JPH0468670B2 - - Google Patents

Description

[Detailed Description of the Invention] (Technical Field) The present invention relates to a recognition method that can cope with the slow input speed of handwritten characters written by a scribe in an online character recognition device used as an input device for information processing equipment. .

(Background technology) Conventionally, in online character recognition systems, strokes belonging to the same character frame are considered to be the same character, and strokes belonging to the same character frame are considered to be the same character, and strokes belonging to the same character frame are considered to be the same character, and strokes belonging to the same character frame are When a stroke belonging to the same character frame is input, the character is extracted by considering the previous strokes belonging to the same character frame as one character (hereinafter referred to as character extraction by character frame), and a method is used to recognize the character. Ta. However, in this method of cutting out characters using a character frame, the character will not be cut out unless the next character is written down, so it may be necessary to check the recognition result by writing down each character one by one, or to write down the characters one after another. When checking the final character in a case where a character is entered, the user must touch a separately set entry character frame with a pen to extract the character and perform recognition processing, which is a cumbersome operation. . In addition, in order to avoid this troublesome operation, after writing a character, if the next character is not written within a certain period of time,
Until now, a method has been adopted in which the strokes input within the same character frame are regarded as one character and the character is extracted (hereinafter referred to as character extraction based on time) to recognize the character. However, in this time-based character extraction method, the same characters may be separated if they are written slowly, so it is necessary to set a sufficiently long time for time extraction, which requires extra waiting time in addition to the recognition processing time. As a result, it took a long time for the recognition results to be output, which was a time nuisance.
In reality, these two methods are used to deal with the problem, but it has not yet been able to sufficiently cope with the slow input speed of handwritten characters written by scribes, and it is said that this places a burden on the scribes. There were flaws.

(Objective of the Invention) In order to eliminate the above-mentioned drawbacks, the present invention eliminates the waiting time during extraction in the time-based character extraction method until the recognition result is output to the display, and the recognition result is output. The present invention provides a recognition processing method that can cope with the slow input speed of handwritten characters written by a scribe by minimizing the time required for writing.

(Summary of the Invention) In character recognition processing, this method performs character segmentation based on time, so each stroke of a written character is the same character without waiting for a segmentation time (waiting time) that is set at a certain time. Each time an input is made within the frame, the stroke up to that point is temporarily regarded as one character, character extraction is performed, and recognition processing is started. If there is no stroke input after starting the recognition process, the recognition result is output as a recognized character to the internal display and further to the external interface in the minimum amount of time. If the next stroke is written in another input character frame within the character extraction time after the recognition process starts, the recognition result will be
After the strokes are output to the internal display and the external interface at the same time as the recognition ends, a new recognition process is started for the strokes written in the separate character frame as separate characters. Also, if the next stroke is written within the same character frame within the character extraction time after the recognition process starts, the recognition result is invalidated, the added stroke is considered as one character, and a new recognition process is started. Start.

Thereafter, this process is repeated every time a stroke is input within the same character frame to advance the recognition process. By minimizing the waiting time other than the recognition processing time until the recognition result is output to the internal display through the recognition processing explained above, it can cope with the slow input speed of handwritten characters written by the scribe. This eliminates the troublesome operation and time required for the scribe to check the recognition results, and will be explained in detail below.

