JP6480304B2 - Signal search method, apparatus, and program - Google Patents

Description

  The present invention relates to a signal search method, apparatus, and program.

  Conventionally, regarding a signal search method, a target signal that is time-series data from an accumulated signal that is time-series data, such as a signal detection system, a signal detection server, a signal detection method, and a program (see Patent Document 1). There is known a method for searching for a location similar to.

Japanese Patent No. 4358229

  However, the method described in Patent Document 1 has a problem in that detection omission occurs when there are partial matches of various lengths in the target signal.

  The present invention has been made in consideration of the above circumstances, and can search for a section of an accumulated signal that partially matches the target signal at an arbitrary length without specifying the length to be detected in advance. It is an object to provide a search method, apparatus, and program.

  In order to achieve the above object, a signal search method according to the present invention includes an accumulated signal acquisition unit, an accumulated feature extraction unit, an accumulated feature quantization unit, a database creation unit, an objective signal acquisition unit, an objective feature extraction unit, an objective feature quantum. A signal search method in a signal search device including a conversion unit, a linked list creation unit, and a search unit, wherein the stored signal acquisition unit acquires a stored signal stored in a database, and the stored feature extraction unit includes: Extracting accumulated feature data from the accumulated signal, the accumulated feature quantizing unit obtaining accumulated quantized data from the accumulated feature data, and the database creating unit extracting the accumulated quantized data into the database The target signal acquisition unit acquires the target signal to be a query, and the target feature extraction unit includes the target signal. Extracting the target feature data from the target feature data, the target feature quantization unit acquiring the target quantized data from the target feature data, and the linked list creating unit presetting the accumulated quantized data For each of the collated intervals, the pair corresponding to each other is searched for from each pair of the portion of the target quantized data and the portion of the collated interval of the accumulated quantized data, and the searched A step of creating a linked list in which corresponding pairs are registered in the matching section based on a pair, and the search unit searches a section corresponding to the accumulated signal and the target signal based on the linked list. Steps.

  The signal search device according to the present invention includes an accumulated signal acquisition unit that acquires an accumulated signal accumulated in a database, an accumulated feature extraction unit that extracts accumulated feature data from the accumulated signal, and accumulated quantized data from the accumulated feature data. A storage feature quantization unit to be acquired, a database creation unit that registers the stored quantization data in the database, a target signal acquisition unit that acquires a target signal to be queried, and a purpose of extracting target feature data from the target signal A feature extraction unit; a target feature quantization unit that obtains target quantization data from the target feature data; and a portion of the target quantization data for each of preset collation intervals of the accumulated quantization data; From each pair of the stored quantized data with the portion of the matching section, the pair corresponding to the portion is searched for, and the searched pair A linked list creating unit for creating a linked list in which corresponding pairs are registered in the matching section, and a search unit for searching for a corresponding section between the accumulated signal and the target signal based on the linked list; , Including.

  The linked list creation unit deletes the linked list other than the longest linked list from the linked list of each of the matching sections, and the search unit searches for a corresponding section. The step of searching for a corresponding section by the search unit searching for a corresponding section of the accumulated signal and the target signal based on a linked list is based on the linked list exceeding a predetermined length, It is possible to search for a section corresponding to the accumulated signal and the target signal.

  When the linked list creating unit creates the linked list for the verification section, the searched combination of the time of the accumulated signal and the time of the target signal of the pair corresponding to the verification section is The pair can be registered in the linked list when it is within a predetermined range.

  The program according to the present invention is a program for causing a computer to execute each step of the signal search method.

  As described above, according to the signal search method, apparatus, and program of the present invention, the portion of the target quantized data and the stored quantized data for each of the collation intervals set in advance in the accumulated quantized data. From each of the pairs with the part of the matching section, search for a pair corresponding to each other, create a linked list in which the corresponding pair in the matching section is registered based on the searched pair, and based on the linked list, By searching for a corresponding section between the accumulated signal and the target signal, it is possible to search for a section of the accumulated signal that partially matches the target signal at an arbitrary length without specifying the length to be detected in advance. The effect is obtained.

