JPH07296105A - Recognizing and converting device - Google Patents

Abstract

PURPOSE:To customize a dictionary for recognition such as making characters in the running style into a dictionary while keeping the stability of recognition by judging whether any reflection is performed or not by a user corresponding to the degree of matching with a recognizing character and reflecting an input feature amount while considering the stability of following recognition. CONSTITUTION:An input hand is judged by a recognizing part 130, the input feature amount is expressed by a feature amount spatial coordinate, a distance from the input hand to the spatial coordinate is calculated, and characters satisfying the degree of certainty and a prescribed reference inside the dictionary are nominated. Afterwards, a character code is possessed by the recognizing part 130 and displayed out on the display screen of a tablet 110 with integrated display, and the selection of an input intended character is accepted. In the case of that selection, when the user selects any character after the second rank although the degree of certainty for a maximum certainty character is high, a dictionary execution confirming part 190 decides whether the degree of certainty for a character in the first rank is higher than the fixed reference value or not. When the degree of certainty is higher than the reference value, it is presented and displayed, and the danger of mixed recognition is warned to the user.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention recognizes handwriting information obtained by running a pen and reading information obtained by an optical reading process by a scanner, and converts it into a character code and a figure code. Regarding the converter.

[0002]

2. Description of the Related Art The recognition conversion device is a device that can substitute the code information input by a key punch for a pen running operation or the like, and has been actively used in recent years. Among them, what is called a handwriting recognition device, in which the above-mentioned input operation can be performed in a handwriting sense, has been rapidly spread in recent years, and the recognition conversion device will be described as a device that performs a handwriting recognition function.

First, the handwriting recognition function will be described. As shown in FIG. 10, the number of strokes, the degree of bending of the first stroke and the second stroke, a1, a2, a3, a4 ..., Direction between the start points of the first stroke and the second stroke, b1, b2, b3, b4, start point Several types of appraisal information for distinguishing character-shaped characteristic parts such as directions c1, c2, c3, c4 between end points are prepared. Furthermore, the appraisal result by the appraisal information is made into a dictionary for each character.

After the dictionary is formed, the input handwriting is evaluated by these pieces of evaluation information, the degree of coincidence with the dictionary-based evaluation result is calculated, and all calculation results higher than a certain standard are nominated. All of these nominated characters are displayed and output for selection. The user selects one of these display outputs that is intended for input. When selected, the character code of the selected character is input.

The information about the above characteristic portion is called a characteristic amount. In the above dictionary, what kind of feature amount is selected is selected from the following two types according to the purpose. The former blindly learns the character input by the user to form a recognition dictionary, and is a method premised on customization for one user. In addition, this method also enables the use of dictionaries, habits, and the like in a dictionary.

The latter method collects the handwritings on the monitor by several tens or more, and uses the characteristic parts found in any of the collected handwritings to create a dictionary, which is a method for generalizing the recognition dictionary. is there. Of course, even with this latter method, input feature quantity reflection is possible to some extent.

[0007]

However, the recognition conversion device in the above-mentioned prior art has problems in recognition accuracy and usability. The problem of the recognition accuracy may occur due to the former dictionary. For example, the input handwriting t100 shown in FIG.
Suppose you went in. The input handwriting t100 is considered to be the character "wa" rather, and the matching degree is in the order of "wa" and "re". Regardless of such a degree of coincidence, when the user determines the character "re" as the input character,
The learning function reflects the characteristic portion of the input handwriting t100 in the dictionary for the second character "re". As a result of such reflection, the contents of the recognition dictionary for the characters "wa" and "re" are very similar, and the input handwriting is "wa",
It becomes impossible to accurately recognize which one is “re”. This means deterioration of the recognition dictionary, and a decrease in the degree to which the recognition result can be accurately derived (called recognition stability).

In the latter method, the recognition stability is maintained due to the versatility of the recognition dictionary. On the contrary, since the input feature quantity is filtered from the viewpoint of maintaining the recognition stability, the reflection of the feature quantity that is intended to be input in habitual characters is almost rejected. Therefore, it is almost impossible to input habit characters, and the user is obliged to input correct handwriting.

SUMMARY OF THE INVENTION In view of the above problems, it is an object of the present invention to provide a recognition conversion device capable of customizing a recognition dictionary, such as a dictionary of broken characters and habit characters, while maintaining recognition stability.

