JPH0646418B2 - Feature extraction method - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a feature extraction method applied to a character recognition device or the like, and particularly feature extraction for effectively and stably extracting a pattern difference that has only a minute difference. Regarding the scheme.

In the conventional recognition of character / graphics, a method is widely adopted in which strokes are extracted from a character / graphic pattern, and the positions and lengths of the extracted strokes and mutual relationships between the strokes are used for recognition. The method is as follows: (1) Calculate the curvature of a contour point sequence detected by tracing the contour of a character figure, divide the contour sequence with points with large curvature values as dividing points, and combine the divided sequences. The strokes are extracted by combining them, or (2) the character / graphic pattern is thinned to form a skeleton, and the connectivity and skeleton pattern of the skeleton pattern is traced to detect a sudden change point of the angle, etc. The strokes extracted and extracted in (1) and (2) above are identified by extracting geometrical features and the like. In addition, (3) the distance from each side of the character rectangular frame of the character / graphic pattern to the character / graphic pattern is determined, and the protrusion of the character / graphic pattern corresponding to each side of the character rectangular frame is extracted using the distance, Japanese Patent Laid-Open No. 60-230282 also proposes a method of recognizing characters by using the extracted protrusion information of the protrusions.

(Problems to be Solved by the Invention) However, in the method (1), when the character / graphic pattern becomes large and the character / graphic pattern becomes complicated, the processing amount increases and the processing speed decreases. In the method (2), it is necessary to make the character / graphic pattern thin, and there are problems such as pattern distortion and beard due to the thinning, and the subsequent processing is complicated. In the method of (3), the extracted features become unstable when the shape of the input character is deformed, and especially "water"
The effect is significant between similar characters such as "ice" and "large" and "thick".

An object of the present invention is to divide each of (M × N) regions based on the peripheral distribution of a character / graphic pattern without performing very complicated processing such as contour tracing and thinning in the conventional recognition method. On the other hand, by extracting the information about the poles that satisfy a predetermined condition from the outer coordinate series that represents the outer shape of the character / graphic pattern, the pattern is stable even when the character shape changes, and there are only very small differences. The object of the present invention is to provide a feature extraction method that effectively extracts the difference between the two.

(Means for Solving the Problems) The present invention includes a pattern register for storing a pattern obtained by reading and quantizing a character graphic on a medium as an original pattern, and based on the original pattern, the characteristics of the character graphic are determined. In order to solve the above-mentioned problems of the prior art, the feature extraction method for extracting is to be used. (A) A means for detecting a character rectangular frame formed of the circumscribed side of the original pattern stored in the pattern register, and (b) the original Means for projecting the pattern onto desired two axes respectively to obtain respective black bit number distributions, and (c) first, along each axis, the black bit number distribution within a range limited by the character rectangular frame. The respective barycentric coordinates of the black bit number distribution are determined for the respective ranges obtained by dividing the range limited by the character rectangular frame by the respective barycentric coordinates detected until then. The process of making decisions multiple times Means for repeatedly obtaining respective barycentric coordinate series and dividing the inside of the character rectangular frame into (M × N) regions, determining means for the respective divided coordinates based on the barycentric coordinates, and (d) the original pattern Means for vertically scanning the inside of the character rectangular frame from each side of the character rectangular frame and detecting the coordinate series of the first change point from the background part to the character part, (e) based on the change point coordinate series Then, means for calculating the amount of change of each change point, (f) means for detecting a pole point whose increase and decrease change according to the above change amount, and (g) for each of the (M × N) areas Then, a means for extracting the information of the extreme point satisfying a predetermined condition is provided.

(Operation) According to the present invention, since the feature extraction method is configured as described above, each technical means operates as follows.

The means (a) detects the character rectangular frame that is the circumscribed frame of the character figure by scanning the original pattern stored in the pattern register, and outputs it to each means (b) (c) and (d). To do.
The means of (b) creates a black bit number distribution in each direction of the desired two axes (for example, X-axis and Y-axis directions) from the original pattern,
Output to the method of (c). The means of (c) receives the character rectangular frame from the means of (a) and each black bit number distribution from the means of (b), and the center of gravity of the original pattern based on the peripheral distribution with respect to the desired two axes. The division coordinates are detected and the division coordinates are output to the means (g). The means of (d) detects the coordinates of the first change point when scanning vertically from each of the four sides of the upper, lower, left and right sides of the original rectangular frame of the original pattern over the entire range of the above side. By creating a change point coordinate series,
Output to the means of (e). The means (e) calculates the amount of change of each change point on the basis of the change point coordinate series from the means (d) and outputs it to the means (f). The means of (f) detects the pole point where the increase and decrease change based on the amount of change from the means of (e),
Output to the means of (g). The means (g) receives the division point coordinates from the means (c) and the pole point information from the means (f), and satisfies the predetermined condition for each of the above (M × N) areas. We extract the poles and provide them as features for character and figure recognition. Therefore, it becomes possible to effectively and stably extract the pattern difference having only a minute difference, and the problem of the conventional technique is solved.

Embodiments Embodiments of the present invention will be described in detail below with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of a character recognition device to which the feature extraction method of the present invention is applied.

In FIG. 1, 1 is an optical input from a form, and 14 is a character name output of the character recognition device of this embodiment. This character recognition device includes a photoelectric conversion unit 2, a pattern register 3, a character frame detection unit 4, a character projection creation unit 5, a center of gravity detection unit 6, a character frame division point determination unit 7, an outer coordinate sequence detection unit 8, and a polar coordinate sequence detection. It includes a unit 9, a pole counting unit 10, a pole counting region table 11, an identifying unit 12, and a dictionary memory 13.

The optical input 1 is input to the photoelectric conversion unit 2. The photoelectric conversion unit 2 decomposes one character planned area into 128 × 128 pixels and converts each pixel into a binary digital signal (hereinafter referred to as an input character pattern), which has an average size. One character is represented by an input character pattern of about 60 × 60 bits. The pattern register 3 stores the input character pattern in a format capable of reproducing the X and Y coordinates of each pixel in the character planned area, and has a capacity of 128 × 128 bits corresponding to the character planned area.

In the present embodiment, the character frame detection unit 4 detects the circumscribing frame of a character by expressing it in the pattern register 3 by the left end coordinate X _l , the right end coordinate X _r , the upper end coordinate Y _t , and the lower end coordinate Y _b .

The character projection creation unit 5 projects the input character pattern of the pattern register 3 onto two desired axes, that is, the X axis and the Y axis (horizontal direction and vertical direction in the two-dimensional coordinates of the pattern register 3 respectively) in the present embodiment, and black. The distribution of the number of bits is obtained and black bit number distributions SX (x) and SY (y) are created. The black bit number distributions SX (x) and SY (y) are expressed by the following equation.

However, x and y are 0 to 127 in the pattern register 3, respectively.
Where Y _t and Y _b are the upper and lower end coordinates in the Y axis direction of the character frame, X _l and X _r are the left and right end coordinates in the X axis direction, and P (x, y) Means a black bit or a white bit, and P (x, y) for a black bit (significant color)
= 1 and a white bit (background color), P (x, y) = 0.

Fig. 2 shows the black bit number distribution SX (x) for the input character patterns (Fig. 2 (a)) of the Chinese characters "water" and "ice" in (b) and (c).
, SY (y).

The center-of-gravity detection unit 6 determines the entire range X _{l in} the X and Y axis directions of the character frame.
˜X _r , Y _t ˜Y _b and each range divided by the barycentric coordinates detected in the previous process, the barycentric coordinates of each black bit number distribution SX (x), SY (y) of the input character pattern The series X (M _p ) and Y (M _q ) are obtained by dividing the sum of the first-order moments in each range by the black bit sum in that range. However, M _p, M _q is the centroid coordinate numbers assigned in the order of magnitude of the coordinate values, M _p
= 1 to MX (MX is the number of centers of gravity in the X-axis direction), M _q = 1
To MY (MY is the number of centers of gravity in the Y-axis direction). In the following description, the case of detecting seven barycentric coordinates X (M _p ) will be described. First, the range of X-axis direction of a character frame X _l to X _r
As shown in equation (3), the first moment sum of the black bit number distribution SX (x) of the input character pattern is divided by the black bit sum of the range, and the center of gravity coordinate number of 4 becomes Calculate the barycentric coordinate X (4).

Then, each range divided by the barycentric coordinate X (4),
_{X l ~X (4), X} (4) ~X r 2 two barycentric coordinates as the target X (2), determined by the X (6) (4) equation.

Then, the barycentric coordinates detected so far X (2), X (4),
Range divided by the _{X (6) X l ~X (} 2), X (2) ~X (4)
, X (4) to X (6), X (6) to Xr, the four barycentric coordinates X (1), X (3), X (5), and X (7) are _expressed by equation (5). Ask by.

In the detection of the barycentric coordinate Y (M _q ) in the Y-axis direction, if the number of barycentric coordinate MY to be detected is 7, first, the range Y of the character frame is detected.
Black bit distribution S of input character pattern for _{t to} Y _b
Detecting the Y center coordinates Y of the (y) (4), then the range Y _t to Y for 2 minutes character frame center of gravity coordinates (4), Y (4) ~Y b black bit distribution SY respectively as the target ( barycentric coordinate Y (2) of y)
, Y (4) are detected, and the respective ranges Y _{t to} Y (Y (2), Y (4), and Y (6) in which the character frames in the Y-axis direction are divided are detected. 2), Y (2) ~ Y (4), Y (4) ~
Y (6), Y (6 ) ~Y b black bit distribution as target SY
By detecting the barycentric coordinates Y (1), Y (3), Y (5), and Y (7) of (y), a total of seven barycentric coordinates Y (1) to Y (7) are detected.

The barycentric coordinates X (1) to X (7) for the input character patterns (Fig. 2 (a)) of the Chinese characters "water" and "ice" are shown in Figs. 2 (b) and (c), respectively.
Y (1) to Y (7) are illustrated together with the black bit number distribution.

The character frame division point determination unit 7 receives X, which is received from the center of gravity detection unit 6,
The barycentric coordinate series X (M _p ) and Y (M _q ) in each direction of the Y-axis are used as the divisional coordinate candidates, and the barycentric coordinate numbers M _p and M _q are divided unit areas whose number is NX, NY in the character frame of the input character pattern. Divided coordinate series D in association with divided coordinate numbers k _i and k _j
X (k _i ) and DY (k _j ) are determined.

The division format of the division unit area in this embodiment can take four formats of NX = 4,5,6,8 as the division number in the X-axis direction, and NY = 4,5 as the division number NY in the Y-axis direction. , 6,8
The division coordinate number in the X-axis direction is k _i (k _i = 1 to NX-1, NX
= 4, 5, 6, 8) and the division coordinate number in the Y-axis direction
As k _j (k _j = 1 to NY-1, NY = 4, 5, 6, 8), the divided coordinate series DX (k _i ), DY (which divides the character frame into the divided unit areas NX, NY. k _j ). X,
Divided coordinate series DX by associating barycentric coordinate numbers M _p and M _{q in} each Y-axis direction with divided coordinate numbers k _i and k _j in the X and Y-axis directions.
In order to determine (k _i ), DY (k _j ), a table as shown in Table 1 is prepared, and the number of divisions NX in each direction of the X and Y axes,
Corresponding to NY, from this table, the barycentric coordinate numbers M _p , M
_q is read, and the barycentric coordinates X (M _p ), Y (M _q ) corresponding to the barycentric coordinate numbers M _p , M _q are divided into coordinates DX (k _i ), D.
Determine as Y (k _j ).

In this table, the number of barycentric coordinates MX and MY detected by the barycentric detecting unit 6 is 7, but in a general case, the barycentric coordinates of the number of divisions in each direction of the X and Y axes are included. , And in that case, if there are extra barycentric coordinates remaining, the barycentric coordinates may be added so as to include one more barycentric coordinate in order from the regions at both ends.

FIG. 3 shows a case where the number of divisions NX = NY = 5 is designated as the number of divisions NX, NY in each direction of the X and Y axes.
The divided coordinate series DX (k _i ), DY (k _j ) and the barycentric coordinate series X (M
_p ), Y (M _q ) and the division unit areas (k _i , k _j ) set by the division coordinate series DX (k _i ), DY (k _j ). There is.

The division numbers NX and NY are determined according to the complexity of the input character. If it is rejected once, the number of divisions N
Change X and NY to make character recognition again.

As described above, in the character frame division point determination unit 7, the division format of the division unit area is NX = as the number of divisions in the X-axis direction.
It is possible to adopt four formats of 4, 5, 6, 8 and 4 formats of NY = 4,5, 6, 8 as the number of divisions in the Y-axis direction. In the present embodiment, the number of divisions is NX =
The following description will be given assuming that NY = 8. In this case, in the X-axis direction, the barycentric coordinates X (1), X (2), X (3), X (4), X
Division coordinates DX (1), DX corresponding to (5), X (6), X (7)
(2), DX (3), DX (4), DX (5), DX (6), DX
(7) In the Y-axis direction, barycentric coordinates Y (1), Y (2),
Divided coordinates DY (1), DY (2), DY (3), DY (4), DS corresponding to Y (3), Y (4), S (5), S (6), and Y (7)
(5), DY (6) and DY (7) are determined.

Although the barycentric coordinates and the divided coordinates are associated with each other by using the table in the present embodiment, the two can be associated with each other by executing the arithmetic processing of the flowchart shown in FIG. The results of the division in FIG. 4 are all rounded down to the right of the decimal point.

In FIG. 4, a number M _α is obtained by dividing the number of centroids MX by the number of divisions NX in step S1, and MX is obtained in steps S2 and S3.
The remainder R _{1 of} / NX and R ₂ obtained by dividing R ₁ by 2 are obtained. In step S4, the central value k _{α of the} number of divisions is obtained, and in steps S5 and S6, the division number k _i and the center of gravity number M _p are set to 0. Also, in steps S7, S8, and S9, the division number
Each time k _i is incremented by 1, the previously set quotient R ₂ is incremented by 1
Then, the center of gravity number M _p is increased by M _α . It is checked in step S10 that the quotient R ₂ is not negative, and unless the quotient R ₂ is negative, the number of centroid numbers is incremented by 1 in step S11, and the centroid number M _{p is associated} with the division number k _i in step 12,
The division coordinate DX (M _p ) is determined. If the quotient R ₂ is negative,
In step S13, it is determined whether or not the current division number k _i is larger than the median k _d , and if it is larger, the center of gravity number is increased by 1,
If it is smaller, the division coordinate DX (M _p ) is determined by the center of gravity number set in step S9, and it is detected in step S14 that the division number k _i matches (NX-1), and the process ends.

The outer coordinate sequence detection unit 8 targets the inside of the character frame, and scans the entire range of each of the upper side, lower side, left side, and right side of the character frame in the direction perpendicular to the first black dot. Of the outer coordinate series GT (x), GB (x), GL
Create (y) and GR (y). However, x and y are two-dimensional coordinates of 0 to 127 in the pattern register 3, respectively, and G
T stores the y coordinate obtained by the above process for the upper side, GB stores the y coordinate obtained by the above process for the lower side, and GL stores the x coordinate obtained by the above process for the left side. , GR stores the x-coordinates obtained by the above processing on the right side.

FIG. 5 shows the outline coordinate series from the upper side for the input character pattern of the Chinese character “water”, and FIG. 8 shows the values of the outline coordinate series from the upper side for the input character pattern of the Chinese character “ice”.

The polar coordinate series detection unit 9 detects the ranges X _{l to} X _r and Y _{t to} Y _b in the outline coordinate series detected by the outline coordinate series detection unit 8.
(6) to (9), the points where the value of the equation becomes zero are ignored, and the coordinate series of the polar points where the increase and decrease change (TX (n ₁ ), TY (n ₁ )), (BX (n ₂ ), BY (n ₂ )), (LX (n ₃ ), LY (n ₃ )), (RX (n ₄ ), RY (n ₄ ))) is detected.

dGT (x) = GT (x + 1) -GT (x) x _l ≤ x <x _r … (6) dGB (x) = GB (x + 1) -GB (x) x _l ≤ x <x _r … (7) dGL (y) ＝ GL (y + 1) -GL (y) y _t ≦ y <y _b … (8) dGR (y) ＝ GR (y + 1) -GR (y) y _t ≦ y <y _b … (9) where start point (x _l , GT (x _l )), (x _l , GB (x _l )), (GL (y _t ), y _t ), (GR
(y _t ), y _t ) and end point (x _r , GT (X _r )), (x _r , GB (x _r )), (GL (y _b , y
_b ), (GR (y _b ), y _b ) are unconditionally poles. Where n ₁ ,
n _2, n _3, n ₄ is coordinate number of pole against outer coordinate sequence _{each, 2 ≦ n 1 ≦ Tn,} 2 ≦ n 2 ≦ Bn, 2 ≦ n 3 ≦ Ln, 2
≦ n ₄ ≦ Rn, where Tn is the outer coordinate series GT for the upper side
, Bn is the total number of poles in the outer coordinate series GB for the lower side, Ln is the total number of poles in the outer coordinate series GL for the left side, and Rn
Is the total number of poles in the outer coordinate series GR for the right side. Further, (TX (n ₁ ), TY (n ₁ )) is a polar coordinate series in the outer coordinate series GT with respect to the upper side, and (BX (n ₂ ), BY
(n ₂ )) is a polar coordinate series in the outer coordinate series GB for the lower side, (LX (n ₃ ), LY (n ₃ )) is a polar coordinate series in the outer coordinate series GL for the left side, and (RX (n ₄ ), R
Y (n ₄ )) is a polar coordinate series in the outer coordinate series GR for the right side.

FIG. 6 shows an example of the polar coordinate series from the upper side for the input character pattern of the Chinese character “water”, and FIG. 9 shows an example of the polar coordinate series from the upper side for the input character pattern of the Chinese character “ice”.

The pole counting unit 10 calculates the target pole coordinate series (PTX (pn ₁ ), PTY (p
n ₁ )), (PBX (pn ₂ ), PBY (pn ₂ )), (PLX (pn ₃ )), PLY (pn ₃ )), (P
RX (pn ₄ ), PRY (pn ₄ )) top side, bottom side, left side, SE indicating the type of right side, and the start point division point number S of the target area in the reference side when extracting the polar coordinate series And the end point division point number E (if SE indicates the upper side and the lower side, S
≦ E ≦ NX, S ≦ E ≦ NY when indicating the left side and the right side)
And the area end division point number D (SE indicates the upper side and the lower side of the target area in the direction perpendicular to the reference side, When showing the left side and the right side, And S from the pole point counting area table 11
With respect to the polar coordinate series determined by E, the division point determined by the character frame division point determination unit 7 for the starting point division point number S of the reference side and the character frame division point for the end point division point number E of the reference side A direction toward the reference side with respect to the division point determined by the determination section 7, the reference side, and the area surrounded by the division points determined by the character frame division point determination section 7 with respect to the area end division point number D When n is positive, the number of maximum points and the number of maximum points are n, respectively.
_It is counted as _max (i) and the number of minimum points n _min (i).

However, in this embodiment, the type SE is the polar coordinate series (TX (n ₁ ), TY (n) where SE = 1 and the upper side is the reference side.
₁ )), SE = 2, and the polar coordinate series (BX (n ₂ ), BY (n ₂ )) with the lower side as the reference side, SE = 3
And the polar coordinate series (LX (n ₃ ), L
Y (n ₃ )), and SE = 4, and a polar coordinate series (RX (n ₄ ), RY (n ₄ )) with the right side as the reference side. Further, i represents the number of each set when the type SE, the starting point division point number S, the end point division point number E, and the area end division point number D in the pole counting area table 11 are set as 1 set, and 1 ≦ i ≦ i and _max, i _max is the total number of sets of pole count region table 11. S, E, and D are zero and represent x ₁ or Y _t , and NX or NY are x _r or Y.
Indicates _b .

In FIG. 7, the maximum and minimum points from the upper side of the Chinese character “water” corresponding to FIG. 6 and the pole counting areas for S, E and D, the first
FIG. 0 shows the maximum and minimum points from the upper side of the Chinese character “ice” corresponding to FIG. 9 and the pole counting areas for S, E and D, respectively.

The identification unit 12 determines the number of maximum points and the number of minimum points for each target area counted by the pole counting unit 10, and the dictionary memory 13.
The standard pattern stored in the above-mentioned standard pattern is compared with the maximum point number range and the minimum point number range for the relevant area, and the standard pattern determined to be within the range in all relevant areas in the dictionary memory 13 is stored. The character name of is output to the character name output 14.

Here, the effectiveness of the number of local maximum points and the number of local minimum points, which is one type of the feature information of the feature extraction method of the present invention, is shown below.

