JPS5921075B2 - Braille detection method - Google Patents

Description

Detailed Description of the Invention The present invention optically detects the concave and convex pattern of Braille, which is a necessary technology when reading Braille paper for the blind and inputting it into an information processing machine such as a computer. It concerns a method of extracting only necessary information.

As shown in Figure 1, Braille is made up of six positions 1 to 6 (this is called a square in dot order terminology), and depending on whether there is a convex point at each position, if there is a convex point, The code is formed as logic ``1'', otherwise logic ``0''.

This is called the Braille code. In the same square, the distance between the center points of 1 and 4 is the distance between horizontal points, the distance between 1 and 2 is the distance between points, and the distance between point 4 and point 1 of the adjacent square is called the inter-cell distance, and the distance between point 3 and point 1 in the row below is called the inter-row distance. Maximum 32 squares in one line, 1
It is standard to have 17 to 24 lines of information on a page.

In this way, the amount of information that can be put on a single sheet of Braille paper is limited by the size of the Braille squares (which affects how easily they are readable), so it is possible to make the most of the limited space available. Double-sided Braille paper has become popular for use in As shown in Figure 2, double-sided Braille paper uses the line spacing to print dots on the back side, so that the information that can be stored on one sheet of paper is approximately 1.5 times that of single-sided Braille paper.

Double-sided Braille paper has raised dots 21 and 22 (raised dots on the back side) arranged alternately on each line, but when reading by touching with fingers, the raised points 22 are not felt, so there is no problem. However, when reading by optically detecting shadows, it is not possible to distinguish between the convex points 21 and 22, and double-sided Braille paper cannot be read using conventional optical reading methods. The present invention
To provide a method that can optically detect unevenness, distinguish point unevenness using light shadows and highlights, and read both single-sided and double-sided Braille paper without distinction.

This will be explained in detail below using examples. The dots on the Braille paper created by the Braille creation tool are convex dots on the front side and convex dots on the back side.

As shown in FIG. 3A, when the convex point 22 is irradiated with light 26 obliquely from the upper right, a shadow 21 is created in the part where the light does not hit, and a highlight part 28 is created in the part where the light hits almost perpendicularly. Figure B shows the illuminance level of each part.
The level above the standard level (the flat surface of the Braille paper) is the bright highlight area, and the level below it is the dark shadow area. As shown in FIG. 3C, when the convex point 21 is irradiated with light from the same direction, a shadow portion 27 and a highlight portion 28 are generated. Figure D shows the illuminance level of each part when the convex point 21 is irradiated with light. Figure B. ! Comparing l:D, the highlight part 28 and shadow 27 in the direction of light irradiation.
It can be seen that the process is reversed. The present invention utilizes this property to distinguish between the convex dots 21 and concave dots 22, and is devised so that double-sided Braille paper can be read. Next, it will be explained in detail. Braille information is whether there are dots or not.
Since it is information of (existence) and 0 (absence), when handling such information, it is easier to handle it all digitally.

First, the scanning method and sampling method for inputting data will be described.

In scanning, the short side direction of the Braille paper is the main scanning direction, and the long side direction is the sub scanning direction. This scanning method is the same as reading Braille with a finger, and the subsequent recognition process is simplified. Furthermore, since the size of the dots was approximately 1.2 to 1.50 m, the scanning density was set to 3 lines/m. Normal points have a width of three or more scanning lines, and you can also pick up crushed points. Sampling is similarly performed at a density of 3 lines/Mn in the main scanning direction. This value was selected as a value that allows the distance between horizontal points and the distance between cells to be distinguished by the difference in the number of sampling points. When the scanning method and sampling method are determined in this way, the Braille paper can be thought of as being partitioned into grids with an interval of 1/311I [1/311I] in length and width, and the size of dot shadows and highlight areas is expressed by the number of grids. The interval between dots or between squares can be expressed numerically as the number of sample points or scanning lines.

This will be explained below using numerical expressions. To simplify the explanation, consider a double-sided Braille paper with dots in all positions as shown in FIG. 5.