[0010]

In order to achieve the above object, the recognition conversion apparatus according to claim 1 is an input feature amount extraction means for extracting a feature amount which is information peculiar to input data, and characters, symbols, figures, etc. And a storage unit that stores a recognition reference character that is a feature amount that can be taken by the shape of each recognition reference character as a recognition dictionary, and a matching degree between the input feature amount and the feature amount of each recognition reference character. Of the recognition reference characters, one of the recognition candidates, a nomination unit that nominates a recognition reference character whose degree of coincidence with the input feature amount satisfies a predetermined reference as a recognition candidate.
When the selection means to be selected by the operator and the recognition reference character having the second highest degree of matching in the selection are selected, it is determined whether or not the recognition reference character having the first highest matching degree satisfies a predetermined criterion. Determination means based on the determination result, first reflection means for reflecting the input feature amount in the recognition dictionary when the first-ranked recognition reference character does not satisfy a predetermined criterion, and the first-ranked recognition character. When the recognition reference character satisfies a predetermined reference, warning means for warning the possibility of confusion of recognition and warning means for allowing the operator to select whether or not to reflect the input feature amount, and operation for reflecting the input feature amount despite the above warning A second reflecting means for reflecting the input feature quantity in the recognition dictionary when the person selects it.

According to another aspect of the recognition conversion apparatus of the present invention, the input feature quantity extraction means obtains n-dimensional information by the appraisal processing using the n kinds of appraisal information as the input feature quantity for the input data, and stores it in the storage means. When the shape of each recognition reference character is identified by the identification process, n-dimensional information that can be identified is stored as a feature amount, and the nomination means uses the n-dimensional information that is the feature amount as a reference of an n-dimensional space. It is used as the axis coordinate value, and the inter-coordinate distance between the input feature amount and the feature amount of the recognition reference character is used as the degree of coincidence.

The recognition conversion device according to claim 1 or 2 further includes sub-selection means for accepting a selection of a recognition reference character outside the nomination from an operator, and in the selection by the selection means instead of the determination means. Second degree of coincidence
If the recognition reference characters after the rank are selected, or if the recognition reference characters outside the nomination are selected by the sub-selection means, it is determined whether or not the recognition reference character with the first matching degree satisfies the predetermined criterion. It is characterized in that it is provided with a second judging means for judging based on the degree.

[0013]

According to the recognition conversion device of the present invention, the feature quantity which is the information peculiar to the input data is extracted by the input feature quantity extraction means, and the recognition reference which is a character, a symbol, a figure or the like is stored in the storage means. The character and the feature amount that the shape of each recognition reference character can have are stored as a recognition dictionary. After the input feature amount is extracted, the degree of coincidence between the input feature amount and the feature amount of each recognition reference character is calculated by the nomination means, and the recognition criterion among the recognition reference characters whose degree of coincidence with the input feature amount satisfies a predetermined criterion. The character is nominated as a recognition candidate. The recognition candidate is provided for selection by the selection means, and the matching degree second
When the recognition reference character of the rank or higher is selected, the determination means determines whether or not the recognition reference character of the first matching degree satisfies a predetermined criterion based on the matching degree.
As the predetermined criterion, it can be adopted whether or not the degree of coincidence of the recognition reference character having the first highest degree of coincidence exceeds a predetermined threshold value. Alternatively, it can be adopted whether or not the difference between the degree of coincidence of the first-ranked recognition reference character and the second-ranked one is equal to or more than a predetermined threshold value. As a result of the judgment, the first
If the rank recognition character does not meet the predetermined criteria,
The input feature quantity is reflected in the recognition dictionary by the first reflecting means. If the first-ranked recognition reference character satisfies a predetermined standard, the warning unit warns of the possibility of recognition confusion, and the operator is allowed to select whether or not to reflect the input feature amount. When the operator selects the reflection of the input feature amount despite the warning, the input feature amount is reflected in the recognition dictionary by the second reflecting means.

Further, in claim 2, the input feature amount extracting means performs the appraisal process using the n types of appraisal information on the input data, and the n-dimensional information is obtained as the input feature amount. Further, in the storage means, n-dimensional information that can be evaluated when the shape of each recognition reference character is evaluated by the evaluation processing is stored as a feature amount. The n-dimensional information, which is the feature quantity, is used as a reference axis coordinate value in the n-dimensional space by the nominating means, and the inter-coordinate distance between the input feature quantity and the recognition feature quantity is regarded as the degree of coincidence.

