JPS638886A - On-line continuous character recognizing device - Google Patents

Abstract

PURPOSE:To use position information and to recognize a character which cannot be recognized at a direction angle by making a border value into a position origin at the time of matching an input character string and a standard pattern. CONSTITUTION:The device is composed of a normalizing part 270 to execute the normalization of the position information of an input pattern with the border between characters, a standard pattern memory part 130 to store a standard pattern to link a stroke between the characters from the terminal point of the character to the starting point of a next character to an isolated character pattern, a terminal point table 200 to store the terminal point of the isolated character pattern and a recognizing part to recognize the input pattern based upon a character string standard pattern with the position information. At the time of matching the input character string and the standard pattern, the border value when the optimum channel is obtained, in short, the border between the characters of the input pattern is preserved. By making the border value into the origin, the position is normalized concerning one character in the input character string and the position information is used as the distance between points at the time of matching.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to an online continuous character recognition device that recognizes one or more consecutively written characters from the handwriting of characters input online.

(Prior Art) Conventionally, a device that recognizes a string of characters written in succession without the user instructing the end of one character includes, for example, "Japanese Patent Application No. 173421/1983 CV
There was an online continuous character recognition device described in Reference (1) below. Hereinafter, a conventional online continuous character recognition device will be explained using this device as an example.

This device uses continuously written characters, including characters written without breaks, as an input pattern, and recognizes the characters by pattern matching this pattern with a standard pattern stored in the recognition device in advance. .

The input character string is exchanged into a time-series pattern of directional angles ai of line segments constituting the character. Let this time-series pattern be input pattern A. Similarly, the standard pattern for each character is direction angle b
It is held in the device as the time series pattern BC= of j. Here, n represents a character name. This time-series pattern is referred to as a continuous character pattern BC. The continuous character pattern BG consists of an isolated character pattern BI representing a pattern from the beginning to the end of one character and an inter-character stroke 83 representing the inter-character stroke from the end of writing to the beginning of writing.

An end point table Ej (n) = sn (5) storing the end point S of the isolated character pattern BI included in the continuous character pattern BC is held in the device in the same way as the continuous character pattern BG.

One or more of these continuous character patterns BC are connected to synthesize a character string standard pattern for matching with the input pattern. However, only one pattern to be connected at the end is an isolated character pattern BI up to the end point listed in the end point table Ej.

Here, the distance between the first data of input pattern A and the third data of standard pattern BC is determined by the direction b
It is expressed as the angle formed between j. This distance is defined as the point-to-point distance. The inter-pattern distance between the input pattern A and the character string standard pattern is calculated by changing the character string standard pattern to the input pattern on the time axis so that the cumulative value of this point-to-point distance do (i, j) in the time direction is the minimum. Let this be the cumulative value of the inter-point distance d, (i, j) when matched. The combination of consecutive character patterns BC that minimizes the inter-pattern distance between the character string standard pattern synthesized in this way and the input pattern A is taken as a recognition result.

This matching on the time axis can be performed by the DP matching method described in document (1). DP matching with the continuous character pattern BC at time i to the input pattern is the recurrence formula %) j・1,2. ... By calculating for j, the minimum cumulative distance up to time i can be obtained. However, the initial value is g, (0,0) =O, gn(0,j) = o
o (j=1.2°...jo) is given.

At the same time as this recurrence formula calculation, the path value required to trace the combination of standard patterns that later gives the minimum distance between patterns. Regarding (i, j), perform the assignment operation such that (i, j) =. However, the initial value is (0
,0)=0. Route value. (i, j) has time i
, j, when performing the recurrence formula calculation of equation (7), the boundary value where the boundary distance is given, that is, where the starting end of the standard pattern corresponds to the input pattern is stored.

Character name n2 The minimum cumulative distance to time i is J, and the minimum value T for n among (i, +)
=m, i, n Qn(i, J> (9) becomes the boundary distance of the recurrence formula calculation at the next time i+1.

