JP2000276478A - Method and device for detecting time-series data and recording medium where program thereof is recorded - Google Patents

Abstract

PROBLEM TO BE SOLVED: To speedily and securely detection object time-series data to be detected by matching the electronic watermarks of input data and original data and determining matching data when similar data are found through a matching process between the input data and original data. SOLUTION: When a specific CM is retrieved from a broadcast signal like a CM, the original data of the CM is recorded previously in an original data storage part 141, electronic watermark information embedded in the original data of the CM is recorded in an electronic watermark storage part 142, and a retrieved broadcast radio wave is recorded in a broadcast signal storage part 143. Then the original data and broadcast contents are made into feature signals by a feature signal process part 131, and the similarities of all original data feature signals are calculated, clustered, and recorded in an original data similarity storage part 147. Then a matching process part 133 performs a matching processing between the feature signals of the broadcast signal and original data and collates the electronic watermark read out of the original data position in the broadcast signal with the electronic watermark of the original data when there are similar data to determine matching data.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a time series data detection method and apparatus for accurately detecting time series data registered in advance from time series data.

[0002]

2. Description of the Related Art In some cases, it is required to accurately detect previously registered time-series data from input time-series data. For example, in a sound or video of a broadcast, or a signal of both, a part where a specific commercial is broadcast is detected to automatically verify whether the commercial is broadcast as specified, or a specific music or video is detected. This is a case where a video is detected and the use of a video or music work is automatically monitored.

Conventionally, methods for detecting target time-series data from time-series data include a method using "digital watermark" and a method using "matching".

[0004] The method using the "digital watermark" is an identifier (hereinafter referred to as ID) for authenticating the identity of the data or guaranteeing the original.
) Is embedded in the data and extracted to extract it. This method requires ID embedding processing for the original data, but accurate detection is possible as long as the ID is confirmed.

[0005] The method based on "matching" basically compares the waveform (or characteristic amount) of the target time-series data with the time-series data (audio or video) to be detected, and compares them. When the degree is equal to or more than a certain value (hereinafter, this certain value is referred to as a matching threshold value), it is a method of confirming the identity of time-series data to be detected as existing.

[0006]

However, in the case of a detection system using a digital watermark, since the appearance position of the digital watermark is unknown, it is necessary to check the digital watermark over all the time-series data. There is a problem in that detection requires a lot of processing time. Also,
Even in the case of a search system using matching, it is necessary to perform matching processing on all time-series data. However, an active search method and the like described in Japanese Patent Application No. 10-151723 have been developed, and the processing has been speeded up and the practicability has been improved. However, this method is similar due to the limitations of the processing method. There is a problem that it is difficult to distinguish data having such a waveform (feature amount). For example, in the case of a CM, in addition to the original data of the CM to be examined in the input data, the original data of the CM in which a part is slightly different from the original data of the CM to be examined, for example, most C
In some cases, the original data of the CM is used, such as changing the words of a certain part according to the target audience, or deleting a certain part of the scene from the same CM to make it a simplified version of the CM. In such a case, it is difficult to perform the division by the matching process, and there is a possibility that the whole is detected as the original data of the CM to be examined.

FIGS. 7A and 7B are schematic diagrams for explaining a problem of a detection method by matching using a CM submission confirmation as an example. FIG. 7A shows a state before detection, and FIG. 7B shows a state after detection. . A set 72 of CMs to be detected exists in a set 71 of input data to be detected. The set represented by the ellipse as a result of the matching is the CM 73 detected by the matching, and the CM 73 includes the CM 74 correctly detected by the matching and the CM 76 erroneously detected by the matching. CM7 to be detected on the other hand
The portion excluding the CM 74 that has been correctly detected from No. 2 remains as a CM 75 that has not been detected.

An object of the present invention is to solve the above-mentioned problems of both types, and an object thereof is to quickly and surely detect time-series data to be detected in all regions of the time-series data. To provide a time-series data detection method and a detection device for the same and a recording medium on which the program thereof is recorded.

[0009]

A method for detecting time-series data according to the present invention comprises detecting time-series data in which time-series data to be searched (original data) is detected from input time-series data (input data). In this method, digital watermark information capable of uniquely identifying the original data is embedded in the original data and the input data. The original data to be searched is input, converted into a feature signal, stored as an original data feature signal, and all original data feature signals are matched using a predetermined threshold to calculate similarity. The data is stored as a data similarity table, and digital watermark information of the original data is input and stored to complete necessary original data related data. Input data is input, processed into a feature signal, and stored as an input data feature signal to complete necessary input data related data.