Further, in claim 3, claim 1 or 2
Since the described recognition conversion device further includes the sub-selection means, the sub-selection means accepts the selection of the recognition reference character outside the nomination, and instead of the determination means, the second
Since the second determination means selects the recognition reference character having the second highest degree of coincidence in the selection by the selection means, or the sub-selection means selects the recognition reference character outside the nomination. When selected, it is determined based on the degree of coincidence whether or not the recognition reference character having the first highest degree of coincidence satisfies a predetermined criterion.

[0016]

DESCRIPTION OF THE PREFERRED EMBODIMENTS An electronic memo device will be described below as one embodiment of the recognition conversion device according to the present invention. The electronic memo device of this embodiment is capable of inputting hiragana by handwriting recognition. The configuration of the electronic memo device is shown in FIG. The electronic memo device 100 includes a display-integrated tablet 1
10, a coordinate detection unit 120, and an input handwriting detection unit 125
The input handwriting storage unit 127, the recognition unit 130, the recognition dictionary 140, the recognition confirmation input unit 150, the display font ROM 160, the memo buffer 170, the dictionary registration unit 180, and the dictionary conversion confirmation unit 190. And the control ROM 20
0 and the CPU 210.

The display-integrated tablet 110 has a display function as a liquid crystal panel and a traveling plate function as a pen tablet for the pen 105. FIG. 2 shows an example of display contents of the display-integrated tablet 110. FIG. 2 is a display screen for inputting handwriting, and includes mode icons m1, m2, m3, m4, ... For setting the recognition input mode.
Recognition icon s that accepts pen touch input for recognition execution
110 and a handwriting input frame s120 for displaying the dragged track as a handwriting image h100 are displayed.

The coordinate detecting section 120 detects the position of the pen tip from the voltage induced by the contact of the pen tip. As shown in FIG. 2, the pen tip position is given by the X and Y coordinates when the left end of the input frame is the reference coordinates (0, 0) and is associated with the coordinates on the screen. The above-mentioned detected coordinates are regarded as one set from pen touch to pen up. This unit is called a stroke, and is the image of the input handwriting.
Equivalent to.

The input handwriting detection unit 125 is a coordinate detection unit 12.
The input strokes detected by 0 are regarded as the first stroke, the second stroke, and the third stroke from the earliest input time, and are sequentially stored in the input handwriting storage unit 127. Further, the input handwriting detection unit 125 counts the timeout after pen-up,
If a time-out is detected, the input stroke group is made into one character input handwriting.

The recognition unit 130 is provided with a recognition dictionary 140 and determines which character the input handwriting corresponds to. The appraisal information for the appraisal includes, as shown in FIG. 3, the number of strokes, the degree of bending of the first stroke, and the second stroke, a10, a20, a3.
0, a40, direction b10 between the start points of the first and second strokes, direction c10 between the start points and end points, etc. n types of appraisal criteria with different dimensions to be compared (n = 1,2,3 ...) are prepared. However, the appraisal results using these appraisal information are all expressed in scalar quantities. Furthermore, the feature of the input handwriting is expressed by the spatial coordinates (x1, x2, x3 ..., xn) with this scalar quantity x1, x2, x3 ... xn as the reference axis component (the above space is It is called a feature space.
The feature amount described in the related art is expressed by this spatial coordinate. ).

FIG. 4 shows a three-dimensional feature quantity space. In this figure, the x-axis coordinate is the number of strokes, the z-axis coordinate is the degree of bending of the second stroke,
The y-axis coordinate is the direction between the start points of the first and second strokes. In the recognition dictionary 140, the characteristic amount that each character in the dictionary can take is stored in the same dimension as the characteristic amount of the input handwriting in the format shown in FIG. This is the characteristic amount of the monitor handwriting collected with about 100 people as a guide, and is regarded as the handwriting of the general public. Each feature stored in this way is
Similar to the input feature amount described above, it is expressed in the feature amount space. After expressing the feature quantity by coordinates, the distance to the spatial coordinates of the input handwriting is calculated for each dictionary character, and the shortness of this distance is taken as the certainty factor of the dictionary character. The certainty factor is a degree indicating how much confidence the appraisal result can bring, and is synonymous with the coincidence factor. Incidentally, the certainty factor matching degree can be set to a value more suitable for recognition by further increasing or decreasing a value resulting from a special sub-process for determining similar characters.

