like a neural network.Machine learningIt takes much longer to develop than.

Online handwriting recognition

Online handwriting recognition is specialTablet,PDAAutomatically recognizes the text entered usingSensorCollects the movement of the pen tip and the raising and lowering of the pen as data.These data are called digital inks and can be regarded as expressions of handwriting movements.This signal is pattern-matched with a character database according to a recognition algorithm, and the written characters are converted into text information and passed to the application.

Of this methodinterfaceGenerally includes the following elements.

  • Pen or stylus used by the user to write
  • It may be integrated with or adjacent to a flat input area output display that senses contact. (Input deviceAs a tablet orTouch panelIs used.
  • A software application that translates and interprets pen or stylus movements and the resulting curves into digital text.


KeyboardCommercial products that adopted handwriting recognition as an alternative to input appeared in the early 1980s.As an early handwriting terminal product Pencept Penpad [2] And InforitePOSTerminal[3]There is.Personal computerAs a huge consumer market, PenCept as an alternative to keyboards and mice[4], CIC[5]The handwriting recognition system was released by.As the first tablet portable computerGRiD Systems OfGRiDPadWas released in September 1989.For thisMS-DOSIt was equipped with an operating system based on.

Early 1990sNCR,IBM,EO Manufacturers jointlyGOCompany PenPoint OS (In a broad sense)Tablet pcHas been released. PenPoint had handwriting recognition and gesture capabilities that could be applied to third-party software. IBM initiallyThinkPadIt used the brand name of, and used its own handwriting recognition technology.This recognition system laterMicrosoft Of Windows for Pen Computing Ported to IBMOS / 2Also transplanted.Both of these have failed commercially.

Advances in electronics have made the equipment required for handwriting recognition even smaller than tablet computers.PDAThe handwriting recognition function has come to be used.The first PDA to offer handwriting input wasApple Computerof"Newton".PDA "Message Pad" with Newton OS has been streamlinedUser interface The advantage of (UI) was made known to the world.However, Newton was not commercially successful.In addition to price and processing speed issues, the unreliability of software that learns user writing patterns is also cited as one of the factors that hindered its widespread use.Newton OS 2.0 has greatly improved handwriting recognition, including unique features not found in current recognition systems, such as modeless error correction, but before that a bad first impression was formed.Apple Newton was discontinued and its features were ported to Mac OS X 10.2 and laterInkwellIt became.

palmOne (currently Palm Inc.) IsGraffitiDeveloped and sold a series of PDAs based on the recognition system. What is "graffiti"?The alphabetIt is a collection of one-stroke patterns that correspond one-to-one with, and by adopting this, the recognition accuracy has dramatically improved.The user had to memorize input patterns that are similar to the alphabet but have different stroke orders and shapes (some of which match), but once memorized, they can input comfortably with fast and accurate recognition. I was able to do it.However, the alphabet andArabic numerals, To enter characters other than some symbolsInput methodConversion work was required.The original graffiti method wasXeroxIt turned out to be infringing the patent held by Xerox, and later acquired a license for the CIC system that existed prior to the Xerox patent and switched to it.

sharpWas manufacturing PDA hardware with the Newton operating system in partnership with Apple Computer.Sharp was considering localizing the Newton OS in Japanese and selling it in Japan, but when Newton's commercial failure became clear, it was integrated into its own route.It was manufactured and sold by the companyElectronic notebookHas been extended to incorporate a pen-operated UI like Newton.The PDA "PI-3000" under its own brand was an exceptional hit for a similar product in Japan.The company is a PDAZaurusIt has been made into a series under the nickname of, and has been on sale to this day after several architectural changes.The handwriting recognition engine has been improved year by year, and even a certain amount of quirky or broken characters can be recognized with low accuracy (however, if the stroke order is incorrect, the accuracy may drop significantly).The time required for recognition is rather long, and it is not possible to input at very high speed.