A matching process is performed between the input data characteristic signal and the original data characteristic signal using a predetermined threshold to detect the number and position of the original data in the input data. If there is similar data in the cluster corresponding to the threshold value of the similarity table, the digital watermark embedded in the detected original data position in the input data is read out and compared with the digital watermark information of the original data to obtain similar data. Is eliminated, and the data number and data position of the original data in the input data are determined.

Further, a time schedule indicating the expected positional relationship of the original data in the input data is stored in the input data-related material, and the number of the original data detected by the matching process is included in the input data calculated from the time schedule. If the number of original data is less than the planned number of original data, the matching threshold is changed to a lower value until the detected number of original data is equal to or greater than the planned number of original data included in the input data calculated from the time schedule. The number of data and the data position of the original data in the input data may be detected by repeatedly performing the matching process.

Further, by accumulating and analyzing the result of performing the matching process and the data of the threshold value, a matching threshold value corresponding to the matching process is appropriately set, and the number of original data in the input data is reduced. The data position may be detected.

A time-series data detecting apparatus according to the present invention is a time-series data detecting apparatus for detecting time-series data (original data) to be searched from input time-series data (input data). Unit and original data storage unit, digital watermark input unit and digital watermark storage unit, input data input unit and input data storage unit, time schedule input unit and time schedule storage unit, feature signal processing unit, original data feature signal storage unit, An input data feature signal storage unit, a similarity calculation unit, an original data similarity storage unit, a matching processing unit, an appearance frequency determination unit, a similarity determination unit, a watermark extraction / determination unit, a determination result output unit, and a control unit I have.

The original data input unit inputs the original data and stores it in the original data storage unit. The digital watermark data input unit inputs the digital watermark data embedded in the original data.
The data is stored in the digital watermark data storage together with the embedding destination data, and the input data input unit inputs the input data,
The data is stored in the input data storage unit, and the time schedule input unit inputs the time schedule of the input data and stores the input data in the time schedule storage unit.

The feature signal processing unit performs a feature signal processing on the original data to generate an original data feature signal, stores the signal in the original data feature signal storage unit, performs a feature signal processing on the input data, and executes the input data feature processing. A signal is generated and stored in an input data feature signal storage unit, and a similarity calculation unit calculates a similarity between the original data feature signals by a matching process based on a predetermined threshold value, and generates an original signal having a similar feature. The clustering process is performed so that the data becomes the same cluster, and the data is stored in the original data similarity storage unit together with the cluster information.

The matching processing unit performs a matching process between the original data characteristic signal and the input data characteristic signal using a predetermined threshold value, and the appearance number determination unit determines the expected number of appearances of the original data on the time schedule. Compare the number of appearances of the original data by the matching process, change the threshold value of the matching processing unit as necessary, and the similarity determination unit determines the clustering information of the original data corresponding to the finally selected threshold value. Is read, and it is determined whether there is only one original data or similar data exists in the cluster, and the result is sent to the watermark extraction determining unit. In this case, the matching processing result of the matching processing unit is output to the judgment result output unit. If similar data exists, the digital watermark of the original data is read out and input. The digital watermark embedded in the position of the original data of the data matching processing result is read and collated, and the number of times of matching of the identified original data and the position information are output to the judgment result output unit. The data is output to the outside, and the control unit controls and executes the processing of each unit.

Further, a recording medium is provided, and the operation of the control unit is as follows.
The control may be performed by a time-series data detection system control program recorded on the recording medium.

The recording medium on which the program of the present invention is recorded records a control program for detecting time-series data (original data) to be searched from input time-series data (input data).

A unique ID is embedded in the original data as digital watermark information, and all the original data in the input data can be identified by the digital watermark information. The similarity between the original data is calculated. In calculating the similarity, the same method as that used in the matching process is used.
After the calculation, the data whose similarity is equal to or larger than the clustering threshold value is regarded as similar data.