Characters in the dictionary are classified by comparing the certainty factors, and the ones having the certainty factors satisfying a predetermined criterion are nominated. After nominated, get the character code of the nominated character. In the example of FIG. 4, the feature quantities x1 and x2 of the characters “re” and “wa” are represented by spatial coordinates. The certainty factor for the input feature amount p1 is the distances d1 and d2 in this figure.
It is represented by. The details of handwriting recognition are described in various documents such as "Online Handwritten Kanji / Hiragana Recognition", Technical Report of IEICE, PRL-82-37, and the description is omitted here.

The recognition confirmation input section 150 expands the display font ROM 160 with the character code of the nominated character,
A font corresponding to the nominated code is displayed and output on the display screen of the display-integrated tablet 110, and after the output, the user selects a character intended to be input. If the user makes a selection, the character code of the corresponding character is inserted into the memo buffer 170. An example of the display by the recognition confirmation input unit 150 is shown in FIG. By the recognition confirmation input unit 150, on the screen of the display-integrated tablet 110, a handwriting image h200 in which the input handwriting is imaged, and display fonts f100 and f1 in which the codes of each nominated character are fontified.
10, f120, f130, and rank messages m100, m11 indicating the rank of the certainty factor of each nominated character.
0, m120 ... Is displayed. While the display is being performed in this manner, the frames w1 and w surrounding the character fonts are displayed.
The input intention character is selected by designating the inside of 2, w3 and w4 by pen touch.

The memo buffer 170 includes a recognition confirmation input unit 1
It is a non-volatile memory in which the character code input from 50 is accumulated. The dictionary registration unit 180 uses the recognition dictionary 140.
The feature amount of the input handwriting is reflected in the item of the input intention character in. In the example of FIG. 4, when the input feature amount p1 is regarded as “wa” and is made into a dictionary, as shown in FIG. 6, the spatial coordinates of the input feature amount p1 are also the features of the character “wa” as in the feature amount x1. Quantity x
The recognition dictionary 140 for recognizing the character "wa"
Be registered with. When the feature amount reflection for the input feature amount is repeated in this way, the space character of one dictionary character can take a plurality of space coordinates, such as x4, x5 ... In FIG. Each of these is a target of distance calculation in the above-described certainty factor calculation.

[0025] The dictionary conversion confirmation section 190 confirms that the input handwriting is
If it is a habit character that is erroneously recognized as another character, it is suggested that the reflection of the feature amount of the input handwriting specializes in the recognition dictionary, and asks whether the reflection should be done. As shown in FIG. 7, the confidence factor k1 of the maximum confidence factor character is k1.
Is higher than the others, but the character c1 of the second place or higher is selected by the user, the above presentation is performed.
FIG. 8 shows an example of the presentation display by the dictionary-based gambling confirmation section 190. In this presentation display, despite the maximum confidence,
A display font c1 of a character that is not selected as an input intention character, an information message s230 that warns the user of the possibility of confusion recognition due to recognition dictionary specialization, and a selection icon that receives a pen touch input as to whether or not to reflect a feature amount s200 and s210 are displayed.

Here, if the user merely makes a mistake in pen operation and thinks that the recognition stability is deteriorated and the input feature quantity should not be reflected, and the selection icon s210 is touched with the pen, the feature quantity is not reflected. The process ends. Further, even if the user guides a recognition error, it is unavoidable because this is his own handwriting, and when the user touches the selection icon s200 with a pen, the feature amount is reflected and the process ends.

When the certainty factor of the maximum certainty factor is less than a certain reference value, the feature amount is reflected without displaying the presentation screen of FIG. This is because the reflection of features does not affect recognition stability. In the control ROM 200,
A control program for the electronic memo device 100 is recorded. The flow chart of this program is shown in FIG.