That is, Qn (i, O) = T (10)
becomes. Also, the boundary value of the route at that time is hn(i, O)
= i (11).

Also, the character name nt that satisfies equation (9) and the nt route value at that time. t(i,,+) respectively as boundary character name N
(i) and the boundary route L(i) as follows.

n(i) = nt (12)
L(i) = h (i, Jn[) (
13) In the nt boundary character name n(1) and the boundary route L(i), the character name nt of the standard pattern whose cumulative distance is the minimum at the end J of the standard pattern and the position of the starting edge of the standard pattern are stored. Ru. In other words, it can be seen that the standard pattern of the character name n(i) matches the portion from time L (i) to i.

On the other hand, for the last part of the input pattern A, the minimum inter-pattern distance with respect to the isolated character pattern BI must be determined. Therefore, at the end point i=1 of the input pattern A, the end point 5n of the isolated character pattern BI is determined from the end point table Ej (n) = Ej (n) (14)
Find (7) and perform the recurrence formula operation and assignment operation of the (&) expression as j=1.2. ...This is carried out only for the isolated character pattern BI part of So. As a result, the distance between patterns can be determined as follows.

From the above, the recognition results are obtained as follows. The above-mentioned boundary character name N (i) stores the character name n of the standard pattern whose cumulative distance is the minimum at time i. Also,
The boundary value is stored in the boundary route L u). i
=I, value of boundary character name N (i) nr=N(i) (16)
becomes the character name nr of the standard pattern that matches the portion from time L(i) to I. Next, the value of i is replaced with L(1) and the value of N(i) is found in the same way.

By tracing L(1) until L(i)=0, recognition results are obtained in reverse order of time.

(Problem to be solved by the invention) Conventionally, the distance d between the points of the input pattern and the standard pattern
, (i, j) did not use position information, but instead used direction angle, etc. This is because the boundaries between characters in the input pattern were not known, and the positional information of the input character string could not be compared with the positional information of the standard pattern separated for each - character. However, when recognizing the katakana letters ``ko'' and ``yu,'' the only difference is the lateral position of the end point of the final stroke, and it is not possible to distinguish based on the direction angle.

As described above, conventionally, there were characters that could not be recognized because character position information could not be used. An object of the present invention is to provide an online character recognition device that solves this problem.

(Means for Solving the Problems) The online continuous character recognition device according to the present invention includes an input unit that reads the handwriting of written characters as a time-series human cover pattern; and a normalization of position information of the input pattern between characters. a standard pattern storage unit for storing a standard pattern in which inter-character strokes from the end point of a character to the start point of the next character are connected to an isolated character pattern representing one character; ; An end point table that stores the end point of the isolated character pattern;
A recognition unit that recognizes the input pattern using positional information based on a character string standard pattern in which an isolated character pattern whose end point is the end point stored in the end point table is concatenated at the end of a pattern in which two or more connected characters are connected. It is composed of

(Function) In the online character recognition device of the present invention, when matching an input character string with a standard pattern, the boundary value when taking the optimal route at time i-1, that is, the boundary between characters of the input pattern. (hereinafter referred to as character boundaries). Then, by using that boundary value as the origin, the position of one character in the input string is normalized,
The position information is used as the distance between points during matching.

In this device, the input character string is a time series pattern of the position coordinates (aXH, aye) of the input points that make up the character A=((ax・, a”yH) l i=1.2.・
, I). Let this time-series pattern be input pattern A. Similarly, the standard pattern for each character is held in the device as position coordinates (bx., by,). This time-series pattern is referred to as a continuous character pattern BC. The origin of the position coordinates of these patterns (hereinafter referred to as the position origin) is a character string or
This is the starting point for writing a letter. Continuous character pattern BC includes isolated character pattern I, which represents the pattern from the beginning of writing to the end of one character, and BS, which represents the inter-character stroke from the end of writing to the beginning of writing.Isolated characters included in this continuous character pattern BC. An end point table Ej(n)=80 (21) storing the end point S0 of the pattern BI is held in the device in the same way as the continuous character pattern BC.

One or more of these continuous character patterns BC are connected to synthesize a character string standard pattern for matching with the input pattern. However, only one pattern to be connected at the end is an isolated character pattern 81 up to the end point listed in the end point table Ej.

Here, position coordinates (a
X・, aye ) is normalized as follows. Position coordinates of input pattern A at time K and standard pattern B
When the position coordinates of C at time are optimally correlated, that is, when the cumulative distance to time (k, L) is the minimum, the input pattern associated with the character boundary that is the start of tlXQlX-nBG Find the time H. Normalization is performed by subtracting the position coordinates of input pattern A at time i from the position coordinates at time i. The character boundary that gives the minimum cumulative distance matched up to the end time (k, L) is h = h (k, L) (22
), the position coordinates normalized by the position origin at the time (hereinafter referred to as the normal cover turn) are (ax・−
aXh, ayHayh).

Let J d n (+, j, ni, L) = be the point-to-point distance between the regular cover turn for others at time i and the standard pattern of character name N (bx,, by'!) at time j.

The distance between the input pattern A and the character string standard pattern is such that the distance between the points d, (i, j, to, L) is minimized in the time direction. Let this be the cumulative value of the point-to-point distance dn (i, J, to, L) when the standard patterns are aligned on the time axis. The combination of consecutive character patterns BC that minimizes the inter-pattern distance between the character string standard pattern synthesized in this way and the input pattern A is taken as a recognition result.

This matching on the time axis is performed by the DP matching method. At time i of input pattern A, DP matching with continuous character pattern BG is repeated. However, J, (0, 0) = O, go (0, j) = oo (j =
1.2. , J") for j=1, 2, . . . , J, the minimum cumulative distance up to time i can be obtained.

At the same time as this recurrence formula calculation, the path value (i, j) required to trace the combination of standard patterns that gives the minimum inter-pattern distance is calculated as in the prior art (
8) An assignment operation such as the expression is performed.

Further, ho(i, j) is also used as a boundary value for normalizing the input pattern A at the next time 1+1.

The subsequent recognition processing is performed in exactly the same manner as in formulas (9) to (16) of the prior art, and finally, the recognition result is obtained and the process ends.

As described above, by using the boundary value as the positional origin when matching an input character string with a standard pattern, it becomes possible to use positional information, and it is possible to recognize characters that could not be recognized using direction angles in the past. .

In other words, various positional information can be used as the point-to-point distance used in the present invention.

In addition, as a standard pattern “Patent Application 1986-110743
Continuous character patterns and isolated character patterns described in "No." and "
The present invention can be used in exactly the same manner when using the isolated character pattern and the character spacing pattern described in Japanese Patent Application No. 53397/1983.

In addition to equation (24), the recurrence formula used for DP matching is also found in Nikkei Electronics Magazine, November 7, 1981, issue P, 1.
71-P, 208, various recurrence formulas described in the literature published under the title [Two-stage DP matching for efficiently recognizing continuously uttered word sounds] (Sakoe) can be used.

(Example) An example of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of the present invention.

The standard pattern storage unit 130 contains a continuous character pattern BG corresponding to equation (18), and an end point table Ej200.
, the end point S0 corresponding to equation (21) is stored for each character name n (n=1.2...,N).

Recurrence formula value q. (i, j) and the route value.
A memory 170 and an N memory 180 are prepared.

Similarly, the T memory 210 stores boundary distances, boundary character names, and boundary routes corresponding to equations (9), (12), and (13), and is addressed by time i. N memories 230 and 220 are prepared respectively.

Character string data input from the tablet 100 is converted into time series data of position coordinates in the preprocessing section 110. This time series data is sent to the input pattern buffer 120 and held as an input pattern (aXi, ayi). At the same time, j=t, 2. −
, J and rl=1.2°..., Oo (the maximum value allowed by the memory) is set for N. Similarly, the T memory 210 has T (0)=0, T (i)=oo (
i=1.2.・, I) is set, N memory 180
is cleared by O. Also, the end point I of the input pattern
is sent from the preprocessing section 110 to the end point detection section 260 and held there.

When the above initialization is completed, the control unit 250 changes the time signal i to 1, 2, . . . . , ■ and the output beam, and processing is performed in synchronization with this. At this time i, the character designation signal n from the control unit 250 is 1, 2, . ..., changes like N. The input pattern (aXH) is sent from the input pattern buffer 120 according to the time signal i.

a'i/H) is sent to the normalization section 270. Normalization section 2
At 70, the input pattern is normalized as shown in equation (23) using boundary value signals hl and h2 from the DP matching section 160, which will be described later. Then, the normalization unit cover turns a1·, a2i corresponding to hl and h2 are output to the distance calculation unit 140. At the same time, the standard pattern br! is read out from the standard pattern storage section 130 using the character designation signal n and the standard pattern time signal j! is also sent to the distance calculation section 140. In the distance calculation section 140, the normalization section cover turn a1.

a 2 HtR'JR9, / b, &(D+ tvf
n(D; tc The inter-point distances d1 and d2 are calculated as in equation (24). These inter-point distances di and d2 are sent to the DP matching section 160, and the minimum cumulative distance is calculated for the character name n. .

Next, the operation within one cycle for n will be explained. DP
A detailed configuration example of the matching section 160 is shown in FIG. When the character designation signal n is designated, T (i
-1) is read out to the signal line T2 and gl ・Register 16
01, and i-1 is written in the hl register 1606.
A numerical value will be written.

Next, when the standard pattern time signal j is set to 1,
ql ・The contents of register 1601 are q2 ・Register 1
The contents of the hl register 1606 are transferred to the h2 register 1607. G at the same time
G(n, 1) is read from the memory 170 and written to the Ql register 1601, and H(n, 1) is read from the H memory 180.
, 1) are read out and written into the hl register 1606. At this point, the values held in g2 register 1602 and h2 register 1607 are the recurrence formula (2
4), (8) boundary conditions g(i-1,0) and h (
i-1,0). Further, the values held in the ql register 601 and the h register 1606 are each g (i-1, 1). (i-1, 1).

Thereafter, the calculations of equations (24) and (8) are repeatedly executed while the standard pattern time signal j is sequentially increased. The operation when the standard pattern time signal generally takes the value j will be described below. At this point, ql register 6
01. q2 ・Cll in each register 602
=CI (i-1,j), Q2 =C]o(i
-t, j-1) is also hl register 606. h
2. The registers 607 each have boundary values [11=h,
(i-1,j), h2 =hn (i-1,j-
1) is maintained. The values of hl and h2 are sent to the normalization unit 270 described above, and the position coordinates (aXhl, ayh2) of the input pattern at time i = hl are obtained from the input pattern buffer 120, and the equation (23) is
1 (aX・aXhl, ayi aVhl) and normalizes the input pattern. Similarly for h2, a2
Normalization is performed as (aX·-aXh2.aV;-aVh2). These normalization part cover turns al.

a2i is sent to the distance calculation unit 140 as described above, and each standard pattern br! The distance between points dl, d
2 is calculated.

di and d2 sent to the DP matching unit 160 are added to gl and register 1601. by adders 1603 and 1604, respectively. q2 - The sum g1 + d1 of 01 and C12 held in the register 602.

Q2 +d2 is calculated and sent to minimum value detection section 1608. The minimum value detection section 1604 detects the signals Ω1+d1 and g2.
The minimum value of +d2 is determined and output as the minimum value signal gm. At the same time, this minimum value is g1+d1, (72+d
2, a switch signal C having a value of 1 or 2 is output, respectively.