The basic search is performed by a matching method using the similarity as a threshold value, and only original data having similar data is verified using a digital watermark. In this case, the threshold value can be adaptively changed according to the processing content. The adaptive change is a method in which, for example, when the number of time-series data to be detected in the input data is known in advance, the threshold is lowered until the number of detected data is equal to or larger than the detected number. Generation of original data for detection can be prevented.

At the time of extracting the digital watermark, the position in the time-series data to be scanned is derived by the matching method, so that the verification can be easily performed.

With these features, it is possible to prevent the redundant processing at the time of watermark extraction when only the digital watermark is used, and to distinguish the similar original data which is difficult to distinguish when only the matching is used. Data retrieval from time-series data can be realized at relatively low cost and at high speed.

[0023]

Next, embodiments of the present invention will be described with reference to the drawings. The embodiment of the present invention is a case where the input data is a broadcast signal, but the input signal is not limited to the broadcast signal. FIG. 1 is a block diagram of a time-series data detection device according to a first embodiment of the present invention.

The time-series data detecting apparatus 100 according to the first embodiment of the present invention includes an original data input unit 111 for inputting all the original data that may be used and an original data for storing the input original data. A storage unit 141, a digital watermark data input unit 112 for inputting digital watermark data embedded in the original data, a digital watermark data storage unit 142 for storing the input digital watermark data together with the embedding destination data, a broadcast for inputting a broadcast signal A signal input unit 113 and a broadcast signal storage unit 143 for storing the input broadcast signal, a time schedule input unit 114 for inputting a broadcast time schedule, a time schedule storage unit 144 for storing the input time schedule, and a feature signal processing unit 1
31, an original data feature signal storage unit 145, a broadcast feature signal storage unit 146, a similarity calculation unit 132, an original data similarity storage unit 147, a matching processing unit 133, an appearance number determination unit 1
34, similarity determination unit 135, watermark extraction / determination unit 13
6, a determination result output unit 121 and a control unit 138.

The characteristic signal processing section 131 takes in the original data from the original data storage section 141 and performs a characteristic signal processing to generate an original data characteristic signal.
The broadcast feature signal is stored in the broadcast feature signal storage unit 146, the broadcast signal is taken in from the broadcast signal storage unit 141, the feature signal is converted into a feature signal, and the broadcast feature signal is generated.

The similarity calculation unit 132 reads the original data characteristic signal from the original data characteristic signal storage unit 145, and calculates the similarity between the original data stored by the matching process based on a predetermined clustering threshold. The clustering process is performed so that the original data having similar characteristics are in the same cluster, and the original data is stored in the original data similarity storage unit 147 together with the cluster information. However, the clustering threshold here is the same as the matching threshold used in the matching described later, and when a plurality of matching thresholds are used, the clustering is performed in accordance with the matching threshold. Is recorded.

The matching processing unit 133 reads an original data characteristic signal to be searched from the original data characteristic signal storage unit 145, reads a broadcast characteristic signal to be searched from the broadcast characteristic signal storage unit 146, and outputs the original data characteristic signal and the broadcast characteristic. The signal and the signal are subjected to a matching process using a predetermined matching threshold. In the matching process, basically, as described above, the waveform or feature amount of the target time-series data (original data) and the time-series data (input data) to be detected are compared, and the similarity is constant. If the value is equal to or more than the value (matching threshold), it is assumed that the original data to be detected exists in the input data and the identity thereof is confirmed.

The appearance number determination unit 134 reads the time schedule from the time schedule storage unit 144 and compares the number of appearances of the original data on the schedule with the number of appearances of the original data by the matching processing of the matching processing unit 133. If the number of appearances of the original data by the matching process is less than the number of occurrences on the schedule, the matching threshold is changed to a low value in the matching processing unit, and the matching process is repeated. When the number of appearances is equal to or more than the number of appearances, the matching number and the position information are sent to the watermark extraction determination unit 136.

The similarity determination unit 135 reads the clustering information of the original data corresponding to the finally selected matching threshold from the original data similarity storage unit 147, and determines the data existing in the cluster as the original data. It is determined whether there is only one or similar data, and the result is sent to the watermark extraction determination unit 136.