MicrosoftIs an OS with a UI that assumes pen operationWindows MobileDeveloped,licenseFor sale.The OS is a highly versatile embedded OS.Windows CEBased on this, it is a combination of a UI suitable for pen operation and components necessary for using it as a PDA.There used to be a segment called "Handheld PC" equipped with a keyboard, but later it was integrated into pen operation. PDAs that adopted Windows Mobile robbed Palm of customers and secured a certain share, but did not reach new demand.Symbian OSEtc. adoptedSmartphoneIs being pushed.

For the handwriting recognition system based on the desktop OS,Windows XP Tablet PC EditionThere is.Tablet pcIs a special equipped with a touch panellaptopSo, you can enter characters by hand on the screen using a stylus. The OS recognizes the handwritten pattern and converts it into normal text.Microsoft's system does not learn the user's writing pattern, but matches it with an internal recognition database that contains similar glyphs.This system is different from the handwriting recognition system used in Windows Mobile.Windows 7 Has an advanced handwriting recognition system with a learning function, and supports not only English but also Japanese, Chinese, Korean, etc.

As for the one using a general-purpose OS,life insurance-Non-life insuranceIn the industry, handy PCs with a touch panel and no keyboard may be used with a general-purpose OS for desktops that is not Tablet PC Edition and a unique handwriting recognition engine.In addition, as a kiosk terminal, a small PC capable of pen operation may be used.In addition to improving user convenience, this also has the purpose of disabling keyboard operations and ensuring system safety.

Nowadays, with a pen with digital elementspaperAttempts have been made to write letters in and obtain digitized text from them.For exampleI noteTechnology developed by[6]Is relatively well known and has some success in the education market.Whether or not this technology will take root is unknown.

Although handwritten character recognition has become popular as an input method, it has not yet been widely used in desktop computers and the like.KeyboardThe general view is that input by is faster and more reliable. Many PDAs are equipped with a handwriting recognition function, and some can recognize cursive characters, but the low recognition rate is still a problem.Software keyboardMany users find it more efficient.


As an example of a recent handwriting recognition systemMicrosoft OfTablet pcOS for (Windows XP Tablet PC Edition,Windows Vista,Windows 7). It uses a (TDNN) classifier called "Inferno" and was developed by Microsoft itself.Newton OS CalliGrapher, which was licensed by Microsoft from ParaGraph International in 2.0 for the handwriting recognition system used in 1999, is also the second recognition engine along with TDNN. New generation version of CalliGrapher[7]Is available as a Windows Mobile version from PhatWare Corp, which was acquired by ParaGraph in 2001.

(Parascript RiteScript, which was developed by EverNote Corporation from 2000 to 2004, is said to be the third-generation handwriting recognition technology. Evernote It is used in software. ritePen integrates riteScript with the handwriting recognition function originally provided in Windows Vista to improve the recognition rate.

CellWriter[8]It is,LinuxForOpen SourceIt is a handwriting recognition program of.University of MinnesotaUndergraduate Research Opportunity Program[9] Developed by Michael Levin et al. As part of. CellWriter is a method of learning writer's habits and requires training before it can be used.Equipped with an English word recognition engine to enhance the recognition rate, software keyboard function andContext menuYou can fine-tune or correct recognition mistakes with.Unlike other recognition engines, the source code and algorithms are publicly available and can be downloaded for free.

MyScript Builder[10]Is a handwriting recognition released by Vision ObjectsSDKIt is also often used to develop applications using the digital pen developed by Anoto.


Handwritten character recognition is an area of ​​active research.ICFHR (International Conference on Frontiers in Handwriting Recognition) is held in even-numbered years, and ICDAR (International Conference on Document Analysis and Recognition) is held in odd-numbered years as the largest international conference in this field.BothIEEEThe paper scrutinized by is published.The fields where research is particularly active are as follows.

  • Online recognition
  • Offline recognition
  • Signature verification
  • Interpretation of mail address
  • Check processing

Chronological Table

  • 1915: Stylus-based handwriting recognition user interface patented in the United States[11][12].
  • 1957: Stylator Tablet: Tom Dimond ComputerInput deviceReleased an electronic tablet that can recognize handwritten characters[13].
  • year 1961: RAND Tablet Invention: Relatively more famous than its predecessor[14][15].
  • 1962: Cursive handwriting recognition[16].
  • 1969: GRAIL system: Handwriting recognition system with electronic ink display and gesture commands[17].
  • 1973: Applicon CAD/CAMシステム[18]..Adopted Ledeen's recognition system for handwriting recognition[19].
  • 1980s: General handwriting recognition system: Pencept[4] And CIC[5] Both have released a handwriting recognition system using a tablet for the general PC market. The Cadre System has added handwriting recognition to Inforite POS terminals.[20]
  • 1989: Portable Handwriting Recognition Computer: GRiDPad[21] from GRiD Systems.


