JP4079665B2 - Search device - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a search device, and more particularly to a search device that searches for a pattern.
[0002]
[Prior art]
With the development of the information society in recent years, the amount of information handled has increased, and in order to realize advanced multimedia services and wider areas, the speed of symbol string (character string) search processing is strongly increased. It has been demanded. In the search process, a specified symbol string called a pattern is searched from a given symbol string called text.
[0003]
In the conventional simplest search algorithm, if a mismatch is detected in the search process, the pattern is shifted by one symbol and the matching is repeated again. As typical algorithms that enable faster processing, there are Knuth, Morris, There is a KMP method proposed by Pratt and a BM method proposed by Boyer and Morre.
[0004]
The basic search operation of the KMP method and the BM method is as follows: when matching pattern symbols (the KMP method matches the pattern from the beginning to the front to the back, and the BM method uses the pattern from the tail to the back to the previous direction. When a mismatch is detected, the pattern is greatly moved based on a certain condition. By performing such a search, the search performance can be improved as compared with a simple method in which a pattern is always shifted by one symbol when a mismatch of symbols occurs.
[0005]
[Problems to be solved by the invention]
In the field of information communication, a pattern having a structure including consecutive identical symbols is often searched. Examples of the pattern in which the same symbol continues include an HDLC flag and a header existing in a stream compressed by MPEG.
[0006]
When a search process is performed on a pattern having such a specific structure by the conventional KMP method or the BM method as described above, there is a problem in that useless symbol matching occurs and the search performance cannot be improved. It was.
[0007]
Even the KMP method and the BM method, which represent fast search algorithms, are not effective for patterns of any structure, and the search performance for patterns having the same symbol string is not necessarily the best.
[0008]
The present invention has been made in view of these points, and an object of the present invention is to provide a search device that performs a search process for a pattern having a structure including consecutive identical symbols at high speed.
[0009]
[Means for Solving the Problems]
In the present invention, in order to solve the above-described problem, the same symbol string recognition unit for recognizing the same symbol string composed of the same consecutive symbols in the pattern in the search apparatus 10 for searching for a pattern as shown in FIG. 11, a symbol intends row matching search processing unit 12 of the pattern and the text, the search apparatus 10, characterized in that it has a provided.
[0010]
Here, search processing unit 12, with a number of 1~n from the head of the same symbol string to the end, from the end of the symbols of the same symbol string, starting symbols matching towards the beginning of the same symbol string When the symbol collation of the symbol k (1 ≦ k ≦ n) among the symbols in the same symbol string has failed , the pattern for k symbols from the failed position to the tail end is obtained. Shift and perform symbol collation again from the last symbol of the same symbol string .
[0011]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings. Figure 1 illustrates the principle of a search device. The search device 10 searches for a pattern from text (or binary data).
[0012]
The same symbol string recognition unit 11 recognizes the same symbol string composed of the same consecutive symbols in the pattern. The search processing unit 12 starts symbol matching (or simply referred to as matching) between the pattern and the text from the last symbol in the same symbol string. If the symbol collation of the last symbol fails, the pattern is shifted by the number of the same symbols, and the symbol collation is performed again to perform the search.
[0013]
For example, if the pattern is “000B”, the same symbol string of this pattern is “000”. When performing a search, the text and the left end of the pattern are combined to start symbol matching with the text from the last symbol “0” of “000”. If the symbol matching of the last symbol “0” fails (the text in the figure is “C”, it does not match “0”), it is shifted by the number of the same symbol string (here, 3 symbols). Then, the symbol collation is performed again. Detailed operation contents will be described later with reference to FIG.
[0014]
Next, an outline of the KMP method and the BM method will be described. FIG. 2 is a diagram for explaining the KMP method. In order to find the position where the pattern appears in the text, first the text and the left end of the pattern are matched, and the text and the text are collated one by one from the beginning (left end) to the back.
[0015]
Then, when a symbol mismatch occurs (○ and △ in the figure), if there is the same partial sequence (thick solid line frame in the figure) on the left side of the pattern from the position, Shift to the right by the size of the figure. The substring is selected with the maximum length.