The CPU 210 controls the processing of the electronic memo device 100 according to the control program in the control ROM 200. The recognition input operation of the electronic memo device 100 will be described below. When the user activates the electronic memo device 100, the handwriting input frame s1 is displayed on the display-integrated tablet 110.
20 is displayed. When the user runs the pen tip on the input frame, the drag track is displayed as a handwriting image. The input strokes detected by the coordinate detection unit 120 are considered to be the first stroke, the second stroke, and the third stroke from the earliest input time, and are sequentially stored in the input handwriting storage unit 127 (step S300). After the storage, the recognition unit 130 evaluates which character the input handwriting corresponds to. As a result of this judgment, the input feature amount is expressed in the feature amount space coordinates, and the feature amount that can be recognized by each character, figure, or symbol is expressed in the feature amount space like the input handwriting. After expression, for each character in the dictionary, calculate the distance to the spatial coordinates of the input handwriting,
This distance is regarded as the certainty factor of the character in the dictionary. After calculating the certainty factor, the characters in the dictionary having the certainty factor satisfying a predetermined criterion are nominated (step S310). After the nomination, the recognition unit 130 is caused to acquire the character code of the nominated character. After the acquisition, the recognition confirmation input unit 150 expands the display font ROM 160 with the character code of the nominated character, and displays the font corresponding to the nominated code on the display screen of the display-integrated tablet 110 (step S3).
20), the selection of the input intention character is accepted (step S
330). If the user makes a selection, the character code of the corresponding character is inserted into the memo buffer 170.
In the above selection, when the second and subsequent characters are selected by the user despite the higher certainty factor of the maximum certainty factor character than the others (step S340), the certainty factor of the first most significant character is a constant criterion. Dictation confirmation unit 190
To determine (step S350). Here, instead of determining whether or not the certainty factor of the first rank character is equal to or greater than a certain reference value, for example, it is determined whether the difference between the certainty factor of the first rank and the certainty factor of the second rank is a certain value or more. It may be set as If it is determined that the certainty factor is equal to or higher than the reference value, the display is performed and the user is warned of the risk of confusion recognition (step S
360).

Here, it is considered that the user has only made a mistake in pen operation and should not reflect the input feature amount in consideration of the deterioration of recognition stability, and the selection icon s210 shown in FIG.
When is touched with a pen (step S370), the process ends without reflecting the feature amount. Further, even if the user introduces a recognition error, it is unavoidable that this is his handwriting, and when the user touches the selection icon s200 with a pen, the feature amount is reflected and the process ends. Furthermore, if the input handwriting is terrible habit and the certainty factor of the misrecognized character is less than a certain reference value, the presentation screen shown in FIG. 8 is not displayed and the input feature amount does not affect the recognition stability. The feature amount is reflected (step S38).
0).

As described above, according to the present embodiment, by displaying and warning the fear of recognition confusion based on the certainty factor, how much the input feature quantity to be reflected affects the versatility of the recognition dictionary. This can be presented to the user.
Therefore, the user can customize the recognition dictionary while considering later recognition stability. In this embodiment, the hiragana character is used as the recognition reference character for convenience of explanation. However, JIS code first level kanji / second level kanji, katakana, romaji / russian / Greek,
It may be any of various characters and symbols such as Arabic numerals, Arabic numerals, descriptive symbols / unit symbols / scholarly symbols, and figures such as polygons, circles, and ellipses.

Although the distance in the feature amount space is used as the certainty factor to determine whether or not recognition confusion occurs, other items can be used as the certainty factor if the input handwriting and the shapes of the characters in the dictionary can be compared. You may use. In the above embodiment, the input intention character is selected only from the nominated character group, but it may be selected from the character other than the nominated character. In this case, it suffices if the auxiliary selection unit is provided with a configuration that enables code input without handwriting input, and this auxiliary selection unit may be, for example, a keyboard that allows code input by key punch, or a memoized character. An input menu display unit that displays a code list and accepts designation of an input intention character by pen touch, and a configuration in which additional functions are added to these can be provided.

Further, in the above embodiment, the user confirms the lexicalization bold line by checking whether or not the certainty factor of the maximum confident character exceeds a predetermined threshold value. May use other ones. As an example, first place,
For example, the judgment may be made based on the difference in the second place confidence. Further, in the above embodiment, the online data is used, but it may be offline data read by an optical reader. In this case, the input data is image data, and the recognition is performed by comparing the feature amount of this image data with the feature amount in the recognition dictionary.