This pattern is difficult to distinguish, so the generality of the explanation is not lost. Let's consider the shadows and highlights created by irradiating this Braille paper with light from about 4" diagonally above the paper in the directions shown in FIGS. 4 and 5. 6th
Figure A corresponds to one vertical column of information surrounded by a broken line, which is sampled and read by scanning the double-sided Braille paper in Figure 5. The dotted portion 61 corresponds to a shadow), and the diagonal lined portion 62 corresponds to a highlighted portion. Numbers a to h are given at the left end of the figure for convenience and to distinguish between concave and convex points. As is clear from this figure, shadows and highlights appear alternately for both concave and convex points, but the order of appearance is different. Then, in the part where the concave point changes to a convex point (concave point c and convex point d), the highlights of both points are dull, and in the part where the convex point changes to a concave point (concave point f and concave point g), both points are highlighted. The shadow of the dot is falling. If only convex or concave points can be cut out from such a pattern, double-sided Braille paper can be read and recognized in the same manner as single-sided Braille paper. Next, we will explain step by step how to cut out only the concave points.

Of course, it is also possible to cut out only the convex points using a similar method. (1) Increase the highlight information by one scanning line in the sub-scanning direction.

(Fig. 6B) It depends on the irradiation angle of the light on the paper,
The highlight information 62 is thinner in the vertical direction (sub-scanning direction) than the shadow information 61.

If a point is damaged, as can be easily imagined from FIG. 3, it may be difficult to output highlight information. Therefore, first, preprocessing is performed to compensate the highlight information 62 so that it has the same size as the shadow information 61. This can be done by logically ORing the information one scanning line before and the information picked up in the current scanning line as new information. (2) Shadow information 61 is delayed by four scans.

(6th C) The size of the point is 1.0 to 1.
Since the distance is approximately 5 Im, the shadow information 61 and highlight information 62 of one point are also detected separated by approximately the same distance as the diameter of the point.

(When the light is irradiated from around 40 degrees) This distance corresponds to 3 to 5 scanning lines, so if the shadow information 61 is delayed based on 4 scanning lines, the shadow information 61 of the concave point and the highlight The information 62 is provided at almost the same timing, but the shadow information 61 of the convex point and the illumination information 62 are increasingly separated from each other. However, at a convex point, there is an inconvenience that the shadow information of the previous point and the highlight information of the point below overlap. (
In Figure 6 B and C, there is a temporal overlap between the shadow of point d and the highlight of e, and the shadow of point e and the highlight of point f)
For small points, the delay of four scan lines is too large, but it is compensated for by the preprocessing described in 1.

Although the delay is insufficient for large points, since the points are originally large, both the highlight information 62 and the shadow information 61 are large, so a slight deviation does not pose a problem. (3) Perform a logical product of shadow information and highlight information.

Shadow information 61 and highlight information 6 are obtained by the processing described in 2.
Since the temporal timings of 2 are now almost equal, if we calculate the logical product of both information, we will no longer have the same dots that occurred in Figure 6 A (7) c and D, and f and g.) , are separated without significantly damaging the shape of the points.

(See Figure 6D) The concave point was separated by the above operation, but as shown in points E and f in Figure 6D, the shadow of the point immediately before the convex point and the highlight of the point below Unnecessary points caused by overlapping are also separated at the same time.

The reason for this is that the distance between the vertical points mentioned in Figure 1 is 2.0~2.
．． 5[11n corresponds to 7 to 8 low scanning lines, and since the distance is twice the distance between the shadow information and the highlight information, if only the shadow information is delayed by an additional 4 scanning lines, the lower point This is because it overlaps with the highlight information. Next, I will explain how to remove this unnecessary point. (4) Eliminate unnecessary points according to rules.

Since point information always includes shadow information and highlight information as a pair, unnecessary points can be determined using rules using both pieces of information.

The method will be described with reference to FIG. In FIG. 7, the shadow information is modeled as 9 for highlight information, and the logical product of both information is modeled as . Of course, these pieces of information have a complicated form as shown in FIG. 6, but for convenience of explanation, they are simply modeled, and the preprocessing shown in FIG. 6B is omitted. FIG. 7A shows the result when light is irradiated to convex points (two points), and the information is arranged in the order of highlight 71 and shadow 742 from the top.