[0016]
For example, when the pattern “EFGHEF2” is found in the text of the figure, if the matching succeeds up to “EFGHEF” and a mismatch occurs at the last “2”, the substring “EF” to the left of “2” Since the same pattern as "" is also present at the head of the pattern, the head of the pattern can be shifted to the position P1 for the next pattern position.
[0017]
3 and 4 are diagrams for explaining the BM method. In order to find the position where the pattern appears in the text, the text and the left end of the pattern are first matched, and the text and the text are collated in order one by one from the tail (right end) of the pattern.
[0018]
In the BM method, there are two algorithms for how much the pattern is shifted to the right when a symbol mismatch occurs (circles and triangles in the figure). FIG. 3 shows the first algorithm, and FIG. 4 shows the second algorithm.
[0019]
In contrast to FIG. 3, the first algorithm is similar to the KMP method, and if the same partial sequence as the partial sequence (bold solid line frame in the figure) at the position before the mismatch from the right end of the pattern is on the left side, Shift the pattern to the right so that the subsequence overlaps the text subsequence.
[0020]
For example, when the pattern “NEFGHEF2” is found in the text of the figure, if the matching succeeds up to “EF2” and a mismatch occurs at the next “H”, the substring “EF” at the position before “H” Since the same partial sequence is also on the left side of the pattern, it can be shifted to the position where the partial sequences overlap (the beginning of the pattern comes to the position P2).
[0021]
In contrast to FIG. 4, the second algorithm looks at the symbol of the text when there is a mismatch, and when there is the same symbol on the left side of the pattern, the pattern is shifted so that it overlaps that symbol of the text.
[0022]
For example, when the pattern “NKFGHEF2” is found in the text of the figure, if matching succeeds up to “EF2” and a mismatch occurs at the next “H”, the symbol of the text when the mismatch occurs is “K” It is. Since “K” is also on the left side of the pattern, it can be shifted to a position where the pattern and text “K” overlap (the beginning of the pattern comes to position P3).
[0023]
In general, it is known that the search speed of the BM method is higher than that of the KMP method. In the BM method, it is known that the search speed of the second algorithm is higher than that of the first algorithm.
[0024]
Search operation of search apparatus 10 in the following be described in detail while comparing with the conventional BM method for. Hereinafter, the search operation of the search device 10 is also referred to as a pattern search method.
[0025]
Figure 5 is a diagram illustrating a search procedure. The text is a symbol string shown in the figure, and the pattern is “000B”. First, attention is focused on the longest of the same symbol string "000" to be continuous in pattern. Then, the text and the left end of the pattern are combined, and the text is first collated from the last symbol “0” of “000” in the pattern “000B”.
[0026]
If a symbol mismatch is found as a result of collation, there is only the character “0” that failed to collate in the head direction of the pattern. In this case, the symbol is shifted by one character and two characters. It is clear that even if they are matched, they do not match.
[0027]
Therefore, the pattern is shifted by three characters (the number of the same symbol string), and similar symbol matching is repeated again. In the figure, the rightmost character of “000” matches during the third shift. Therefore, next, the text is compared with the second character of “000”.
[0028]
Here, if a mismatch between the symbols is found, it is clear that even if the symbols are matched by shifting one character, it is clear that they do not match. In general, when numbers 1 to n are assigned from the beginning to the end of the same symbol string and the k (2 ≦ k ≦ n) th symbol collation fails, the pattern is shifted by k. It will be.
[0029]
By performing such a search process, the number of shifts can be greatly reduced. In the example shown in the figure, the pattern “000B” is detected in the text by the third shift (the matching order at the third shift position is “0” at the end of “000”, “ The pattern is detected by collating in the order of “0” and leading “0”, and then collating “B”.
[0030]
FIG. 6 is a diagram showing a conventional BM method search procedure. The text and pattern symbol strings are the same as in FIG. The figure shows the case of the second algorithm of the BM method. In the BM method, the search is started from the last character of the pattern.
[0031]
If the symbols do not match, look at the unmatched symbol in the text, and if there is the same symbol to the left of the pattern, shift the pattern so that this symbol is in the position where the mismatch was found. Here, since the symbol in the text that does not match is “0”, the pattern is shifted so that this “0” and the second “0” from the end of “000B” overlap.