[0033]

According to the recognition conversion device of the first aspect,
Since the user is made to judge whether or not the reflection is performed according to the degree of coincidence with the recognition reference character, the user can reflect the input feature amount while considering the recognition stability later.
Therefore, the recognition dictionary can maintain the recognition stability at a certain level or more, and further, the recognition dictionary can be customized so that the habit character and the broken character are made into a dictionary.

Further, according to the recognition conversion device of the second aspect, since the n-dimensional information by the n kinds of appraisal processing is used for the calculation of the degree of coincidence, even if the recognition processing is established from multidimensional appraisal processing, All the appraisal results for one input can be reflected in the recognition dictionary.

[Brief description of drawings]

FIG. 1 is an electronic memo device 100 according to an embodiment of the present invention.
It is a block diagram of.

FIG. 2 is a diagram showing an example of a screen display when handwriting is input.

FIG. 3 is a diagram showing an example of appraisal information.

FIG. 4 is a diagram showing a three-dimensional feature amount space and a feature amount expressed in the space.

FIG. 5 is an example of a display screen when accepting selection of an input intention character.

FIG. 6 is a diagram showing what happens to the feature quantity expression in the feature quantity space as the input feature quantity is reflected in the recognition dictionary.

FIG. 7 is a diagram showing an example of causing recognition confusion in a feature amount space.

FIG. 8 is an example of a screen display presenting recognition confusion to a user.

FIG. 9 is a flow chart showing the control contents of the electronic memo device 100.
It is a chart.

FIG. 10 is a diagram showing an example of appraisal information for each character.

FIG. 11 is a diagram showing an example of recognition confusion.

[Explanation of symbols]

 100 electronic memo device 105 pen 110 display-integrated tablet 120 coordinate detection unit 125 input handwriting detection unit 127 input handwriting storage unit 130 recognition unit 140 recognition dictionary 150 recognition confirmation input unit 160 display font ROM 170 memo buffer 180 dictionary registration unit 190 dictionary conversion control unit 200 control ROM 210 CPU

Claims (3)

[Claims] 1. An input feature amount extraction means for extracting a feature amount which is information peculiar to input data, a recognition reference character which is a character, a symbol, a figure, etc., and a feature amount which the shape of each recognition reference character can take. And a degree of coincidence between the input feature amount and the feature amount of each recognition reference character is calculated, and the degree of coincidence with the input feature amount of the recognition reference character is a predetermined criterion. Nomination means for nominating a recognition reference character satisfying the above condition as a recognition candidate, selection means for allowing the operator to select one of the recognition candidates, and recognition reference characters having a second matching degree or higher in the selection are selected. Determination means for determining whether or not the recognition reference character having the first highest degree of coincidence satisfies a predetermined criterion, based on the degree of coincidence, and the result of the determination, the first highest recognition criterion If the character does not satisfy the predetermined criterion, the first reflecting means for reflecting the input feature amount in the recognition dictionary and, if the first-ranked recognition reference character satisfies the predetermined criterion, warn of the possibility of recognition confusion and input. A warning unit that allows the operator to select whether or not to reflect the feature quantity, and a second means for reflecting the input feature quantity in the recognition dictionary when the operator selects the input feature quantity reflection despite the warning. A recognition conversion device, comprising: 2. The input feature quantity extraction means obtains n-dimensional information as an input feature quantity by the appraisal processing using n kinds of appraisal information for the input data, and the storage means stores each of the n-dimensional information by the appraisal processing. When the shape of the recognition reference character is identified, the n-dimensional information that can be identified is stored as a feature amount, and the nomination means uses the n-dimensional information that is the feature amount as a reference axis coordinate value of an n-dimensional space, The recognition conversion device according to claim 1, wherein the inter-coordinate distance between the input feature amount and the recognition reference character is used as the degree of coincidence. 3. The recognition conversion device according to claim 1, further comprising a sub-selection unit for accepting a selection of a recognition reference character outside the nomination from an operator, wherein the selection unit replaces the determination unit. In the selection, when the recognition reference character having the second highest degree of coincidence is selected, or when the recognition reference character other than the nomination is selected by the sub-selection means, the recognition reference character having the highest degree of coincidence becomes the predetermined reference. The second to determine whether or not to satisfy, based on the degree of agreement
A recognition and conversion device comprising: