CN102981753B - The hand-written Chinese character input recognition method improved - Google Patents

Abstract

An improved device and method for handwritten Chinese character input recognition for electronic devices. Wherein, the device includes means for recognizing Chinese characters written by the user in a writing area of the electronic device, wherein the recognition is only completed when the following conditions are met: the starting point or the middle point of the previous stroke is not in a Predetermined starting area, the first few strokes constitute a completed character, and the starting point or middle point of the current stroke is in the described starting area.

Description

Improved Handwritten Chinese Character Input Recognition Method

This application is a divisional application with application number 200310115725.7, application date November 28, 2003, and titled "Improved Handwritten Chinese Character Input Recognition Method".

technical field

The present invention relates to handwritten Chinese character input recognition, in particular to an improved handwritten Chinese character input recognition method for electronic devices, which can tolerate timeout or even not be affected by timeout, and can allow partial modification of inputted characters.

Background technique

Handwritten Chinese character input is a commonly used input method for computers and other electronic devices with touch screens or touch pads (writing pads). With the wide application of notebook computers, mobile phones and personal data assistants (PDAs), fast, efficient and accurate handwritten Chinese character input recognition has become the wish of many people.

For writing Chinese characters, different people have different writing styles and writing habits. A difficult problem in handwritten Chinese character input recognition is how to determine whether the user has finished writing a word. Therefore, an obstacle to fast and accurate recognition is the so-called timeout, the waiting time after the last stroke, because timeout is the most effective way to detect whether the user has finished writing a word. In order to make the recognition of handwritten Chinese character input practical, the timeout is actually very short, usually between 0.2-0.8 seconds, and generally can be adjusted between 0.2-2.0 seconds by the system or the user.

However, using timeouts to determine input completion is sometimes problematic. First of all, it is difficult to set an optimal timeout, because the timeout varies greatly from user to user, that is, writing speed and breaks between strokes and words are largely user-dependent. Second, the timeout varies from word to word. For some characters, the user has to move a long distance between two strokes. When the distance is too long, it takes longer than normal stroke-to-stroke. In this case, a timeout occurs before the user finishes writing a word, and strokes written before and after the timeout will be recognized as two characters.

In order to solve this problem caused by timeout, a common solution is to provide the user with visual feedback in the form of handwriting, so that it can confirm when it is recognized. In addition, most handwritten Chinese character input recognition software allows the user to manually set the length of timeout. Some solutions provide the user with two or more handwriting input areas to force the user to write only one word in one input area.

In fact, the handwriting feedback must be placed below the handwriting, otherwise, the user will not notice that the recognition is going on. This approach cannot be used on devices without a touch screen, such as those with only a tablet. Allowing users to set the timeout by themselves can provide a more flexible waiting time (timeout), but it cannot solve the problem. Because there is only one same timeout for all words. In practice, it is unrealistic to require a truly flexible timeout for each character and for different strokes within a character. For those portable electronic devices with only a relatively small screen, there is only a relatively limited writing area, therefore, multiple writing areas cannot be provided.

Also, since handwriting recognition is not 100% accurate, users may sometimes rewrite a word to ensure it is recognized correctly. Since Chinese characters require a complicated writing process, it is a burden for users to write a word repeatedly, both physically and psychologically. Furthermore, this prevents quick and efficient identification. In addition, if only part (for example, a part) is correct in the primary candidate word, and if the system provides a candidate word list, the user may also want to browse all candidate words (usually 8-10 words), which makes The selection process is more difficult. And if the user thinks that the recognition is wrong, he has to rewrite the whole word.

In order to reduce rewriting and facilitate input and recognition, various solutions have been proposed, such as: classifying or decomposing characters according to radicals, etc. (such as: Wubi font and Cangjie code, etc.). US Patent No. 5,923,778 discloses a method for identifying using radical information, and proposes a database for hierarchically constructing character information based on radical information. US Patent 6,272,243 discloses a method and device for inputting into a handwritten notepad, which has a predefined sub-input area. However, it is only suitable for inputting Japanese.

