JPH0199190A - Distance arithmetic circuit - Google Patents

Abstract

PURPOSE:To execute arithmetic operation rapidly with a simple circuit constitution by providing an arithmetic part which calculate the distance between recognition object data and dictionary data to output the calculation result in a positive output and a reverse output, a selector part to select the positive output or the reverse output depending on carry from the arithmetic part, and an accumulative arithmetic part. CONSTITUTION:The distance between the invert data of an input recognition object data ID2 and dictionary data ID1 is calculated in the arithmetic part 400a, in case the calculation result (ID1-ID2) is positive, a carry 3 in 'H' is transmitted to the selector part 400b to let is 400b select a positive output '1'. In case the result is negative, a carry 3 in 'L' is transmitted to the part 400b to let is 400b select an inverse output 2, in order that the absolute value ¦ID1-ID2¦ of an output from the part 400b is always secured. And N-pieces of values obtained by repeating the above processing for N-times, are accumulated by the accumulative arithmetic part 400c, and thus one character that is the object of recognition is recognized. In such a way, a distance arithmetic circuit of simple constitution capable of rapid arithmetic operation can be obtained.

Description

[Detailed Description of the Invention] [Summary] Regarding a distance calculation circuit that calculates dictionary data and recognition target data in a character recognition device as an absolute value of a city block distance, there is a distance calculation circuit that can realize quick calculation processing with a simple circuit configuration. The purpose is to provide an arithmetic circuit, and it includes a recognition dictionary section that stores data on feature parameters that character patterns belonging to each character category should have, and a recognition dictionary section that stores data on feature parameters that character patterns belonging to each character category should have, and that extracts dictionary data stored in the recognition dictionary section from feature points of recognition target data. A device comprising a recognition unit that searches and calculates the distance between the recognition target data and the dictionary data to recognize the recognition target data, and the recognition unit calculates the distance between the recognition target data and the dictionary data. An arithmetic section that outputs the positive and inverted outputs of the arithmetic results, and outputs a carry of "H" when the output is positive, and a carry of "L" when the output is inverted, and an input depending on the state of the carry from the arithmetic section. It consists of a selection section that selects a normal output or an inverted output, and an accumulation calculation section that accumulates the results of repeated calculations performed by the calculation section N times to obtain cumulative data for one character.

[Industrial application field]

The present invention relates to a distance calculation circuit that calculates dictionary data and recognition target data in a character recognition device as an absolute value of a city block distance.

With the development of recognition technology such as character recognition, there has been a demand for faster dictionary search processing for recognition.

In order to perform high-speed dictionary search processing for recognition,
It is necessary to configure dedicated hardware to deal with it, but
Since the distance calculation circuit that calculates the distance between the dictionary data and the recognition target data plays an important role in the dictionary search process, it is necessary to achieve faster processing with a simpler circuit configuration.

[Conventional technology]

FIG. 4 is a block diagram illustrating a conventional example, and FIG. 5 is a diagram illustrating an outline of the configuration of a character recognition device.

Normally, a character recognition device has a basic configuration as shown in FIG. That is, a preprocessing unit 1 performs processing to remove various noises (for example, variations in character line width, external noise, etc.) included in the character pattern of the recognition target character and to minimize errors in the recognition process. , a feature extraction unit 2 that detects parameters characterizing character patterns of characters to be recognized, a recognition dictionary unit 3 that stores data on characteristic parameters that character patterns belonging to various character categories should have, and characters of characters to be recognized. The recognition unit 4 includes a recognition unit 4 that performs recognition processing of characters to be recognized by comparing parameters characterizing the pattern with characteristic parameter data stored in the recognition dictionary unit 3.

There are various types of recognition processing performed by the recognition unit 4, such as distance calculation and character recognition based on the accumulated calculation results. This will be done at O.

FIG. 4 shows an outline of the configuration of the distance calculation circuit 30 described above, which includes a comparator (CMP)-41 that compares the dictionary data 101 and the recognition target data ID2, and a comparator (CMP)-41 that compares the dictionary data 101 and the recognition target data ID2. As input, comparator (CMP)
selectors (SEL) 42 and 43 that select one of them based on the output signal from 41;
) 42 and the inverted output of the selector (SEL) 43;
It is provided with an inverter circuit 45 that sends out the inverted output of the adder (ADD) 43, and an accumulation operation circuit 46 that accumulates the outputs of N blocks of the adder (ADD) 44.