[0032]
Such processing is performed from the 0th shift to the 2nd shift, the 3rd shift to the 6th shift. Also, when shifting from the second time to the third time, “A” in the text and “B” in the pattern do not match, and “A” does not exist to the left of the pattern, so the shift from the second time to the third time shifts. When going, the pattern is shifted by 4 characters.
[0033]
As shown in the figure, the BM method requires a total of seven shifts. Thus, in the case of patterns including the same symbol string, it can be seen that the search process by the search device 10 is faster.
[0034]
The search device in the case of introducing search efficiency in the following be described. FIG. 7 is a diagram showing the configuration of the search device. The search device 20 includes a search efficiency calculation unit 21 and a search processing unit 22. Hereinafter, the search operation of the search device 20 is also simply referred to as a search method.
[0035]
The search efficiency calculation unit 21 calculates the search efficiency for each symbol of the pattern. Search efficiency refers to the expected value of the shift amount, which is the sum of the shift amount that can be taken by the shift operation when symbol matching fails and the probability that the shift amount will occur, and the number of symbol matches. It is expressed as a ratio. The search processing unit 22 performs a search by collating symbols in descending order of the search efficiency value.
[0036]
Next, the search efficiency and the operation of the search device 20 will be described in detail. In order to increase the search speed, it is necessary to increase the amount of shift and eliminate unnecessary symbol matching. In this case, it is necessary to provide an index as to which symbol in the pattern is used for collation, and whether the shift amount can be increased when the collation fails. What shows this index is the search efficiency. The search efficiency is the expected value of the shift amount / number of symbol collations.
[0037]
First, a case where search efficiency is introduced into the pattern search method will be described. As a specific example, consider a case where the text is composed of three types of symbols A, B, and C, and the pattern is “AAABCAB”. Further, numbers (1), (2),..., (7) are assigned to the symbols in the pattern from the beginning of the pattern.
[0038]
In the pattern search method, the search is performed by paying attention to the same symbol string. Therefore, the search efficiency of the respective symbols (3), (2), and (1) with respect to “AAA” is obtained. When the search efficiency is expressed as E [(n)] (n is a number given to the pattern), E [(3)] to E [(1)] are expressed by the following equations (E [(3 )] Is the search efficiency for “A” at the end of “AAA”, E [(2)] is the search efficiency for “A” in the middle of “AAA”, and E [(1)] This is the search efficiency for “A” at the left end of “AAA”).
[0039]
[Expression 1]
E [(3)] = (P (A) × 0 + P (B) × 3 + P (C) × 3) / 1 = 2 (1a)
E [(2)] = (P (A) × 0 + P (B) × 2 + P (C) × 2) / 2 = 4/6 (1b)
E [(1)] = (P (A) × 0 + P (B) × 1 + P (C) × 1) / 3 = 2/9 (1c)
P (A) in each expression is the probability that the symbol “A” appears in the text, P (B) is the probability that the symbol “B” appears in the text, and P (C) is the symbol “C” in the text. "Is the probability of appearance. Here, it is assumed that P (A) = P (B) = P (C) = 1/3.
[0040]
On the other hand, the numerator of the equation E [(n)] is an expected value of the shift amount, and is represented by a weighted average using the probability of the shift amount as a random variable as a weight. For example, P (A) × 0 in the formula (1a) matches if the last symbol “A” of “AAA” is collated with text and the symbol of the text is “A”. In this case, since there is no shift operation, the shift amount is zero. That is, when “A” at the end of “AAA” is collated with text, in order for the shift amount to be zero, “A” should appear in the text at that position, and the probability is P (A ).
[0041]
Further, when “A” at the end of “AAA” is collated with text for P (B) × 3 and P (C) × 3 in Expression (1a), the symbol of the text is “B” or “ If it is “C”, it is inconsistent. In this case, the shift operation is performed for three symbols as described above with reference to FIG. That is, when “A” at the end of “AAA” is collated with text, in order for the shift amount to be 3, “B” or “C” should appear in the text at that position, and the probability Are P (B) and P (C).