If only the shadow 72 is delayed by 4 lines, the highlight 71 of the lower point and the shadow 72 of the upper point overlap, and when the logical product is taken, an unnecessary point 73 occurs. 1 upward from the center of point 73 that occurred at this time
There must be highlight information between the 1st scan and the 12th scan. On the other hand, as shown in FIG. 7B, in the case of a concave point, the shadow information 72 appears first, then the highlight information 71, and only the shadow information 72 is delayed by 4 lines and is ANDed with the highlight information 71. There is no highlight information between 1 and 12 scans upward from the center of the generated information 73. Therefore, by appropriately selecting the values of 1 and 112, it is possible to distinguish between necessary points and unnecessary points.
Since the distance between vertical points is 2 to 2.5nI01, 1
,=6, and 12=8. 7th using this method
Let us consider a case where a plurality of convex points (in the figure, three convex points, P, , P2, and P3) are consecutive as shown in Figure C. In this case, first of all, there is a highlight 71P1 in the 1st to 12th scanning questions on the point information 73P1P2 created by the AND of the shadow 72P of P, and the highlight 71P2 of P2, so this point 73P,P
Delete 2. Shadow of P2 72 Highlights of P2 and P3 7
Since the highlight information 71P2 generated by the point P2 is also detected in the upper 11-12 scans of the point information 73P2P3 generated by the logical product of 1P3, the point 73P2P3 is similarly erased. In this way, unnecessary points are deleted one after another.
Next, consider the case where a plurality of concave points are lined up in succession as shown in FIG. First, on point information 73Q1 created by the AND of Q1's shadow, 72Q, and highlight 71Q,
Since there is no illite in the 1st to 12th scans, the point information 73Q1 generated by this logical product is not erased, but the highlight information 71Q of Q is erased. Point information 73 generated by the logical product of Q2's shadow 72Q2 and highlight 71Q2
Originally, there was a highlight 71Q1 of Q1 in the upper 1st to 12th scans of Q2, but it no longer exists because it has already been erased. Therefore, the information 73Q2 resulting from the logical product of the shadow 72Q2 of Q2 and the highlight 71Q2 is not deleted, but the highlight information 71Q2 of Q2 is deleted. Point Q3
Similarly, the point 73Q3 generated by the logical product is not erased. In this way, in the information generated by the logical product of shadows and highlights, unnecessary information is erased but necessary information is not deleted, so the point information generated by the logical product is information only for depressed points. This means that concave points and convex points can be distinguished. Next, (1) to (4) mentioned above.
A circuit configuration for realizing the method will be explained.

Figures 8 and 9 show its configuration. First, we will explain step by step starting from Figure 8. As described in (2), S, to S4 are data buffers for delaying the shadow information by four scans, and are memories composed of four shift registers. Each memory stores information for one scan as a digital signal,
Each time the sampling clock is input, the data is shifted one bit at a time. A delay of exactly four scans is obtained from the input of S, to the output of S4. H1 to Hl2 are delay memories for illite information for 12 scans), S1
- The configuration is similar to S4. The or gate 30 is the above 1
This implements the preprocessing described in , and expands the highlight point information in the sub-scanning direction by ORing the input signal of one highlight information and the highlight information from one scan ago. . The AND gate 31 is a circuit that calculates the AND of the highlight signal subjected to the preprocessing described in (3) and the shadow signal delayed by four scans. The circuit block 32 is called an erase buffer and is a memory composed of six shift registers, each of which stores 8 bits of information.

Each memory is connected in series with delay memories H, .about.H, and is driven by the sampling clock+8 clocks. The contents of the erase buffer 32 can be reset by an external erase signal. 33
is a buffer memory for detecting highlight information, and each memory has three 8-bit shift register configurations.
Connected to O-Hl2.

The central 4 bits of each 8-bit memory are OR gates 34
If any one bit is logic 1, the output of the OR gate 34 (highlight information detection signal) is obtained. In FIG. 9, P, .about.P, are memories having a shift register configuration, which are connected to a cascade and driven by a sampling clock.

This memory is the output signal of gate 31 in FIG. Logical product data of shadow information and highlight information is stored in each memory for one scan, for a total of seven scans. P, ~P
, are connected to memories M1 to M, each having an 8-bit shift register configuration. : Configuring the matrix memory 40. Assuming that each bit of the 8-bit memory group is Mij (1=1-7, j?1-8) as shown in the figure, these bits form a matrix as shown in FIG.

This matrix scans the paper surface while moving to the right in one clock cycle and moving downward in each scan. That is, within this matrix, time information is converted to plane information. The point information is of a size that is completely included in this matrix, so when scanning the paper with this matrix, the point information generated by the logical product of shadow information and highlight information will always be located at the center of the matrix. There is a point in time when it is located at . FIG. 11 shows the situation at that time, and A. is a case where the size of the point information is an even number horizontally x an odd number vertically, and B is a case where the size of point information is an odd number horizontally x an even number vertically. In the latter case, center on the left side) bottom side). Reference numeral 41 denotes a circuit that outputs an output signal when the point information is located at the center, and is composed of a gate circuit.