It is a block diagram which shows one structural example of the signal search apparatus which concerns on embodiment of this invention. It is a conceptual diagram explaining the process which acquires accumulation | storage feature data and accumulation | storage quantization data from an accumulation | storage signal. It is explanatory drawing for demonstrating the process of the connection list preparation part with respect to the pair of the part of the objective quantization data part and the accumulation | storage quantization data part. It is explanatory drawing for demonstrating an example of the creation method of a linked list. It is explanatory drawing for demonstrating an example of the creation method of a linked list. It is explanatory drawing for demonstrating an example of the creation method of a linked list. It is explanatory drawing for demonstrating an example of the creation method of a linked list. It is explanatory drawing for demonstrating an example of the creation method of a linked list. It is a flowchart which shows the content of the accumulation | storage process routine in the signal search apparatus which concerns on embodiment of this invention. It is a flowchart which shows the content of the search process routine in the signal search apparatus which concerns on embodiment of this invention. It is a figure for demonstrating an example of the conventional method.

<Overview>
In the conventional method, when searching for a section that partially matches the target signal, for example, a signal corresponding to the length of the partial match is extracted from the target signal, and the section of the accumulated signal that partially matches the target signal. Had to explore. Further, in the conventional method, as shown in FIG. 11, when the horizontal axis is the time of the accumulated signal and the vertical axis is the time of the target signal, it is assumed that the correspondence between the signal parts in the verification section is linear. Therefore, it is considered difficult to handle a situation where successive fluctuations occur in the signal portion.

  In the conventional method, it is conceivable to use the fact that the correspondence between the signal portions is linear, for example, to detect a straight line by Hough transform and search for a corresponding section between the accumulated signal and the target signal. However, the Hough transform method is an approach that narrows the allowable range globally for signals. Considering fluctuations in the signal part, it is possible to increase the parameter that represents the slope of the straight line obtained by the Hough transform, but as the parameter increases, the voting space for searching becomes enormous, and the memory and processing time are unrealistic. Time.

  Therefore, in the present embodiment, when the corresponding portions of the target signal portion and the accumulated signal portion are within a predetermined range, they are registered in the linked list, and the target signal and the arbitrary signal are specified by the linked list. The section of the accumulated signal partially matching with the length of is searched. In the present embodiment, while viewing the allowable range of fluctuation in the signal portion locally, pairs of matching portions are registered in the linked list, and candidates for sections in which the accumulated signal and the target signal partially match are narrowed down.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

<System configuration>
FIG. 1 is a block diagram showing an embodiment of a signal search apparatus according to the present invention. A signal search apparatus 100 shown in FIG. 1 is an apparatus that identifies an accumulated signal corresponding to a target signal. Specifically, a CPU (Central Processing Unit), a RAM, and a program for executing processing routines described later. The computer is provided with a ROM that stores the above, and is functionally configured as follows.

  The signal search apparatus 100 includes an input unit 10, a calculation unit 20, and an output unit 50.

  The input unit 10 receives an input of an accumulation signal and a target signal serving as a query. In this embodiment, a case where the length of the target signal is equal to or shorter than the length of the accumulated signal will be described as an example. As the accumulation signal and the target signal, for example, time series data such as an acoustic signal and a moving image signal can be used.

  The computing unit 20 includes an accumulated signal acquisition unit 22, an accumulated feature extraction unit 24, an accumulated feature quantization unit 26, a database creation unit 28, a database 30, an objective signal acquisition unit 32, an objective feature extraction unit 34, and an objective feature quantization unit 36. A linked list creation unit 38 and a search unit 40.

  The accumulated signal acquisition unit 22 acquires the accumulated signal received by the input unit 10. The accumulated signal is accumulated in the database 30 described later.

  The storage feature extraction unit 24 extracts storage feature data from the storage signal acquired by the storage signal acquisition unit 22.