[0042]
Therefore, since the expected value is the sum of the value of the random variable multiplied by the probability that each value is realized, the numerator (the expected value of the shift amount) of Equation (1a) is P (A) × 0 + P (B) × 3 + P (C) × 3. The same concept applies to formula (1b) and formula (1c).
[0043]
On the other hand, the denominator in each expression indicates the number of symbol matching between the pattern and the text. In the case of “A” at the end of “AAA”, symbol collation is performed first, and therefore 1 in Formula (1a).
[0044]
Similarly, in the case of “A” in the middle of “AAA”, the last “A” is collated and then the symbol collation is performed. The number of symbol verifications between text and text is two. For this reason, it is 2 in the formula (1b). Furthermore, in the case of “A” at the left end of “AAA”, since the last “A” and the middle “A” are collated and then the symbol collation is performed, the left end “A” is collated. For this purpose, the number of symbol verifications between the pattern and text is three. For this reason, it is 3 in the formula (1c).
[0045]
When the magnitudes of the search efficiency values calculated in this way are compared, (1) <(2) <(3). Therefore, the higher the search efficiency, the larger the shift amount can be obtained when the collation fails. Therefore, the values are increased in the order of (3), (2), (1) (“AAA It will be understood that symbol matching should be performed in the order of “tail, middle, left end”.
[0046]
Next, a case where search efficiency is introduced into the second algorithm of the BM method will be described. As an example, the text is composed of three types of symbols A, B, and C, and the pattern is “AAABCAB”, as described above. Also, numbers (1), (2),..., (7) are added from the beginning of the pattern. Retrieval efficiencies E [(7)] to E [(1)] based on the BM method for each symbol of the pattern are expressed by the following equations.
[0047]
[Expression 2]
E [(7)] = (P (A) × 1 + P (B) × 0 + P (C) × 2) / 1 = 1 (2a)
E [(6)] = (P (A) × 0 + P (B) × 2 + P (C) × 1) / 2 = 1/2 (2b)
E [(5)] = (P (A) × 2 + P (B) × 1 + P (C) × 0) / 3 = 1/3 (2c)
E [(4)] = (P (A) × 1 + P (B) × 0 + P (C) × 4) / 4 = 5/12 (2d)
E [(3)] = (P (A) × 0 + P (B) × 3 + P (C) × 3) / 5 = 2/5 (2e)
E [(2)] = (P (A) × 0 + P (B) × 2 + P (C) × 2) / 6 = 4/18 (2f)
E [(1)] = (P (A) × 0 + P (B) × 1 + P (C) × 1) / 7 = 2/21 (2g)
Here, for example, the expression (2c) will be explained. When the fifth “C” from the beginning of “AAABCAB” is collated with the text for P (C) × 0, the text symbol is “C”. If there is a match. In this case, since there is no shift operation, the shift amount is zero. That is, when the fifth “C” from the beginning of “AAABCAB” is collated with the text, in order for the shift amount to be 0, “C” should appear in the text at that position, and the probability is P (C).
[0048]
For P (A) × 2, if the fifth “C” from the beginning of “AAABCAB” is collated with the text and the symbol of the text is “A”, there is a mismatch. In this case, the shift operation is performed for two symbols as described above with reference to FIG. 4 (because “A” in (3) is matched with the mismatched position), the shift amount is 2. That is, when the fifth “C” from the beginning of “AAABCAB” is collated with text, in order for the shift amount to be 2, “A” should appear in the text at that position, and the probability is P (A).
[0049]
Further, when the fifth character “C” from the head of “AAABCAB” is collated with text for P (B) × 1, if the symbol of the text is “B”, it is inconsistent. In this case, since the shift operation is performed for one symbol as described above with reference to FIG. 4 (because “B” in (4) is aligned with the position where there is a mismatch), the shift amount is 1. That is, when the fifth “C” from the head of “AAABCAB” is collated with the text, in order for the shift amount to be 1, “B” should appear in the text at that position, and the probability is P (B).
[0050]
On the other hand, for the denominator of equation (2c), the fifth “C” from the beginning of “AAABCAB” is matched third from the end of the pattern, so the number of symbol matching between the pattern and text is three. It is 3 because it requires. The other formulas have the same concept as described above.