In the watermark extraction determining section 136, the original data 1
If there is only one, the matching processing result of the matching processing unit 133 is output to the determination result output unit 121. When similar data exists, the digital watermark of the original data is read from the digital watermark storage unit 142, the broadcast signal is read from the broadcast signal storage unit 143, and the broadcast signal is read from the original data position information of the matching processing result of the matching processing unit 133. Of the original data read from the digital watermark storage unit 142, and identifies the original data in correspondence with the digital watermark of the original data, and determines the number of times of matching of the identified original data and the position information. Output to the result output unit 121.

The judgment result output section 121 outputs the judgment result to the outside. The control unit 138 controls and executes the processing of each unit.

When performing only a search for which the presence / absence of the original data in the input data is uncertain so as to detect any music from the broadcast signal, the time schedule storage section 144 and the number-of-appearance determination section 134 may be omitted. Good.

Next, a method of detecting time-series data according to the first embodiment of the present invention will be described. FIG. 2 shows the first embodiment of the present invention.
5A is a flowchart of a preparation stage of the time-series data detection method according to the embodiment, FIG. 4A is a flowchart of an original data-related material accumulation step, and FIG. 4B is a flowchart of a broadcast signal input step. FIG. 3 is a flowchart of a detection process of the time-series data detection method according to the first embodiment of the present invention.

First, the original data related material storing step will be described. If the original data that may be included in the input data to be detected, for example, the original data is a CM, all the original data of the CM that may be broadcast are registered. The original data-related materials may be created each time the original data to be detected is included, including data similar to the original data, or all the original data that may be broadcast, such as CM
May be created as a database, and a portion related to desired original data may be read and used each time.

As the original data-related data, the original data characteristic signal obtained by converting the original data into a characteristic signal, the electronic watermark of the original data, and all registered original data are subjected to matching processing with a plurality of predetermined threshold values. An original data similarity table is created based on the original data created by performing clustering and a threshold.

When the process is started (S101), the original data is input and stored in the original data storage unit 141 (S10).
2) The input original data is converted into a characteristic signal (S10).
3) and store it in the original data feature signal storage unit 145 (S10)
4) The same threshold value as the plurality of threshold values used in the matching process is set as a clustering threshold value, and all the original data are processed using the respective threshold values (S105) by the same matching processing method as the detection process. The similarity calculation is performed by matching the characteristic signals (S106), and the process is repeated for all thresholds (S107), and stored in the original data similarity storage unit 147 as an original data similarity table based on the original data and the thresholds. (S108). Next, digital watermark information of the original data is input (S109) and stored in the digital watermark storage unit 142 (S110). This processing is repeated for all the original data (S111), and necessary original data-related materials are completed (S112).

Next, the broadcast signal input step will be described.
When the process starts (S201), a broadcast signal is input (S201).
202), the input broadcast signal is stored in the broadcast signal storage unit 143 (S203), the input broadcast signal is converted into a characteristic signal to generate a broadcast characteristic signal (S204), and stored in the broadcast characteristic signal storage unit 146. (S205), a time schedule of the input broadcast signal is input (S206), stored in the time schedule storage unit 144 (S207), and a desired range is input (S208) to complete necessary broadcast signal related materials. (S209). Steps 206 and 207 may be omitted in some cases.

Next, the detection process of the time-series data detection method will be described. When the process is started (S301), a broadcast signal to be detected is selected (S302), and the broadcast characteristic signal of the target broadcast signal is stored in the broadcast characteristic signal storage unit 146.
(S303), and reads the time schedule of the broadcast signal to be detected from the time schedule storage section 144 (S304).

Next, original data to be detected is selected (S30).
5) Read the original data characteristic signal of the selected original data from the original data characteristic signal storage unit 145 (S306), read the digital watermark of the selected original data from the digital watermark storage unit 142 (S307), and determine the similarity of the original data. Storage unit 147
Read the similarity table of the original data selected from (S
308), the number of original data to be included in the broadcast signal is calculated from the time schedule (S309).

Next, a predetermined matching threshold value is selected from a plurality of matching threshold values (S310), and a matching process between the broadcast feature signal and the original data feature signal is performed using the matching threshold value. Is detected (S311). In this case, a matching processing method whose processing speed is improved as much as possible is used. Also, it is preferable that the matching threshold value selected first starts from the highest threshold value, but it is preferable to use the last final selection threshold value from the next time onward. If the detected number of original data is smaller than the expected number of original data included in the broadcast signal calculated from the time schedule (S312N), the process returns to step 310 to change the matching threshold value to a lower value and perform the matching process. If the detected number of original data is equal to or larger than the expected number of original data included in the broadcast signal calculated from the time schedule (S312Y), the process proceeds to step 313.