(Structure and operation of the invention) FIG. 1 is a block diagram showing an embodiment of the present invention, in which 1 is a tablet for inputting coordinate data of written characters, and 2 is a tablet for determining whether the pen is on or off based on information from the tablet. Pen on/off detection section, 3
4 is a character extraction detection unit that performs character extraction according to the character frame of the input character based on the input coordinate data from the tablet and the pen on/off signal from 2, and 4 is the character extraction detection unit that performs character extraction for each stroke based on the input coordinate data of the written character from the tablet when the pen is on. Processing such as stroke feature extraction is performed, and further, based on a character extraction signal output from the character extraction detection unit 3 or a character extraction timer (M) 5, which will be described later,
A stroke data processing section 5 manages strokes for each character, and 5 is a character cutting timer that starts counting with the pen-off signal outputted from 2 and is composed of a counter etc. that is reset with the pen-on signal, and outputs a character cutting signal based on time. (M), 6 is composed of a counter that starts counting with a pen-off signal and is reset with a pen-on signal, like 5, and a display timer (D) that outputs a display enable signal to display the recognition result on the internal display; Reference numeral 7 indicates a character dictionary section in which character feature information composed of stroke feature information for each character is stored, and 8 indicates a recognition process by comparing and collating the stroke information of characters in the stroke data processing section 4 and the character dictionary section 7. A recognition unit 9 outputs the recognition result from the recognition unit 8 and displays it on a display timer (D) 6.
An internal display 10 displays the recognition result output from the recognition unit 8 by a display enable signal output from the character extraction unit 3 or externally by a character extraction signal output from the character extraction timer (M) 5. This is an external interface unit that outputs data to the external interface. In an embodiment of the above configuration,
Coordinate data output from the tablet 1 is sent to a pen-on-off detector 2, pen-on-off is determined from the coordinate data from the tablet, and the pen-on-off detector 2 outputs a pen-on signal or a pen-off signal.

The character cutting unit 3 detects in which character frame each stroke of the input character is written based on the pen-on signal output from the pen-on-off detection unit 2 and the coordinate data output from the tablet 1, and distinguishes between different characters. When written in the frame, the previous stroke is regarded as one character, character cutting is detected using the character frame, and a character cutting signal using the character frame is output. Character cutting timer (M) 5
is a timer that starts counting with the pen-on signal output from the pen-on-off detection section 2 and is reset with the pen-on signal, and monitors the time after each stroke talk writing of the input character is completed, and calculates the elapsed character cutting time (T _M ). Outputs the trailing character cutting signal. The display timer (D) 6 is similar to the character cutting timer (M) 5, and outputs a display enable signal after a certain display timing time (TD) has elapsed.

The stroke talk data processing unit 4 takes in the coordinate data of the written input character from the pen-on signal output from the pen-on-off detection unit 2 and the coordinate data from the tablet 1, and performs noise removal, smoothing, normalization, etc. After performing preprocessing, known stroke extraction, stroke feature extraction, etc. are performed. Furthermore, a character cutting command is output to the stroke data processing section 4 in response to a character cutting signal output from the character cutting section 3 or the character cutting timer (M) 5. The following processing is performed until this command is output. The strokes written in sequence until this command is output are considered as strokes that constitute one character,
Stroke management of input characters is performed sequentially. Each piece of managed stroke information is sequentially output to the recognition unit 8 as input character information. The recognition unit 8 is
The stroke information of the unknown input characters sequentially output from the stroke data processing section 4 and the stroke information of the standard pattern of recognized characters stored in the character dictionary section 7 are subjected to stroke processing using techniques such as the known stroke analysis method. Each time the characters are sequentially output from the unit 4, they are compared and verified to recognize unknown input characters. The internal display 9 is a display timer (D)
An internal display or external interface 10 displays the recognition result output from the recognition unit 8 in response to a display enable signal output from the character extraction unit 3 or the character extraction timer (M) 5. By the character cutting signal,
This is an external interface that outputs the recognition result output from the recognition unit 8 to the outside.