[0051]
Here, the values of the respective search efficiencies according to the equations (1a) to (1c) when introduced into the pattern search method and the values of the respective search efficiencies according to the equations (2a) to (2g) when introduced into the BM method. Comparing the magnitude, the formula (1a) = 2 is the largest, and the formula (2a) = 1 is the second largest.
[0052]
Therefore, when combining the pattern search method and the BM method, the pattern search method first matches the pattern “AAABCAB” from (3), and then uses the BM method (7), (6), (5), ( If the respective symbols are collated in the order of 4), (2), and (1), the search performance can be improved.
[0053]
Next, the case where the pattern search method and the KMP method are combined and the search efficiency is introduced into these will be described. As an example, the text consists of three types of symbols A, B, and C, and the pattern is “ABCABAAA”. Also, numbers (1), (2),..., (8) are added from the beginning of the pattern. For each symbol of the pattern, search efficiencies E [(8)] to E [(6)] based on the pattern search method are expressed by the following equations.
[0054]
[Equation 3]
E [(8)] = (P (A) × 0 + P (B) × 3 + P (C) × 3) / 1 = 2 (3a)
E [(7)] = (P (A) × 0 + P (B) × 2 + P (C) × 2) / 2 = 4/6 (3b)
E [(6)] = (P (A) × 0 + P (B) × 1 + P (C) × 1) / 3 = 2/9 (3c)
Since the concept of each formula in the pattern search method has been described above, a description thereof will be omitted. On the other hand, search efficiencies E [(1)] to E [(8)] based on the KMP method are expressed by the following equations.
[0055]
[Expression 4]
E [(1)] = (P (A) × 0 + P (B) × 1 + P (C) × 1) / 1 = 2/3 (4a)
E [(2)] = (P (A) × 1 + P (B) × 0 + P (C) × 1) / 2 = 2/6 (4b)
E [(3)] = (P (A) × 1 + P (B) × 1 + P (C) × 0) / 3 = 2/9 (4c)
E [(4)] = (P (A) × 0 + P (B) × 1 + P (C) × 1) / 4 = 2/12 (4d)
E [(5)] = (P (A) × 1 + P (B) × 0 + P (C) × 1) / 5 = 2/15 (4e)
E [(6)] = (P (A) × 0 + P (B) × 3 + P (C) × 3) / 6 = 2/6 (4f)
E [(7)] = (P (A) × 0 + P (B) × 1 + P (C) × 1) / 7 = 2/21 (4 g)
E [(8)] = (P (A) × 0 + P (B) × 1 + P (C) × 1) / 8 = 2/24 (4h)
Here, for example, the expression (4f) will be described. When the sixth “A” from the beginning of “ABCABAAA” is collated with the text for P (A) × 0, the text symbol is “A”. If there is a match. In this case, since there is no shift operation, the shift amount is zero. That is, when the sixth “A” from the beginning of “ABCABAAA” is collated with text, in order for the shift amount to be zero, “A” should appear in the text at that position, and the probability is P (A).
[0056]
In addition, when P (B) × 3 and P (C) × 3 in Expression (4f) are compared with the text “A” in the sixth from the beginning of “ABCABAAA”, the symbol of the text is “B”. Or if it is “C”, it is inconsistent. In this case, since the shift operation is performed for three symbols as described above with reference to FIG. 2 (the partial sequence “AB” is located on the left side from the mismatched position, and this partial sequence is also at the top of the pattern, so the three symbols are included. The shift amount is 3. That is, when the sixth “A” from the beginning of “ABCABAAA” is compared with the text, the shift amount should be 3, so that “B” or “C” appears in the text at that position. The probabilities are P (B) and P (C).
[0057]
On the other hand, with respect to the denominator of the formula (4f), when “A” is the sixth “A” from the head of “ABCABAAA”, the sixth time from the head of the pattern is collated, so the number of symbol matching between the pattern and text is six. Therefore, it is 6. The other formulas have the same concept as described above.