In step 313, the number of similar data in the cluster corresponding to the final matching threshold is read from the similarity table of the original data read in step 308, and if the similar data is not in the cluster (313Y),
Proceeding to step 318, the detected matching number and position are output as matching data (S318). If there is similar data (313N), the broadcast signal storage unit 143
(S314), marks the position of the original data in the detected broadcast signal (S315), and reads the digital watermark embedded in the scheduled position (S31).
6), collate with the digital watermark of the original data read in step 307, identify that there is original data at the matched position (S317), and output the identified number of original data and data position as matching data (S318). ).

If detection of other original data is to be continued, the flow returns to step S305 to repeat the processing (S319Y). If processing of all the original data is completed (S319N), the flow returns to step S302 to return to other processing. The process shifts to the detection of the broadcast signal (S320Y), and when the detection of all the broadcast signals is completed (S320N), the process ends (S32).
1).

In the time-series data detection method of the present invention, 1) a unique ID is embedded in the original data as digital watermark information so that all the original data in the broadcast signal can be identified by the digital watermark information; ) For the set of original data, calculate the similarity between all the original data, and in calculating this similarity, use the same method as that used in the matching process. Data whose similarity is equal to or higher than the clustering threshold is regarded as similar data. 3) A basic search is performed by a matching method using the similarity as a threshold, and digital watermarking is performed only on original data having similar data. 4) Matching processing with improved processing speed is used. For example, an active search method described in Japanese Patent Application No. 10-151723 is used. 5) The threshold value can be adaptively changed according to the processing content.
The adaptive change is, for example, a method of lowering a threshold value until the number of time-series data to be detected is known in advance if the number of detected time-series data is equal to or larger than the expected number of detections. Analyzing the threshold change data accumulated by repeating the detection, semi-automatically adjusting the threshold using the result, 7) When extracting the digital watermark, the time series to be scanned Since the position in the data is derived by the matching method, it can be easily verified, and it is possible to prevent the redundant processing at the time of extracting the watermark when only the digital watermark is used and to perform the matching only. In the case where is used, data retrieval in time-series data that can be distinguished from similar data, which is difficult, can be realized at relatively low cost and at high speed.

Next, two examples in which the specific embodiment of the method for detecting time-series data of the present invention can be applied with a relatively high effect will be described. As a first example, a case where a CM as a specific signal portion is searched from a broadcast signal such as a specific commercial (CM) in which the expected number of original data appearing in advance is known is described as a second example. A case will be described in which a specific signal, such as original music data, is searched for from a broadcast signal in which the possibility of occurrence of original data, such as original music data, which may be used, is unknown. FIG. 4 is a schematic diagram showing a flow of searching for a specific CM from a broadcast signal, and FIG. 5 is a schematic diagram showing a flow of searching for original music data from the broadcast signal.

First, the first embodiment will be described with reference to FIGS. 1, 2, 3 and 4. Certain commercials (C
When searching for a CM which is a specific signal portion from a broadcast signal such as M) in which the expected number of appearing original data is known, first, CM original data (original data used for broadcasting) Is recorded in the original data storage unit 141 via the original data input unit 111 in advance. Here, the original data of the CM recorded in the original data storage unit 141 is all CMs that may be broadcast.

The digital watermark information embedded in the original data of the CM is recorded in the digital watermark storage unit 142 via the digital watermark input unit 112 together with the embedding destination.

Then, the searched broadcast wave is received by the receiver serving as the broadcast signal input unit 113, and the broadcast contents are recorded in the broadcast signal storage unit 143.

The schedule for broadcasting the CM is recorded in the time schedule storage section 144 via the time schedule input section 114.

Next, the recorded CM original data and broadcast contents are converted into a characteristic signal by the characteristic signal processing section 131,
These are recorded in the original data characteristic signal storage unit 145 and the broadcast characteristic signal storage unit 146, respectively. Here, regarding the original data of the CM, the similarity between the original data of the CM is calculated using the original data feature signal, and based on the clustering threshold, the original data of the CM having similar characteristics are in the same cluster. Is performed, and is recorded in the original data similarity storage unit 147 together with the cluster information.