  For example, as shown in FIG. 2, the accumulation feature extraction unit 24 samples the accumulation signal in a time series at a predetermined interval, and generates a feature vector for each sampling. Then, the accumulated feature extraction unit 24 normalizes the feature vector for each of the plurality of feature vectors generated in time series. Then, the accumulated feature extraction unit 24 uses a plurality of normalized feature vectors as accumulated feature data. As a feature vector, for example, a feature quantity such as a power spectrum value for each frequency in sampling can be converted into a multidimensional vector to obtain a feature vector. As a method for normalizing the vector, for example, after calculating an average value and a standard deviation for a predetermined section in the vicinity of the time series direction of each element of the vector of interest, and subtracting the average from the element of interest Use the value divided by the standard deviation.

  The accumulated feature quantization unit 26 acquires accumulated quantized data from the accumulated feature data extracted by the accumulated feature extraction unit 24.

  For example, as shown in FIG. 2, the accumulated feature quantization unit 26 sets each element of the feature vector for each of a plurality of feature vectors of the accumulated feature data extracted by the accumulated feature extraction unit 24 according to a predetermined threshold. Quantization is performed, and a quantized vector having elements obtained by the quantization as elements is generated. Then, the accumulation feature quantization unit 26 acquires the generated plurality of quantization vectors as accumulation quantization data.

  The predetermined threshold value for performing the quantization is the minimum of the mean square error between the numerical value of the element obtained by binarization and the numerical value of the original element when the accumulated feature data is binarized, for example. It can be obtained by selecting a point to be converted. In addition, the quantization method can also use a method other than the method of selecting a point at which the mean square error is minimized. For example, each element of the feature vector is simply divided into fixed intervals and quantized. It is also possible to use a method and a method of performing quantization by dividing a predetermined non-linear interval.

  In the accumulated quantized data, the quantization vectors in each sampling are sequentially arranged in the order of sampling in time series. Note that the number of arrays (the length of the array) is shorter or equal to the target quantized data described later than the accumulated quantized data.

  The database creating unit 28 registers the accumulated quantized data acquired by the accumulated feature quantizing unit 26 in the database 30.

  The database 30 stores the accumulated quantized data acquired by the accumulated feature quantization unit 26. The accumulated quantized data includes a quantized vector that is quantized for each portion of the accumulated signal sampled in time series.

  The target signal acquisition unit 32 acquires the target signal received by the input unit 10.

  Similar to the accumulated feature extraction unit 24, the target feature extraction unit 34 extracts target feature data from the target signal acquired by the target signal acquisition unit 32.

  Similar to the accumulated feature extraction unit 24, the target feature quantization unit 36 acquires target quantization data from the target feature data extracted by the target feature extraction unit 34. The target quantized data includes a quantized vector quantized for each of the time-sampled target signal portions.

  For each of the collation intervals set in advance in the accumulated quantized data stored in the database 30, the linked list creation unit 38 sets a pair of the target quantized data portion and the corresponding quantized interval portion of the accumulated quantized data. From each of these, a pair corresponding to each part is searched. Then, the linked list creation unit 38 creates a linked list in which the corresponding pairs are registered in the matching section based on the searched pairs for each of the matching sections.

  When the linked list creation unit 38 creates the linked list for the matching section, a combination of the time of the accumulated signal and the time of the target signal of the pair corresponding to the matching section searched for is determined in advance. If it is within the specified range, the pair is registered in the linked list.

  FIG. 3 is a diagram for explaining an example of processing of the linked list creating unit 38 for a pair of the target quantized data portion and the accumulated quantized data portion.

  For example, the linked list creation unit 38 sets a predetermined collation section as shown in FIG. Next, the linked list creation unit 38 calculates, for each pair of the target quantized data portion and the accumulated quantized data portion, the similarity of the pair of the portion in the set collation interval, Is equal to or greater than a predetermined threshold value, it is determined that the pair of the parts corresponds. As the similarity, for example, the Euclidean distance or the Hamming distance between the quantized data portions can be used.