[0058]
Here, the values of the respective search efficiencies according to the equations (3a) to (3c) when introduced into the pattern search method and the values of the respective search efficiencies according to the equations (4a) to (4h) when introduced into the KMP method. Comparing the size, the formula (3a) = 2 is the largest, and then the formula (3b) = the formula (4a) = 2/3 is the largest.
[0059]
Therefore, (8) and (7) of the pattern “ABCABAAA” are first collated by the pattern search method, and then (1), (2), (3), (4), (5), If the respective symbols are collated in the order of (6), the search performance can be improved.
[0060]
Description will be given of a case where next to were carried out pattern search method in the software. FIG. 8 is a diagram showing a program for counting the maximum length of the same symbol string. The program 200 counts the maximum length of the same symbol string. For example, 3 is returned when the pattern is “AAAB”, and 4 is returned when “AABBBBCC”.
[0061]
FIG. 9 is a flowchart showing the processing procedure of the pattern search method. Txt indicates text and Pat indicates a pattern.
[S1] The symbol collation start order is stored in the table T [i]. For example, assuming that the pattern is “000B”, numbers (0) to (3) are added from the top of the pattern. Since the search is performed from the end of the same symbol string, T [0] = (2), T [1] = (1), T [2] = (0), and T [3] = (3).
[S2] The shift amount at the time of symbol collation failure is stored in the table S [T [i]]. For example, in the case of the pattern “000B”, if the end of the same symbol string does not match, the pattern is shifted by 3 symbols, so S [T [0]] = 3.
[S3] As initial settings, i = 0 and shift = 0.
[S4] The following loop processing of steps S5 to S7 is performed until i becomes the number of symbols of the pattern (pattern length).
[S5] The text Txt [shift [T [i]] at the shift position shift [T [i]] and the pattern Pat [T [i]] are compared with each other. If so, go to step S7.
[S6] Shift by a shift amount stored in advance (i = 0, shift = shift + S [T [i]]).
[S7] Move to the next symbol to be verified (i ++).
[0062]
It will be described with reference to a flowchart search method to the next. FIG. 10 is a flowchart showing the processing procedure of the search method.
[S11] When performing symbol collation between the pattern and text, the shift amount, which is the sum of the shift amount, which can be taken by the shift operation when symbol collation fails, multiplied by the probability for the shift amount to occur. The search efficiency that is the ratio of the expected value and the number of symbol matching is calculated.
[S12] A search is performed by collating symbols in the order of the symbols having the largest search efficiency values.
[0063]
FIG. 11 is a flowchart showing the processing procedure of the search method when the search efficiency is introduced by combining the pattern search method and the BM method.
[S21] The same symbol string is specified from the pattern.
[S22] The search efficiency is calculated for the symbols in the same symbol string.
[S23] The search efficiency of the BM method is calculated for all symbols.
[S24] The symbol collation order is determined from the search efficiency.
[S25] A search is executed based on the symbol collation order.
[0064]
FIG. 12 is a flowchart showing the processing procedure of the search method when the search efficiency is introduced by combining the pattern search method and the KMP method.
[S31] The position of the same symbol string is specified from the pattern.
[S32] The search efficiency is calculated for the symbols in the same symbol string.
[S33] The search efficiency of the KMP method is calculated for all symbols.
[S34] The symbol collation order is determined from the search efficiency.
[S35] A search is executed based on the symbol collation order.
[0065]
FIG. 13 is a flowchart showing the processing procedure of the search method when the search efficiency is introduced by combining the pattern search method, the BM method, and the KMP method.
[S41] The position of the same symbol string is specified from the pattern.
[S42] The search efficiency is calculated for the symbols in the same symbol string.
[S43] The search efficiency of the BM method is calculated for all symbols.
[S44] The search efficiency of the KMP method is calculated for all symbols.
[S45] The symbol collation order is determined from the search efficiency.
[S46] A search is executed based on the symbol collation order.
[0066]
As described above, according to the search device 20 and the search method, when searching text or binary data, the symbol matching procedure is dynamically changed using the search efficiency to avoid useless symbol matching. This makes it possible to search for a pattern at a higher speed than when searching only by the conventional KMP method or BM method.
[0067]
Next to the images receiving apparatus will be described. FIG. 14 is a diagram showing the configuration of the image communication system. The image communication system 100 includes an image transmission device 120 and an image reception device 110 that are connected via a network 101.