In the above operation description, the relationship between the stroke writing state of the input character and the output of the recognition processing result will be explained in detail using FIG. 2. A display enable signal is output from the display timer (D) 6 after T _D from the completion of stroke writing, and a character cutting signal based on time is output from the character cutting timer (M) 5 after T _M , and the settings are such that T _D < _TM . shall be. Also, the time required for recognition processing
T _R and the time until the next SU talk starts being input are T _OFF . After the (N-1)th stroke is written, the recognition process is started and the result is obtained after _TR . If the next Nth stroke starts writing within the same character frame as the (N-1)th stroke within T _M time, the recognition result by T _R will be invalid.
A recognition process is newly started after the Nth stroke is written. At this time, if T _R < T _OFF , the recognition result has already been obtained, and if T _R < T _D , this invalid recognition result is output to the internal display at the timing of T _D , but it is not output to the external interface. No output. If _TR > T _D , the display enable signal has already been output, so the recognition result will be output to the internal display immediately after _TR . Also, T _R
If the input of the next Nth stroke is started during the recognition process at >T _OFF , the recognition process is stopped midway, and the recognition process is newly started after the Nth stroke is written. At this time, since recognition has not yet been completed, it is naturally not output to the internal display or external interface. Also, the next Nth stroke within T _M time is the (N-th)
1) If writing is started in a character frame different from the writing character frame of the stroke, the recognition results up to the (N-1)th stroke are valid regardless of the magnitude of T _R and T _OFF , and the character frame Because the character cutting signal is output from the character cutting unit 3, the recognition result is output to the internal display and external interface immediately after T _R , and the Nth stroke is the first stroke of the new character.
Recognized and processed as a stroke. Also, if no stroke is input after the Nth stroke, T _R
After the recognition is completed, the recognition result is output to the internal display or external interface according to the time T _D or T _M , and when T _R > T _D or _TR > T _M , the recognition result is output to the internal display or to the external interface immediately after T _R. Output to external interface.

(Effects of the Invention) As explained above, the present invention performs the recognition process by temporarily treating the stroke up to that point as one character after the stroke writing is completed, without waiting for the elapse of the extraction time before starting the recognition process. Start, and while proceeding with the recognition process, monitor the input information and time of the next stroke and display or output the recognition result to the outside, regardless of the input speed of the written character.
This method has the advantage that recognition results can be obtained without requiring any waiting time, and that the time until the recognition processing results are finally output can be minimized.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of an embodiment of the present invention, and FIG. 2 is a diagram illustrating the stroke writing state of input characters and the temporal relationship in which recognition results are output. 1... Tablet, 2... Pen on/off detection section, 3... Character cutting section, 4... Stroke data extraction section, 5... Character cutting timer (M),
6... Display timer (D), 7... Character dictionary section, 8... Recognition section, 9... Internal display, 10...
...external interface.

Claims (1)

[Claims] 1. An online character recognition method for recognizing characters written on a tablet, comprising: a character cutting unit that outputs a character cutting signal using a character frame from coordinate data of a written character; and stroke information of an input character from the coordinate data. a stroke data processing unit that sequentially extracts character information for each stroke written; a character dictionary unit that stores standard patterns of characters; and a character dictionary that extracts stroke information and character information of input characters that are output from the stroke data processing unit. A recognition unit that performs recognition processing based on the information up to the written stroke by comparing it with the standard pattern of the part, and a display timer that monitors the pen-off time after writing each stroke and outputs a display enable signal after the display timing time TD. an internal display that displays the result of the recognition process by the recognition unit using a display enable signal of the display timer; and a character that monitors the pen-off time and outputs a character cutout signal by time monitoring after the character cutout time TM. a cutout timer; and an external interface that outputs the result of the recognition process at that point in time of the recognition unit to the outside in accordance with a character cutout signal from the character frame of the character cutout unit or a character cutout signal based on time monitoring of the character cutout timer. The recognition unit sequentially performs recognition processing based on the stroke information and character information from the stroke data processing unit up to that point after the pen is turned off due to each stroke of the input character, and the display timing time TD is determined by monitoring the time of the display timer. If there is no input of the next stroke within the time, the recognition result of the recognition process performed sequentially by the recognition unit up to the previous stroke is displayed on the internal display, and the character cutting timer is set to be longer than the display timing time TD. If the next stroke is not input even within the character cutting time TM by time monitoring, the recognition result of the recognition unit at that point is determined and outputted to the outside via the external interface, and the display timing time TD or If the next stroke is within the same character frame within the character cutting time TM, that stroke is continuously recognized as a single character with the previous stroke, and the next stroke from the character cutting part is processed within the character cutting time TM. Online character recognition characterized in that when a stroke is determined to be in another character frame, the result of character recognition processing based on the previous stroke is displayed on the internal display and output to the outside via the external interface. Method.