The characters to be recognized are divided into N blocks, the dictionary data IDI for each block and the recognition target data ID2 are compared by a comparator (CMP) 41, and the larger one of the comparison result data is selected by a selector (SEL) 42. Select the smaller data (SEL) 43
control to select.

Next, the output of the selector (SEL) 42 is sent as a positive output to the A terminal of the adder (ADD) 44, and the selector (
The output of SEL) 43 is inverted and sent to terminal B of adder (ADD) 44 via inverter circuit 45.

On the other hand, the adder (ADD) 44 processes to obtain the absolute value 1vi Xtl when the CI terminal is "H".

However, it is assumed that Vi = dictionary data and x = recognition target data.

The cumulative calculation circuit 46 calculates the result obtained by repeating this process N times.
Then, cumulative calculation processing (ΣlVt Xil is determined, i-1 to n) is performed, and the recognized output data OD is sent to the next processing circuit (for example, a post-processing circuit), which is not shown.

[Problem that the invention seeks to solve]

In the conventional distance calculation circuit 40 as described above, a selector (SEL) 42 that selects larger data and a selector (SEL) 43 that selects smaller data are used as selectors for selecting input data. This has the disadvantage that the circuit configuration becomes larger accordingly.

An object of the present invention is to provide a distance calculation circuit that can realize quick calculation processing with a simple circuit configuration.

[Means for solving problems]

FIG. 1 shows a block diagram illustrating the invention in detail.

The principle block diagram of the present invention shown in FIG. 1 shows a functional block of a distance calculation circuit 400 having the same function as that explained in FIG.
and the dictionary data 101, and outputs the positive output ■ and the inverted output ■ of the calculation results, and when the positive output ■ is the “H” carry ■, and when the inverted output ■ is the “L” carry ■. 400b is a selection unit that selects the input positive output ■ or inverted output ■ depending on the state of the carry ■ from the calculation unit 400a; 400c is the selection unit that outputs the carry ■ from the calculation unit 400a; This is an accumulation calculation unit that accumulates the results and obtains cumulative data for one character, and by providing these, it is a means for solving this problem.

[For production]

Recognition target data 102 inputted to a distance calculation circuit 400 that performs recognition processing between recognition target data ID2 of each block and dictionary data 101 when one character to be recognized is divided into N blocks by absolute value processing of city block distance. The calculation unit 4 calculates the distance between the inverted data and the dictionary data 101.
00a, and the calculation result is l TDI -10
If 21 is positive, a HII carry ■ is selected by selection section 4.
00b, causing the selection unit 400b to select the positive output ■, and in the case of negative, sends “L” carry ■ to the selection unit 400h, causing the selection unit 400b to select the inverted output ■, and the selection unit 400b selects the positive output ■. The output is always kept at the absolute value +101-102 1, and the value repeated N times is calculated by the cumulative calculation unit 400.
By performing cumulative processing in c and configuring to recognize one character to be recognized, it is possible to realize a distance calculation circuit capable of rapid calculation processing with a simple configuration.

〔Example〕

The gist of the present invention will be specifically explained below with reference to embodiments shown in FIGS. 2 and 3.

FIG. 2 is a block diagram illustrating the present invention in detail, and FIG. 3 is a diagram illustrating a time chart in an embodiment of the present invention. Note that the same reference numerals indicate the same objects throughout the figures.

The embodiment of the present invention shown in FIG. 2 has an adder 410. Incrementer 420
In this example, the selector 400b is a two-manpower/one-output selector 430, and the accumulative arithmetic circuit 400c is the same accumulative arithmetic circuit 46 as explained in FIG. 4.

In this embodiment, one character to be recognized is divided into 8 parts, and therefore, the cumulative calculation of the cumulative calculation circuit 46 is performed using the clock CLK for accumulating 8 blocks of data, and 8 blocks of data.
This is done using the clear CLR that occurs every clock, and the situation is shown in the time chart of FIG.

Incidentally, D1 to D8 shown in FIG. 3 indicate data of eight blocks.