Summary of the invention

The purpose of the present invention is to provide an improved handwritten Chinese character input recognition method, which is basically not affected by timeout, and can tolerate timeout or even not be affected by timeout at all.

The object of the present invention is also to provide an improved handwritten Chinese character input recognition method, which can reduce rewriting in handwritten Chinese character input recognition, increase input speed and allow partial modification of a word.

According to one aspect of the present invention, an improved handwritten Chinese character input recognition method for an electronic device is provided, comprising the following steps: identifying a Chinese character written by a user in a writing area of the electronic device according to predetermined rules, and generating a The list of candidate words identified by the above recognition step, and displaying the list of candidate words in a display area of the electronic device; wherein the list of candidate words at least includes the current word just recognized and the current word and the current word A combination of previously recognized and displayed candidate words.

According to another aspect of the present invention, there is provided an improved handwritten Chinese character input recognition method for electronic devices, including the step of recognizing Chinese characters written by a user in a writing area of the electronic device, wherein the Chinese character recognition only Just finish when following condition is met: the starting point or middle point of previous stroke is not in a predetermined starting area, the first few strokes constitute a completed word, and the starting point or middle point of current stroke is in described starting area.

According to another aspect of the present invention, there is provided an improved handwritten Chinese character input recognition method for an electronic device, which includes the steps of recognizing a Chinese character written by a user in a writing area of the electronic device, and The step of displaying the list of main candidate characters identified in the recognition step; wherein said display includes prompting different parts of the font of the main candidate characters in the background part, and said prompt part constitutes an independent writing area, in which In the writing area, it can be recognized independently.

Description of drawings

Embodiments of the present invention are illustrated below with reference to the accompanying drawings.

Figures 1A and 1B schematically illustrate the possible misrecognition of the prior art handwritten Chinese character input recognition method utilizing overtime for recognition;

Figure 2 shows an embodiment of the method according to the invention;

Fig. 3 is the flowchart of the embodiment described in Fig. 2;

Fig. 4 is another embodiment according to the method of the present invention;

Fig. 5 is the flowchart of the embodiment described in Fig. 4;

6A and 6B are yet another embodiment of the method according to the present invention.

FIG. 7 is a flowchart of the embodiment described in FIG. 6 .

Detailed ways

Since the handwritten Chinese character input recognition can be realized by various methods well known in the art, and conventional systems or software can be used for recognition, for the sake of simplicity, the recognition process itself will not be described in detail here.

Referring first to FIG. 1 , FIG. 1A schematically illustrates possible misrecognition in the prior art handwritten Chinese character input recognition method using timeout for recognition. As shown in FIG. 1A, the user intends to write a left-right type character "Ming" consisting of two parts. He usually writes the "day" on the left first, and then the "month" on the right. Since "Ming" is composed of two parts, the radical "日" and a "month" that can independently form a character, he needs to move from the last stroke of the left half "日" (the middle lower part of the input/writing area) A long distance to the first stroke of the "month" in the right half (the upper middle part of the input/writing area), which takes a relatively long time and may have a short pause between the two parts. If the time of pause is longer than the preset overtime, the system will recognize the left part as "day" or "month" and display "day" and "month" in the candidate list. Then, the system recognizes the part on the right as "month" or "day" and displays the recognized "month" and "day" in the candidate list. In other words, due to the timeout, the system recognizes the two parts as two separate words.

The same problem will also occur in the recognition of the two parts of the up and down type. As shown in FIG. 1B, the Chinese character "喜" is composed of the radical "土" and the lower part which does not constitute a complete character. Due to the timeout, the system may recognize the character "lucky" as two characters, namely the character "earth" and the character "sheep" which looks like the lower part.

In actual application, adjusting the waiting time (timeout) is not helpful, because the timeout varies with Chinese characters, strokes, writing habits and users. Even for the same user, for whom a fixed timeout does not apply to all words.