For example, in the kanji “water” and the kanji “ice”, when the presence or absence of the upper left point is detected from the upper side, the fluctuation of the two right strokes adversely affects the feature value. Therefore, it is possible to eliminate the above-mentioned adverse effect by limiting the lateral right side range of the pole-counting region by the division points based on the peripheral distribution.
In addition, by limiting the pole counting area in the vertical direction as well, it is possible to eliminate the effects of cutting, blurring, etc. outside the pole counting area, and more stable feature values can be obtained.

Further, by adding the character frame division point determination unit 7 in the present embodiment, the number of pole point counting target areas is reduced, and therefore the processing speed can be increased.

(Effect of the Invention) As described in detail above, according to the present invention, a character / graphic pattern can be formed without performing very complicated pattern processing such as contour tracking and thinning in the feature information extraction of the conventional recognition method. A character frame division point coordinate series for dividing the inside of the character rectangular frame into (M × N) areas is extracted from the barycentric coordinates of the projection on the two axes, and is extracted from the outline coordinate series representing the outer shape of the character / graphic pattern. A feature extraction method is provided that extracts the information of the poles satisfying a predetermined condition from each of the regions and uses them as features to effectively extract the pattern differences that are stable and have only minute differences. it can.

[Brief description of drawings]

FIG. 1 is a functional block diagram showing an embodiment of a character recognition device to which the present invention is applied, FIG. 2 is a diagram showing a relationship between an input character pattern example, a barycentric coordinate series, and a divided coordinate series, and FIG. Is a diagram showing a correspondence relationship between the barycentric coordinate series and the divided coordinate series, FIG. 4 is a flowchart in another method of determining the divided coordinate series, and FIG. 5 is an outer coordinate series from the upper side for the Chinese character “water” of the input character pattern. Explanatory drawing of detection method, FIG.
FIG. 7 is an explanatory diagram of a maximum point / minimum point detection method for the outline coordinate series in FIG. 7, and FIG. 7 shows a maximum point counting area S = 1, E = 6, D = 3.
6 is an explanatory view of the maximum / minimum point counting method of FIG. 6, FIG. 8 is an explanatory view of an outer coordinate sequence detection method from the upper side for the Chinese character “ice” of the input character pattern, and FIG. 9 is an outer contour of FIG. Explanatory diagram of a maximum point and a minimum point detection method for a coordinate series,
FIG. 10 is an explanatory diagram of the maximum point / minimum point counting method of FIG. 9 for the pole counting areas S = 1, E = 6, D = 3. 1 ... Optical input, 2 ... Photoelectric conversion unit 3 ... Pattern register, 4 ... Character frame detection unit 5 ... Character projection creation unit, 6 ... Center of gravity detection unit 7 ... Character frame division point determination unit 8 ... ... External coordinate sequence detection unit 9 ... Pole point coordinate sequence detection unit 10 ... Pole point counting unit 11 ... Pole point counting area table 12 ... Identification unit 13 ... Dictionary memory 14 ... Character name output

Claims (1)

[Claims] 1. A feature extraction method for extracting a feature of a character / graphic based on the original pattern, comprising a pattern register for storing a pattern obtained by reading and quantizing a character / graphic on a medium as an original pattern. ) Means for detecting a character rectangular frame composed of the circumscribed sides of the original pattern stored in the pattern register, and (b) Means for projecting the original pattern on desired two axes to obtain respective black bit number distributions. (C) First, determine the respective barycentric coordinates of the black bit number distribution within the range limited by the character rectangular frame with respect to the respective axes, and then use the respective barycentric coordinates detected until then. The process of determining the barycentric coordinates of each of the black bit number distributions for each range obtained by dividing the range limited by the character rectangular frame is repeated a plurality of times to obtain each barycentric coordinate sequence, and the inside of the character rectangular frame is calculated by (M × N Means for determining respective division coordinates based on the barycentric coordinates for division into individual areas, and (d) scanning the inside of the character rectangular frame of the original pattern in a vertical direction from each side of the character rectangular frame. Then, means for detecting the coordinate series of the first change point from the background portion to the character portion, (e) means for calculating the change amount of each change point based on the change point coordinate series, and (f) Means for detecting a pole whose increase or decrease changes according to the amount of change, and (g) a pole existing in a predetermined area of the (M × N) respective areas among the detected poles. A feature extraction method comprising: a means for detecting the number.