  1. ^ https://sourceforge.net/projects/javaocr/ Java OCR, 05 June 2010]. Retrieved 05 June 2010
  2. ^ Pencept Penpad (TM) 200 Product Literature, Pencept, Inc., (1982-08-15), http://rwservices.no-ip.info:81/pens/biblio83.html#Pencept83 
  3. ^ Inforite Hand Character Recognition Terminal, Cadre Systems Limited, England, (1982-08-15), http://rwservices.no-ip.info:81/pens/biblio83.html#Inforite82 
  4. ^ a b Users Manual for Penpad 320, Pencept, Inc., (1984-06-15), http://users.erols.com/rwservices/pens/biblio85.html#Pencept84d 
  5. ^ a b Handwriter (R) GrafText (TM) System Model GT-5000, Communication Intelligence Corporation, (1985-01-15), http://rwservices.no-ip.info:81/pens/biblio85.html#CIC85 
  6. ^ Anoto Technology: Digital Pen and Paper, Anoto Group AB, http://www.anoto.com 2008/8/23Browse. 
  7. ^ CalliGrapher
  8. ^ CellWriter
  9. ^ Undergraduate Research Opportunity Program
  10. ^ MyScript Builder Software Development Kit
  11. ^ Goldberg, HE (1915-12-28), Controller, United States Patent 1,116,663, http://users.erols.com/rwservices/pens/biblio70.html#GoldbergHE15 
  12. ^ Goldberg, HE (1915-12-28) (PDF), Controller, United States Patent 1,117,184 (full image), http://www.freepatentsonline.com/1117184.pdf 
  13. ^ Dimond, Tom (1957-12-01), Devices for reading handwritten characters, Proceedings of Eastern Joint Computer Conference, pp. 232–237, http://rwservices.no-ip.info:81/pens/biblio70.html#Dimond57 2008/8/23Browse. 
  14. ^ RAND Tablet, (1961-09-01), http://users.erols.com/rwservices/pens/biblio70.html#RAND61 
  15. ^ 50 Years of Looking Forward, RAND Corporation, (1998-09-01), originalArchived as of October 2009, 5., https://web.archive.org/web/20090507193627/http://www.rand.org/publications/randreview/issues/rr.fall.98/50.html 
  16. ^ Harmon, LD (1962-08-01), Handwriting reader recognizes whole words, Electronics, Vol 35, August 1962, http://users.erols.com/rwservices/pens/biblio70.html#Harmon62 
  17. ^ Ellis, TO (1969-09-01), The GRAIL Project: An Experiment in Man-Machine Communications, The RAND Corporation, RM-5999-ARPA, Santa Monica, California, September 1969, http://users.erols.com/rwservices/pens/biblio70.html#EllisTO69a 
  18. ^ Computerized Graphic Processing System: System User's Manual, Applicon Incorporated, (1973-09-01), http://users.erols.com/rwservices/pens/biblio75.html#Applicon73 
  19. ^ Newman, WM (1973-09-01), The Ledeen Character Recognizer, Principles of Interactive Computer Graphics, McGraw-Hill, pp. 575–582, http://users.erols.com/rwservices/pens/biblio75.html#NewmanWM73a 
  20. ^ Inforite Hand Character Recognition Terminal, Cadre Systems Limited, England, (1982-08-15), http://users.erols.com/rwservices/pens/biblio83.html#Inforite82 
  21. ^ The BYTE Awards: GRiD System's GRiDPad, BYTE Magazine, Vol 15. No 1, (1990-01-12), pp. 285, http://rwservices.no-ip.info:81/pens/biblio90.html#GridPad90a 

Related item

外部 リンク


Back to Top