Figure 2 shows an embodiment of the improved method according to the invention. This embodiment utilizes the frequently used and crucial candidate word list for handwritten Chinese character input recognition, so that while including the current word, it also includes the combination of the current word and a character recognized and displayed before the current word . The candidate list can be displayed on the touch screen or a separate display (in the case of a tablet, a separate display is usually required to display the words written on the tablet). This embodiment only adds the most probable combination of the current word and a word recognized and displayed before the current word to the candidate list, and does not require additional display components to occupy the display screen.

As shown in FIG. 2 , it is assumed that the same user wants to write the same character "Ming" consisting of two parts. According to the writing rules of strokes, he will first write the left half of "日" at a speed suitable for his writing habits. If due to his writing habits or other reasons, there is a pause before starting to write the right half, and the pause is faster If the timeout is long, the system will consider that the input has been completed, and recognize the left half as the word "日" by means of a known method. "日" is displayed on a certain part of a display screen as the primary candidate word, and in the present embodiment, the candidate word is displayed above the writing area.

The user continues to write the remaining part, namely the right half "month". After the timeout, the system generates a new list of candidate characters, which lists the "month" he just wrote and the combination of "month" and the previously recognized and displayed "day", that is, the character "Ming" he wants to write . Then, the word "Ming" can be displayed above the writing area as one of the primary candidate characters or candidate characters. The word "Ming" as the primary candidate word can be set in the system to automatically enter the text area or be selected by the user, for example, by clicking or in other ways. After selection, the word is moved to the text area.

In this embodiment, the characters that can be recognized do not have to be complete Chinese characters, but can be components of characters such as radicals. For example: If a user writes the character "Ban" and there is a timeout after completing the left half, which is a radical rather than a complete Chinese character, the system will recognize the radical before he continues to write the right half "Half" and Display the radical in the candidate list. After the user finishes writing the right half, the system will generate a new list of candidate characters, which includes the current character "Half" just recognized and the character "Accompanied by the combination of "Half" and the previously recognized and displayed radicals. ". The user can select the word, and the system will move the word to the text area.

Furthermore, said combination must form a complete word. Depending on the size and layout of the display screen, the candidate list can include up to 10 words for easy selection.

In this way, by only adding the most likely combination of the current word and a word recognized and displayed before the current word in the candidate word list, usually only one word, the efficiency of recognition can be greatly improved without increasing the display parts and occupying the display Screen.

FIG. 3 shows a flowchart of the embodiment in FIG. 2 . Detection begins when the user starts writing on the touch screen or tablet. The system detects at any time whether there is an interruption (timeout) longer than a predetermined time between strokes (step 102). If the system detects overtime, it will identify the character (step 201) formed by the strokes before the overtime, and then display the identified word (step 202) in the candidate list. Meanwhile the system continues to detect (step 103). If another timeout is detected (step 104), it is identified (step 105) and displayed (step 106). In step 106, the system displays the word recognized in step 105 and the combination of the word recognized in step 105 and the word recognized and displayed in steps 201 and 202. If no timeout is detected, continue detection. After the display in step 106, the system places the word selected by the user in the text editing area according to the user's selection.

Figure 4 is another embodiment of the method according to the invention. In addition to using the above-mentioned candidate list, by detecting the position of the stroke of a character on the touch screen or tablet, the influence of timeout on recognition can be almost completely eliminated.

Since the writing of Chinese characters follows the rules from left to right and from top to bottom, for example: on a touch screen/writing pad, writing usually starts from the upper left corner of the writing area and ends at the lower right corner, so you can place a character at the starting position of the pen, that is, The upper left part of the writing area is defined as a "start area". The size of the start area is ideally slightly larger than a quarter of the writing area or writing box so that most users can start writing in the start area. If a character includes two or more parts, for example: left and right or upper and lower parts, you can start writing another part from a place other than the start area. In this embodiment, this feature of Chinese character input is used to detect whether a character has been written and whether another character has been written.

According to this embodiment, in the following cases, a stroke will be considered as the beginning of a new word: 1) the starting point or the middle point of the previous stroke or the previous several strokes are not in the starting area; 2) the previous several strokes constitute a Complete character; 3) The starting point of the current stroke or the middle point of the current stroke is in the starting area. That is to say, only when above-mentioned three conditions are met simultaneously, just think that the word that is formed by preceding several strokes has been written, and current stroke is the beginning of a new word. Therefore, instead of waiting for a timeout, the system will determine whether a word has been written based on the above conditions instead of a timeout.