[0068]
For the image receiving apparatus 110, the stream data receiving unit 111 receives and stores stream data such as MPEG4 transmitted from the image transmitting apparatus 120. Header search unit 112 has a function of search device 10 or the search unit 20 searches the header information from the stream data (including the same symbol string), and outputs the header position information.
[0069]
The information separation unit 113 separates the header information and the image information from the stream signal based on the header position information. The header processing unit 114 processes the header information, and the decoder 115 decodes the image information and displays it on the screen.
[0070]
In the above description, it is applied to the images communication system, the search, such as a database or file other than this, such as synchronization flag detection in communication HDLC etc., apply to the search system, search software in various fields Is possible.
[0071]
(Supplementary note 1) In a search device for searching for a pattern,
The same symbol string recognition unit for recognizing the same symbol string consisting of consecutive identical symbols in the pattern;
The symbol matching between the pattern and text is performed from the last symbol of the same symbol string, and if the symbol matching of the last symbol fails, the pattern is shifted by the number of the same symbol and the symbol matching is performed again. A search processing unit for performing a search,
A search device comprising:
[0072]
(Additional remark 2) When the said search process part attaches the number of 1-n from the head of the said same symbol string to the tail, and performs symbol collation from the last symbol of the said same symbol string toward the head, The search device according to appendix 1, wherein when the k (2 ≦ k ≦ n) th symbol collation fails, the symbol collation is performed again after shifting the pattern by k patterns.
[0073]
(Supplementary Note 3) In a search device for searching for a pattern,
When performing symbol matching between a pattern and text, an expected value of the shift amount, which is a sum of the shift amount that can be taken by the shift operation when symbol matching fails and the probability for the shift amount to occur, and , A search efficiency calculation unit that calculates a search efficiency that is a ratio of the number of symbol matching,
A search processing unit for performing a search by performing symbol matching in the order of the symbols having the highest search efficiency values;
A search device comprising:
[0074]
(Additional remark 4) The said search efficiency calculation part introduce | transduces the said search efficiency into at least one of the pattern search method, BM method, and KMP method of searching from the tail end of the same symbol sequence which consists of the same continuous symbol in a pattern The search device according to supplementary note 3, wherein:
[0075]
(Additional remark 5) The said search process part performs the search based on the order of the symbol with a large value of the said search efficiency, combining the said pattern search method, the said BM method, and the said KMP method arbitrarily. The described search device.
[0076]
(Additional remark 6) In the pattern search method which searches the pattern containing the same symbol string,
Recognize the same symbol string consisting of consecutive identical symbols in the pattern,
From the last symbol of the same symbol string, perform symbol matching between the pattern and text,
A pattern search method characterized in that, when the symbol matching of the last symbol fails, the pattern is shifted by the number of the same symbols and the symbol matching is performed again for the search.
[0077]
(Supplementary Note 7) k (2 ≦ k ≦) is used when symbol matching is performed from the last symbol of the same symbol string to the beginning by assigning numbers 1 to n from the beginning to the last of the same symbol sequence. n) The pattern search method according to appendix 6, wherein when the first symbol collation fails, the pattern is shifted by k patterns and the symbol collation is performed again.
[0078]
(Supplementary note 8) In a search method for searching for a pattern,
When performing symbol matching between a pattern and text, an expected value of the shift amount, which is a sum of the shift amount that can be taken by the shift operation when symbol matching fails and the probability for the shift amount to occur, and , Calculate the search efficiency, which is the ratio of the number of symbol matches,
A search method, wherein a search is performed by comparing symbols in descending order of the search efficiency value.
[0079]
(Supplementary Note 9) The supplementary note is characterized in that the search efficiency is introduced into at least one of a pattern search method, a BM method, and a KMP method for performing a search from the end of the same symbol string composed of the same continuous symbols in the pattern. 8. The search method according to 8.
[0080]
(Additional remark 10) The search method of Additional remark 9 characterized by searching according to the order of the symbol with the said largest search efficiency value, combining the said pattern search method, the said BM method, and the said KMP method arbitrarily.