  Then, when the linked list creation unit 38 creates the linked list for the matching section, the combination of the time of the accumulated signal and the time of the target signal of the pair determined to correspond in the matching section is determined in advance. If it is within the defined range, the pair is registered in the linked list. A linked list is created for each predetermined collation section.

  4 to 6 are diagrams for explaining an example of a linked list creation method according to the present embodiment, which will be described in detail below.

  In the coordinate system shown in FIG. 4, the horizontal axis represents the time of the target signal, the vertical axis represents the time of the accumulated signal, and the plotted points represent pairs corresponding to the respective times of the quantized data in the collation interval. In the present embodiment, for example, the dotted line shown in FIG. 4 is created as a linked list.

  In the present embodiment, as shown in FIG. 5, the combination of the time of the accumulated signal and the time of the target signal of a pair determined to correspond in the verification section is within a predetermined range. The pair is registered in the linked list. The shaded portion in FIG. 5 represents a link target range that is a predetermined range. As the link target range, for example, a range that is a straight line from the start point of the plot point and is surrounded by a straight line having an inclination of 0.9 and a straight line having an inclination of 1.1 can be used.

  Then, as shown in FIG. 6, each pair that is determined to correspond in the verification section and that exists in the link target range is registered in the linked list, and each registered pair is Connected. It is shown that the longer the linked list with linked pairs, the more corresponding parts in the accumulated signal and the target signal, and the shorter the linked list, the corresponding part in the accumulated signal and the target signal. It is expressed that there are few.

  As shown in FIG. 6, when there are a plurality of linked lists including the same single plot point, as shown in FIG. 7, pruning is performed on the branches representing the links. As pruning, for example, one branch is regarded as one hop, the linked list with the larger number of hops is left preferentially, and the linked list with the smaller number of pops is deleted. When the number of hops is the same, as shown in FIG. 8, the longer distance representing the difference between the time of the branch start point and the time of the branch end point is given priority, and the shorter distance is deleted. For example, in FIG. 8, the distance A representing the difference between the time of the accumulated signal that is the start point of the branch and the time of the accumulated signal that is the end point of the branch is compared with the distance B at the branch branch. The one with priority is left and the shorter one is pruned. Therefore, in the example of FIG. 8, a linked list X and a linked list Y are output.

  In FIG. 8, the case where the difference between the time of the accumulated signal that is the start point of the branch and the time of the accumulated signal that is the end point of the branch has been described, but the time of the target signal that is the start point of the branch and the end point of the branch Pruning may be performed by comparing the difference with the time of the target signal. Alternatively, the branch to be pruned may be determined by combining the time of the accumulation signal and the time of the target signal with respect to the branch start point and the branch end point.

  The search unit 40 searches for a corresponding section between the accumulated signal and the target signal based on the maximum-length linked list among a plurality of linked lists created for each verification section by the linked list creating unit 38.

  For example, the search unit 40 determines, based on the linked list X and Y shown in FIG. 8 above, Y that is the linked list of the longest length, the matching section corresponding to the linked list of the longest length between the accumulated signal and the target signal. Search for a corresponding section.

  The output unit 50 outputs the time or coordinates of the section corresponding to the accumulated signal and the target signal searched by the search unit 40 as a result.

<Operation of signal search device>
Next, the operation of the signal search apparatus according to this embodiment will be described.

<Storage processing routine>
When the accumulation signal is input from the input unit 10, the signal search apparatus executes an accumulation process routine shown in FIG. The accumulation processing routine is executed every time an accumulation signal is input.

  First, in step S <b> 100, the accumulated signal acquisition unit 22 acquires the accumulated signal received by the input unit 10.

  In the next step S102, the accumulated feature extraction unit 24 extracts accumulated feature data from the accumulated signal acquired in step S100.

  In step S104, the accumulated feature quantization unit 26 acquires accumulated quantized data from the accumulated feature data extracted in step S102.