As shown in FIG. 4 , the starting area is at the upper left corner of the writing area or writing box, and can be marked with lines or graphics to prompt the user to start writing in this area.

Also take the word "Ming" as an example, which is composed of left and right parts and should be written next to the word "日" on the left. The starting point and the middle point of the first three strokes next to the word "日" are all in the starting area. Although the starting point of the fourth stroke may be outside the starting area, the first three strokes do not constitute a complete word. Therefore, the beginning of the fourth stroke is not considered to be the beginning of a new character. The right half of the "month" is generally written from outside the starting area. Although the left half of "日" can also form a complete character as a radical, the first stroke of the right half of "月" will not be considered as the beginning of a new character, because the first stroke of "月" is at the beginning Outside the district, the third condition above is not met. The system will detect the starting point of each stroke to determine whether a character has been written according to this rule. It can also be set to make the system detect the middle point of each stroke, that is, whether the middle part of the stroke falls in the starting area. The written word will be recognized and displayed in the candidate word area.

Since the user may begin to write the right half of "month" in the start area due to neglect and other reasons, and this may cause the system to misrecognize the right half of "month" as a complete "month", therefore, the candidate word The table should include, in addition to the current word, the combination of the current word and a word immediately preceding the current word, as described with reference to FIG. 2 . In this example, both "月" and "明" are displayed in the candidate list, so errors caused by the user failing to start writing the right half of "月" outside the start area will not affect the correct recognition of the system .

FIG. 5 shows the flow of the embodiment described in FIG. 4 . Detection begins when the user starts writing on the touch screen or tablet. The system first detects whether the starting point/intermediate point of the previous stroke is in the predetermined starting area (step 301). If not, then detect whether the first few strokes form a complete Chinese character (step 302). If so, then continue to detect, until the starting point/intermediate point of detecting a stroke is not in the starting area (step 401). If the first few strokes form a complete Chinese character, then detect whether the current stroke starting point/intermediate point is in the starting area (step 303). If the result of the detection is affirmative, then the characters formed by the first few strokes are recognized (step 304), and the characters recognized in step 304 and the characters recognized and displayed in the previous round of detection are displayed (step 305). If the result of step 303 is negative, go back to step 401. Finally, the system generates and places the selected word in the text editing area according to the user's choice.

Figures 6A and 6B are yet another embodiment of the method according to the present invention, which enables the user to partially modify a word.

According to the Chinese Character Type Index System (SKIP), Chinese characters can be divided into four types, namely the left-right type (for example: "Ming"), the upper-lower type (for example: "Wu"), and the surrounding type (such as : "country" character) and solid type (for example: "big" character). Therefore, in addition to preset a starting area in the writing area of the display screen, some method, such as: color or grayscale (in the case of a monochrome display screen) can be used to prompt the font type of the recognized word.

Each color-prompted part, for example: the left part and the right part of the left and right type form an independent recognition area respectively. Here, the system can recognize a complete character or a radical, etc., but it should be recognized in the entire writing area. The prime candidate for is a complete word.

As shown in FIG. 6A , when inputting the Chinese character "mix" in the writing area of the touch screen, the user starts from the upper left corner (starting area). The system recognizes the user's input. The identification can be carried out by the method described in FIG. 2, or by the method in FIG. 4, or even in a known manner. Because, for example: the handwriting is illegible, the primary candidate character shown in the writing area is "companion" rather than "mixing" that he wants to write, and the background of the writing area prompts that the word is a left-right type with different colors or gray scales. At this point, the user doesn't have to rewrite the entire word because the top candidate is "almost right," meaning it looks like what the user intended to write, except that the left half is wrong. At this point, the user can rewrite only the left half. When the user rewrites the left half, the background color (grayscale) of the left area changes to confirm that the user only modifies the left half, leaving the right half unchanged. The system generates a new primary candidate word "mix" according to the user's modification, which is just the word the user wants to write.