The dictionary data IDI (for example, 111) read from the recognition dictionary section 3 is sent to the A terminal of the adder 410, and the recognition target data 102 (for example, Di) whose feature amount has been extracted by the feature extraction section 2 is sent through the inverter circuit 440. adder 410
are input to the B terminals of each.

Adder 410 calculates IIDI -1021. That is, for example, IDI=2. If ID2 = 1, the calculation result is output as “0001”,
At this time, the carry CO2 output of the adder 410 is "H".

On the other hand, for example, IDI=1. When ID2=2, the result of the operation is output as "1111", and at this time, the carry CO2 output of adder 410 is output as "L".

When the carry CO output of the adder 410 is "H", the selector 430 selects the positive output of the adder 410 and sends it to the accumulation operation circuit 46, and when the carry CO output is "L" In this case, the inverted output (2) via the incrementer 420 is selected and sent to the accumulation calculation circuit 46.

As a result, the output of the selector 430 is always l 101-
The calculation result for ID2 1 is obtained, and this is calculated as the cumulative calculation result for 8 blocks (i.e. city block distance d=Σlv=
Get Xtl. However, output data 00 of one character is output after recognition processing is performed based on i=1 to n).

The cumulative calculation circuit 46 is cleared by the clear CLR every time eight blocks are processed, and the cumulative calculation value is set to zero. By repeating this process as many times as the number of characters in the recognition dictionary section 3, the dictionary is searched and finally one recognized character is obtained.

The above processing speed is achieved by the comparator (CM "P) 4" explained in Fig. 4! ! The delay of r and the delay of incrementer 420 are canceled out, making it possible to perform processing without particularly reducing processing speed.

〔Effect of the invention〕

According to the present invention as described above, it is possible to provide a distance calculation circuit that can perform distance calculations with a simple configuration and without reducing processing speed.

[Brief explanation of the drawing]

FIG. 1 is a block diagram explaining the present invention in detail, FIG. 2 is a block diagram explaining the present invention in detail, FIG. 3 is a diagram explaining a time chart in an embodiment of the present invention, and FIG. 4 is a conventional block diagram. FIG. 5 is a block diagram explaining an example, and FIG. 5 is a diagram explaining an outline of the configuration of the character recognition device. In the figure, 1 is a preprocessing section, 2 is a feature extraction section, 3 is a recognition dictionary section, 4 is a recognition section, 40.400, 400d is a distance calculation circuit, 41 is a comparator (CMP), 42.43.430 is a Selector (SEL), 44.41
0 is an adder (ADD), 45, 440, 450 are inverter circuits, 46 is an accumulation calculation circuit, 400a is a calculation unit, 400b is a selection unit,
400c is the cumulative calculation unit, 420 is the incrementer 4〈Yarito■ Shellfish no Kyo 97 Σ Wrinkle N Zurofu O Tsunozu 1000
1. 式吏謬读绀N楥、 no zue - Be Song Kan）drinking sink hu 4roQ 茅 5
■

Claims (1)

[Scope of Claims] A recognition dictionary unit (3) that stores data on feature parameters that a character pattern belonging to each character category should have;
) is searched for, and the recognition target data (ID2) and the dictionary data (ID1) are searched.
) and a recognition unit (4) that recognizes the recognition target data (ID2) by calculating a distance between the recognition target data (ID2) and the dictionary, the recognition unit (4) calculating the distance between the recognition target data (ID2) and the dictionary. Calculate the distance to the data (ID1) and output the positive output ([1]) and inverted output ([2]) of the calculation results, and carry “H” when the positive output ([1]) (
[3]), and an arithmetic unit (400a) which outputs an “L” carry ([3]) when the inverted output ([2]) is output, and the carry ([3]) from the arithmetic unit (400a). ]) depending on the input positive output ([1]) or inverted output ([2
]), a selection unit (400b) for selecting the
00a) is equipped with an accumulation calculation unit (400c) that accumulates the results of repeated calculations N times to obtain cumulative data for one character, and carries “H” from the calculation unit (400a) ([3]
) is output, the selection unit (400b) selects the positive output ([1]) of the calculation unit (400a), and carries “L” ([3]) from the calculation unit (400a). )
is output, the selection section (400b) selects the inverted output ([2]) of the calculation section (400a), and cumulatively calculates the output of the selection section (400b) into N blocks. A distance calculation circuit characterized by obtaining cumulative data for one divided character.