  In the next step S106, the database creation unit 28 stores the accumulated quantized data acquired in step S106 in the database 30 and ends the accumulation process routine.

<Search processing routine>
When a target signal is input through the input unit 10, a search processing routine shown in FIG. 10 is executed in the signal search device. The search processing routine is executed each time a target signal is input.

  First, in step S <b> 200, the target signal acquisition unit 32 acquires the target signal received by the input unit 10.

  In the next step S202, the target feature extraction unit 34 extracts target feature data from the target signal acquired in step S100.

  In step S204, the target feature quantization unit 36 acquires target quantization data from the target feature data extracted in step S202.

  In the next step S <b> 206, the linked list creation unit 38 sets one collation section in the accumulated quantized data stored in the database 30.

  Then, in step S208, the linked list creation unit 38 creates the portion of the target quantized data acquired in step S204 and the accumulated quantized data stored in the database 30 for the collation interval set in step S206. From each pair with the part of the matching section, a pair corresponding to the part is searched.

  Then, in step S210, the linked list creation unit 38, based on the search result in step S208 in the matching section set in step S206, the time of the accumulated signal of the corresponding pair and the target signal in the matching section. When the combination with the time is within a predetermined range, the pair is registered in the linked list, and the linked list of the matching section is created.

  In step S212, it is determined whether or not the processing in steps S206 to S210 has been executed for the entire range of the accumulated signal. When the processes of steps S206 to S210 are executed for the entire range of the accumulated signal, the process proceeds to step S214. On the other hand, if there is a collation section in which the processes in steps S206 to S210 are not executed, the process returns to step S206.

  In step S214, the search unit 40 searches for a section corresponding to the accumulated signal and the target signal based on the maximum length linked list among the plurality of linked lists created for each verification section in step S210. To do.

  In step S216, the output unit 50 outputs the time or coordinates of the section corresponding to the accumulated signal and the target signal searched in step S214, and ends the search processing routine.

  As described above, according to the signal search apparatus according to the embodiment of the present invention, the target quantized data portion and the stored quantized data are stored in each of the preset matching intervals in the accumulated quantized data. From each of the pairs with the part of the matching section, search for a pair corresponding to each other, create a linked list in which the corresponding pair in the matching section is registered based on the searched pair, and based on the linked list, By searching for a corresponding section between the accumulated signal and the target signal, it is possible to search for a section of the stored signal that partially matches the target signal at an arbitrary length without specifying the length to be detected in advance.

  Further, by managing the corresponding portions of the target signal and the accumulated signal in the linked list, it is possible to detect partial correspondence of an arbitrary length in the target signal.

  Further, it is possible to detect a partial match having an arbitrary length as compared with the prior art.

  Further, in the prior art, in order to detect partial coincidence, it is necessary to cut out a signal having an arbitrary length from the target signal. However, in this embodiment, it is not necessary to cut out a signal having an arbitrary length from the target signal. Instead, it is possible to search for a section of the accumulated signal that partially matches the target signal.

  Further, it is possible to search for a corresponding section between the accumulated signal and the target signal in correspondence with an arbitrary length of the target signal.

  Further, as shown in FIGS. 6 to 8, even if the correspondence between the signal parts is not linear but curved, the correspondence between the parts of the accumulated signal and the target signal is taken into consideration. A corresponding section between the signal and the target signal can be searched.

  Note that the present invention is not limited to the above-described embodiment, and various modifications and applications are possible without departing from the gist of the present invention.

  For example, in the above-described embodiment, the linked list creation unit 38 may delete a linked list other than the maximum linked list from each linked list in the verification section.