Candidate words can also be displayed in a candidate word list at the same time, for example: in the candidate word list above the writing area shown in FIG. 4 .

Another way is to make the list of candidate characters include all (according to the number of characters that can be displayed) characters with radicals or radicals selected by the user. In this case, the user can select the correct part in the writing area instead of modifying the wrong part. As shown in Figure 6B, when the user notices that the displayed primary candidate word is not the word he actually wants to write, he can select the correct part, for example: by clicking on the area with the correct part (right area) to Choose the right part. Then, all the words with the correct part "half" are displayed in the candidate list, in this example, "companion", "mixing", tripping" etc. The user can pass through the candidate list, for example: click Select the word he wants to write (“mix”) and select the word “mix”.

This embodiment is not applicable to a tablet with a separate display screen.

FIG. 7 is a flowchart of the embodiment described in FIG. 6 . After the user finishes writing a word, the system recognizes it with the method shown in Figure 2 or Figure 4, or in a known manner, and then displays the recognized primary candidate word (step 502), and prompts the font type of the word with color/grayscale ( Step 503), different colors/gray levels represent different regions of the font. If the displayed word is exactly the word written by the user (step 504), then the system generates the selected word according to the user's selection and places it in the text editing area (step 505). If part of the word shown in step 503 is correct, the user can choose to confirm the correct part or directly modify the wrong part. If the user chooses to confirm the correct part (step 601), the system will generate a candidate list (step 602), which includes all the words that contain the correct part. If the user chooses to directly modify the wrong part (step 603), the system will prompt the user by changing the color/grayscale of the modified part to identify and display the modified part.

Although the manner of prompting the user is described by taking color/grayscale as an example, obviously other manners well known to those skilled in the art may also be used.

The present invention has been described above in the form of embodiments, but the present invention is not limited to the embodiments, and those skilled in the art can make various modifications to the present invention within the scope of the appended claims.

Claims (12)

1. the equipment of the input of the handwritten Chinese character for the electronic installation identification improved, comprise for identifying that user writes the device of the Chinese character that district writes of described electronic installation, wherein said identification only just completes when meeting following condition: the starting point of previous stroke or intermediate point be not or not a predetermined beginning district, front several stroke forms a word completed, and the starting point of current stroke or intermediate point are in described beginning district.

2. equipment according to claim 1, writes the upper left quarter in district described in wherein said beginning district is positioned at.

3., according to the equipment of claim 1 or 2, described in the area in wherein said beginning district is greater than, write 1/4th of district.

4. equipment according to claim 1, described equipment also comprises the device for generation of an above-mentioned candidate's word table identified for the device identified, and for showing the device of described candidate's word table in a viewing area of described electronic installation.

5. equipment according to claim 4, wherein said candidate's word table at least comprises the current word and this current word and the candidate's combinatorics on words identifying before this current word and show that have just been identified.

6. equipment according to claim 4, also comprises the device for selecting a word in the described candidate's word table from the display of described candidate word viewing area.

7. the hand-written Chinese character input recognition method for electronic installation improved, comprise and identify that user writes the Chinese character step of writing in district of described electronic installation, wherein said identification only just completes when meeting following condition: the starting point of previous stroke or intermediate point be not or not a predetermined beginning district, front several stroke forms a word completed, and the starting point of current stroke or intermediate point are in described beginning district.

8. method according to claim 7, writes the upper left quarter in district described in wherein said beginning district is positioned at.

9., according to the method for claim 7 or 8, described in the area in wherein said beginning district is greater than, write 1/4th of district.

10. method according to claim 7, also comprises the step of generation one at candidate's word table of above-mentioned identification step identification, and shows the step of described candidate's word table in a viewing area of described electronic installation.

11. methods according to claim 10, wherein said candidate's word table at least comprises the current word and this current word and the candidate's combinatorics on words identifying before this current word and show that have just been identified.

12. methods according to claim 10, also comprise the step selecting a word from described candidate's word table of described candidate word viewing area display.