The AND gate 42 does not output an output when the highlight information detection signal, which is the output of the OR gate 34, is logic 1, and outputs an output when it is logic 0. The fact that the highlight information detection signal is logic 1 at this point means that the point information generated by the logical product of the shadow signal and the highlight signal is located at the center of the matrix, as explained in FIG. 7A. At this time, this corresponds to the highlight information being located 6 to 8 scans above the center point, so the circuit 42 does not output point information.
Furthermore, since the highlight information detection signal being logic 0 at this point corresponds to the state shown in FIG. 7B, a point output signal is output. This signal is then used to erase the contents of the erasing buffer 32. As explained with reference to FIG. 7D, this is a circuit that realizes the logic of erasing highlight information when the center of a point is confirmed and is a necessary point. Since the erasing buffer 32 is composed of a matrix of 8 horizontal by 6 vertical, it is larger than the highlight information and is completely erased. So far, we have described the reading method for double-sided Braille paper, but it is clear in principle that single-sided Braille paper, which has only concave dots, can be read using exactly the same method.

By the method described above, only the point information (the center) of the concave point was obtained as the output of the AND gate 42.

By processing the output signal of the AND gate 42 and reading it into a computer, Braille can be recognized as the code explained in FIG. An example of the processing method is shown in Figure 12.
This will be briefly explained with reference to FIG. 50 stores the center information of the points that come in one by one for one row and creates a horizontal line (point sequence 1, 4, point sequence 2, 5, and point sequence 3, 6).
This is a circuit that aligns the

Although FIG. 13 shows one square, the points input to the alignment circuit 50 are not aligned in the horizontal and vertical directions as shown in FIG. 13A. The alignment circuit 50 horizontally aligns the elements A in FIG. 13 as shown in B in the same figure. 51
13B is a circuit that further aligns the circuits shown in FIG. 13B in the vertical direction as shown in FIG. 13C.

Reference numeral 52 is a microcomputer that processes data arranged both vertically and horizontally as shown in Figure 13C.
, 2, 3 dots or 4, 5, 6 dots) are read one by one, and the data is separated into blocks according to the distance between the horizontal dots and recognized as a Braille code.

By converting it into a Braille code in this way, it can be converted into kana and applied to general information processing machines to process a wide range of Braille information. Finally, to summarize the features of the present invention, Braille paper can be read and processed at high speed by the KO I optical method.

11 Both double-sided Braille paper and single-sided Braille paper can be read by the same method without distinction.

Since both the 111 shadow information and the 1/illight information are used, noise for which no highlight information exists (for example, pencil doodles, black dust, etc.) can be removed, improving the recognition rate.

A high recognition rate can be obtained with a compact device using a JV microcomputer.

(In the example, 99.9% was applied to new double-sided Braille paper.
The above reading recognition rate was obtained.

)

[Brief explanation of drawings]

Figure 1 explains the Braille code, Figure 2 explains double-sided Braille paper, Figure 3 shows the principle of optical detection of dots, and Figure 4 shows the scanning method of reading and the use of light. Figure 5 shows the irradiation direction, Figure 5 shows the arrangement of dots on double-sided Braille paper, and Figure 6 shows the results of optically detecting a part of Figure 5 and how to process them. , Figure 7 illustrates the principle of distinguishing between concave and convex points, Figures 8 and 9 illustrate the configuration of a circuit that reads and processes Braille characters, Figure 10 illustrates the configuration of a matrix, and Figure 11 illustrates the configuration of a circuit that reads and processes Braille characters. The figure shows how to find the center of a point in the matrix, and Figure 12 shows a block diagram of how to perform recognition by preprocessing.
FIG. 13 illustrates the operations performed in each block of FIG. 12. 21...Convex point, 22...Concave point, 26.
...Light, 27...Shadow, 28...
Highlight, 61, 72... Shadow information, 62, 7
1...Highlight, 30,34...Or gate, 31,42...And gate, 32
...Buffer for erasing, 33...Buffer memo 18 40...Matrix memory, 4
1...Gate circuit, 50, 51...Alignment circuit, 52...Microcomputer.

Claims (1)

[Claims] 1. Light is irradiated onto Braille paper from an oblique direction and scanned sequentially, and the shadows and highlights occurring on the Braille are converted into time-series electrical signals, and the shadow signal or highlight signal of this electrical signal is Delay one of the signals for a predetermined time and extract the logical product with the other signal, and when there is no signal that is not delayed at a position a predetermined number of scans before the center of the logical product, the logical product is converted into Braille information. A braille detection method characterized by detecting as. 2. The braille detection method according to claim 1, wherein the highlight information is increased in advance by the number of scans in the sub-scanning direction. 3. The braille detection method according to claim 1, wherein the light irradiation angle is approximately 40 degrees and the delay amount is several minutes for four scans.