Moreover, in the said embodiment, although the search part 40 demonstrated as an example the case where the area corresponding to an accumulation | storage signal and a target signal was searched based on the maximum length linked list, it is not limited to this. Absent.
For example, the search unit 40 may search for a section corresponding to the accumulated signal and the target signal based on a linked list exceeding a predetermined length. For example, when there are a plurality of accumulated signals, a linked list having the maximum length is selected for each accumulated signal, a linked list exceeding a predetermined length is selected from the selected linked lists, and the selected linked list is selected. Based on the list, an accumulated signal including a signal corresponding to the target signal and a section corresponding to the accumulated signal and the target signal may be searched.
Further, the search unit 40 may search for a section corresponding to the accumulated signal and the target signal based on the linked list having the top N lengths in the linked list.

  In this embodiment, the case where the accumulated signal has a length longer than the target signal has been described. However, the present invention is not limited to this, and the accumulated signal may be shorter than the target signal. .

  In the present embodiment, the case where there is one accumulated signal has been described, but the present invention is not limited to this. For example, a plurality of accumulated signals may exist. In this case, the portion of the accumulated signal corresponding to the target signal may be specified for each accumulated signal. A part of the accumulated signal corresponding to the target signal may be specified by collecting all the accumulated signals.

  Further, the above accumulation signal and target signal may be an audio signal, a video signal, or a signal in which an audio signal and a video signal are combined.

  In the present embodiment, the case where time series signals are used as the accumulation signal and the target signal has been described, but the present invention is not limited to this. For example, two-dimensional data such as an image, or more multidimensional data may be used for the accumulation signal and the target signal. In this case, if a collation interval is provided in the accumulated signal and a location of a signal similar to the target signal is detected, the search process can be performed using the linked list in the same manner as the above-described process.

  In the present embodiment, as shown in FIGS. 6 to 8 described above, each pair that is determined to correspond in the matching section and exists within a predetermined range is connected. Although the case where each registered pair is connected to the list and connected is described as an example, the present invention is not limited to this. For example, when the sum of the similarities of pairs determined to correspond in the matching section is equal to or greater than a predetermined threshold, each pair is registered in the linked list, and each registered pair is linked May be.

  In addition, the above-described signal search apparatus has a computer system therein, but the “computer system” includes a homepage providing environment (or display environment) if a WWW system is used.

  In the present specification, the embodiment has been described in which the program is installed in advance. However, the program can be provided by being stored in a computer-readable recording medium.

DESCRIPTION OF SYMBOLS 10 Input part 20 Operation part 22 Accumulated signal acquisition part 24 Accumulated feature extraction part 26 Accumulated feature quantization part 28 Database preparation part 30 Database 32 Objective signal acquisition part 34 Objective feature extraction part 36 Objective feature quantization part 38 Linked list creation part 40 Search unit 50 Output unit 100 Signal search device

Claims (7)