[0081]
(Supplementary Note 11) In an image receiving apparatus that performs image reception control,
A stream data receiving unit for receiving image-compressed stream data;
The same symbol string recognition unit that recognizes the same symbol string composed of the same consecutive symbols in the pattern, and the symbol of the pattern and stream data are collated from the last symbol of the same symbol string, and the last symbol If the symbol matching fails, the header search unit configured by a search processing unit that shifts the pattern by the number of the same symbols and performs symbol matching again to search for the header,
An image receiving apparatus comprising:
[0082]
(Supplementary Note 12) The search processing unit assigns numbers 1 to n from the head to the tail of the same symbol string, and performs symbol matching from the tail symbol to the head of the same symbol string. 12. The image receiving apparatus according to appendix 11, wherein when the k (2 ≦ k ≦ n) -th symbol collation fails, the pattern is shifted by k patterns and symbol collation is performed again.
[0083]
(Supplementary Note 13) In an image receiving apparatus that performs image reception control,
A stream data receiving unit for receiving image-compressed stream data;
Expected shift amount, which is the sum of the shift amount, which can be taken by the shift operation when symbol collation fails, and the probability that the shift amount occurs when performing symbol collation between the pattern and the stream data A search efficiency calculation unit that calculates a search efficiency that is a ratio of the value and the number of symbol matching, a search processing unit that searches for a header by performing symbol matching in the order of a symbol having a large value of the search efficiency, A header search unit comprising:
An image receiving apparatus comprising:
[0084]
(Additional remark 14) The said search efficiency calculation part introduce | transduces the said search efficiency into at least one of the pattern search method, BM method, and KMP method which search from the tail end of the same symbol sequence which consists of the same continuous symbol in a pattern The image receiving device as set forth in appendix 13, wherein:
[0085]
(Supplementary Note 15) The supplementary note 14 is characterized in that the search processing unit performs a search based on an order of symbols having a large value of the search efficiency by arbitrarily combining the pattern search method, the BM method, and the KMP method. The image receiving device described.
[0086]
【The invention's effect】
As described above, the search device according to the present invention performs symbol matching between a pattern and text from the last symbol of the same symbol string in the pattern. The pattern is shifted. As a result, useless symbol matching can be avoided, and a search can be performed at high speed.
[Brief description of the drawings]
FIG. 1 is a principle diagram of the search system.
FIG. 2 is a diagram for explaining a KMP method.
FIG. 3 is a diagram for explaining a BM method;
FIG. 4 is a diagram for explaining a BM method;
FIG. 5 is a diagram showing the search procedure.
FIG. 6 is a diagram showing a search procedure of a conventional BM method.
FIG. 7 is a diagram illustrating a configuration of a search device.
FIG. 8 is a diagram showing a program for counting the maximum length of the same symbol string.
FIG. 9 is a flowchart showing a processing procedure of a pattern search method.
FIG. 10 is a flowchart showing a processing procedure of a search method.
FIG. 11 is a flowchart showing a processing procedure of a search method when search efficiency is introduced by combining a pattern search method and a BM method.
FIG. 12 is a flowchart showing a processing procedure of a search method when search efficiency is introduced by combining a pattern search method and a KMP method.
FIG. 13 is a flowchart showing a processing procedure of a search method when a search efficiency is introduced by combining a pattern search method, a BM method, and a KMP method.
FIG. 14 is a diagram illustrating a configuration of an image communication system.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 10 Search apparatus 11 Same symbol sequence recognition part 12 Search processing part

Claims (1)

In a search device that searches for patterns,
The same symbol string recognition unit for recognizing the same symbol string consisting of consecutive identical symbols in the pattern;
And line intends search processing unit symbols match between the pattern and the text,
Have
The search processing unit
Numbering 1 to n from the beginning to the end of the same symbol string, starting symbol matching from the last symbol of the same symbol string toward the beginning of the same symbol string ,
When the symbol collation of the symbols assigned k (1 ≦ k ≦ n) among the symbols in the same symbol string has failed, the pattern corresponding to k symbols from the failed position toward the tail end. Shift and perform symbol collation again from the last symbol of the same symbol string ,
A search device characterized by that.

Images