The minimum value signal gm obtained in this way becomes the calculation result q.(i, j) of the recurrence formula (24) as the signal 5 (12).The signal 5 (12 is also stored in the G memory 7 as a new G(n, j)).
Filled with 0. The multiplexer 1605 is controlled by the switch signal C, and the hl register 1606,
Signals hl and h2 from the h2 register 607 are selected. That is, hl is selected for C=1 and h2 is selected for C=2. As a result, the output signal sh2 is the calculation result ho(i, j) of equation (8). This signal sh2 is written into the N memory 180 as a new value of H(n,j).

When the above process is repeated until j=Jn, C1(i
, J) Tori. (i, J) is obtained.

These signals are output as signals T1 and h, respectively. Thus, the operation of the DP matching section 160 is completed once.

Signal T1 from DP matching section 160 and T memory 21
Signal T2 from 0 is sent to comparison circuit 190 and compared. When TI<72, the comparison circuit 190 outputs a write pulse Wp. As a result, the T memory 210
The signal T1=C], (i, Jn) is Lmet-'J
220 has h=h (i, Jo), and N memory 230 has a character designation signal n T(i), respectively.
It is written as L(i), n(i). On the other hand, when TI>T2, these entry processes are not performed.

Character designation signal n is 1, 2, . ..., N, the same process as above is repeated, and the process at time i ends. As a result, the T memory 210 holds the minimum value for n of <g (i, J') as shown in equation (9). Roughly speaking, this process is repeated at time 1, 2, . ...
, ■ are repeated, but the processing at the end point I is different from the previous one.

The end point detection unit 260 compares the time signal i with the end point ■, and
= I, the end point signal @ is sent to the end point table Ej200.
Send e. When the end point table Ej200 receives the end point signal e, it sends the end point $0 corresponding to the character designation signal n to the standard pattern storage section 130. Standard pattern storage section 1
30 uses this sent one shift to calculate the distance to the distance calculation unit 14.
Let S be the end point of the standard pattern output at 0. In other words, the data for j in the standard pattern are b, (j=
1.2°・-, S). The subsequent distance calculations and recurrence formula calculations are based on j=1.2. ..., S is performed. The subsequent processing is the same as before, and 1 for character designation signal n.
．． 2. . . , N, the processing at time I ends. With the above pattern flowering 11tD is T(
This means that it was found as the value of 1).

Subsequently, a determination process for obtaining a recognition result is started. The determination unit 240 refers to Lm and N(i) from the L memory and the N memory and outputs the recognition result nr as in equation (16). Give the value of i=■ to N memory, then N(i)
The value of is outputted as a result, and then the value of i is calculated and outputted to the memory, and the value of L(i) is 1q.

The value of si(i) is newly given as i to the N memory, and the value of N(i) is output as a result in the same manner. By repeating this process until L(i)-0, the recognition results nr are output in the reverse order of time i.

(Effects of the Invention) According to the present invention, it is possible to discriminate characters based on differences in position coordinates, which was impossible in conventional online continuous character recognition devices, and an online character recognition device with a higher recognition rate can be obtained.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an embodiment of the present invention, and FIG. 2 is a detailed configuration diagram of a DP matching section in the embodiment.

Claims (1)

[Claims] an input unit that reads handwriting of written characters as a time-series input pattern; a normalization unit that normalizes positional information of the input pattern using boundaries between characters; an isolated character pattern representing one character; a standard pattern storage unit that stores a standard pattern that connects intercharacter strokes from the end point of a character to the start point of the next character; an end point table that stores the end point of the isolated character pattern; and a standard pattern that stores the end point of the isolated character pattern; an online recognition unit that recognizes the input pattern using positional information based on a character string standard pattern in which isolated character patterns whose end point is the end point stored in the end point table are concatenated at the end of the pattern; Continuous character recognizer.