In a signal search apparatus including an accumulated signal acquisition unit, an accumulated feature extraction unit, an accumulated feature quantization unit, a database creation unit, an objective signal acquisition unit, an objective feature extraction unit, an objective feature quantization unit, a linked list creation unit, and a search unit A signal search method comprising:
The accumulated signal obtaining unit obtaining an accumulated signal accumulated in a database;
The accumulated feature extraction unit extracting accumulated feature data from the accumulated signal;
The accumulated feature quantization unit obtains accumulated quantized data from the accumulated feature data; and
The database creating unit registering the accumulated quantized data in the database;
The target signal acquisition unit acquires a target signal to be a query;
The target feature extraction unit extracting target feature data from the target signal;
The target feature quantizing unit obtains target quantized data from the target feature data;
The linked list creating unit, for each of the preset matching intervals of the accumulated quantized data, from each of the pair of the target quantized data portion and the matching interval portion of the accumulated quantized data Searching for the pair corresponding to each other, and creating a linked list in which the corresponding pair is registered in the matching section based on the searched pair;
The search unit searching for a corresponding section between the accumulated signal and the target signal based on the linked list ;
The step of creating the linked list by the linked list creating unit deletes the linked list other than the longest linked list among the linked lists of each of the matching sections,
The step of searching for a section corresponding to the search unit searches for a section corresponding to the accumulated signal and the target signal based on the linked list having the maximum length.
Signal search method. In a signal search apparatus including an accumulated signal acquisition unit, an accumulated feature extraction unit, an accumulated feature quantization unit, a database creation unit, an objective signal acquisition unit, an objective feature extraction unit, an objective feature quantization unit, a linked list creation unit, and a search unit A signal search method comprising:
The accumulated signal obtaining unit obtaining an accumulated signal accumulated in a database;
The accumulated feature extraction unit extracting accumulated feature data from the accumulated signal;
The accumulated feature quantization unit obtains accumulated quantized data from the accumulated feature data; and
The database creating unit registering the accumulated quantized data in the database;
The target signal acquisition unit acquires a target signal to be a query;
The target feature extraction unit extracting target feature data from the target signal;
The target feature quantizing unit obtains target quantized data from the target feature data;
The linked list creating unit, for each of the preset matching intervals of the accumulated quantized data, from each of the pair of the target quantized data portion and the matching interval portion of the accumulated quantized data Searching for the pair corresponding to each other, and creating a linked list in which the corresponding pair is registered in the matching section based on the searched pair;
The search unit searching for a corresponding section between the accumulated signal and the target signal based on the linked list ;
The step of creating the linked list includes the time of the accumulated signal and the time of the target signal of the pair corresponding to the collation section searched when creating the linked list for the collation section. When the combination with is within a predetermined range, the pair is registered in the linked list.
Signal search method. The step of searching for a section corresponding to the search unit searches for a section corresponding to the accumulated signal and the target signal based on the linked list exceeding a predetermined length. The signal search method according to claim 1. An accumulation signal acquisition unit for acquiring an accumulation signal to be accumulated in the database;
An accumulation feature extraction unit for extracting accumulation feature data from the accumulation signal;
An accumulated feature quantization unit for obtaining accumulated quantized data from the accumulated feature data;
A database creation unit for registering the accumulated quantized data in the database;
A target signal acquisition unit for acquiring a target signal to be a query;
A target feature extraction unit for extracting target feature data from the target signal;
A target feature quantization unit for acquiring target quantization data from the target feature data;
For each of the preset matching intervals of the accumulated quantized data, the portions corresponding to each other from each of the pair of the target quantized data portion and the matching interval portion of the accumulated quantized data A linked list creating unit that searches for a pair and creates a linked list in which corresponding pairs are registered in the matching section based on the searched pair.
A search unit that searches for a corresponding section between the accumulated signal and the target signal based on the linked list;
Only including,
The linked list creation unit deletes the linked list other than the longest linked list among the linked lists of each of the matching sections,
The search unit searches for a corresponding section between the accumulated signal and the target signal based on the linked list having the maximum length.
Signal search device. An accumulation signal acquisition unit for acquiring an accumulation signal to be accumulated in the database;
An accumulation feature extraction unit for extracting accumulation feature data from the accumulation signal;
An accumulated feature quantization unit for obtaining accumulated quantized data from the accumulated feature data;
A database creation unit for registering the accumulated quantized data in the database;
A target signal acquisition unit for acquiring a target signal to be a query;
A target feature extraction unit for extracting target feature data from the target signal;
A target feature quantization unit for acquiring target quantization data from the target feature data;
For each of the preset matching intervals of the accumulated quantized data, the portions corresponding to each other from each of the pair of the target quantized data portion and the matching interval portion of the accumulated quantized data A linked list creating unit that searches for a pair and creates a linked list in which corresponding pairs are registered in the matching section based on the searched pair.
A search unit that searches for a corresponding section between the accumulated signal and the target signal based on the linked list;
Only including,
When the linked list creating unit creates the linked list for the verification section, the searched combination of the time of the accumulated signal and the time of the target signal of the pair corresponding to the verification section is Register the pair in the linked list if it is within a predetermined range;
Signal search device. The search unit searches for a corresponding section between the accumulated signal and the target signal based on the linked list exceeding a predetermined length.
The signal search device according to claim 4 or 5. The program for making a computer perform each step of the signal search method of any one of Claims 1-3 .