However, the clustering threshold here is set to the same value as the matching threshold used in the subsequent matching, and when a plurality of matching thresholds are used,
Clustering according to each is performed, and a plurality of pieces of clustering information are recorded.

The matching processing section 133 performs a matching process between the CM characteristic signal and the broadcast content characteristic signal by using a previously selected matching threshold value. After this processing, the number of appearances of the original data of the CM on the time schedule is compared with the number of appearances of the original data detected by the matching processing. If there is, the process proceeds to a specific phase of the original data of the CM according to the number of appearances and the position.

When the number of appearances of the matching processing result is less than the expected number of appearances in the time schedule, the matching threshold in the matching processing unit 133 is reduced, and the matching processing is performed again. This process is continued until the number of appearances in the result of the matching process becomes equal to or more than the number of appearances on the time schedule. When the number of occurrences exceeds the number of occurrences on the time schedule, the original data of the CM is identified by the number of occurrences and the position. Move to phase.

In the specific phase of the original data of the CM, the presence or absence of similar data of the original data in the cluster at the final matching threshold value is searched based on the clustering information recorded in the original data similarity storage unit 147. I do. If the original data of the CM existing in the same class as the original data of the CM to be examined is only the original data of the CM to be examined, the electronic watermark information is not extracted, and the number of the original data as the retrieval result is reduced. The data position is the final result.

If similar data exists in the cluster, the broadcast signal is read from the broadcast signal storage unit 143, the digital watermark embedded at the position of the original data as the search result is read, and stored in the digital watermark storage unit 142. The digital watermark of the obtained original data is read out and collated, and assuming that the original data exists at the same position, the number of the original data and the data position are identified to obtain the final result.

With respect to the specified original data of the CM, it is determined that the original data of the CM is present in the broadcast signal, and its position is reported.

The above processing is performed for each original data of the CM to be searched, the processing results are integrated, and the result is compared with the time schedule to determine whether the CM has been properly broadcast according to the time schedule.

Next, a second embodiment will be described with reference to FIG. 1, FIG. 2, FIG. 3, and FIG. When searching for original music data that may be used for broadcasting from a broadcast signal for which the possibility that the original data appears is unknown, first, the original music data (original music data used for broadcasting) Are recorded in the original data storage unit 141 via the original data input unit 111 in advance. Here, the original music data stored in the original data storage unit 141 is all music that may be used.

The digital watermark information embedded in the original music data is recorded in the digital watermark storage unit 142 via the digital watermark input unit 112 together with the embedding destination.

The broadcast wave to be searched is received by the receiver serving as the broadcast signal input unit 113, and the broadcast content is recorded in the broadcast signal storage unit 143.

Next, the recorded music original data and broadcast content data are converted into characteristic signals by the characteristic signal processing section 131, and are recorded in the original data characteristic signal storage section 145 and the broadcast characteristic signal storage section 146, respectively. Is done.

Here, the similarity between the original music data is calculated using the original music data characteristic signal, and based on the clustering threshold, the original music data having similar characteristics belong to the same cluster. A clustering process is performed so that the original data similarity storage unit 147 stores the cluster information together with the cluster information. However, the clustering threshold here is the same value as the matching threshold used in the subsequent matching processing.

The matching processing unit 133 performs a matching process between the original music data characteristic signal and the broadcast content characteristic signal using a predetermined matching threshold value. Move to a specific phase. In the first embodiment, the number of appearances of the time schedule is compared with the number of occurrences of the matching process.
In this embodiment, this comparison is not performed, and the number of appearances and the position of the original music data are searched based on the matching result based on a predetermined matching threshold.

In the specific phase of the original music data, based on the clustering information recorded in the original data similarity storage section 147, the presence or absence of similar data of the original data in the cluster at the matching threshold is searched. If the original data of the music existing in the same class as the original data of the music to be examined is only the original data of the music to be examined, the digital watermark information is not extracted, and the number of the original data as the search result is The data position is the final result.

If similar music data exists in the cluster, the broadcast signal is read from the broadcast signal storage unit 143, the digital watermark embedded in the position of the original data as the search result is read, and the digital watermark storage unit 142 is read. Is read out and compared with the digital watermark, and assuming that the original data exists at the same position, the number of the original data and the data position are identified to obtain the final result. The above processing is performed for all original data of music to be examined.

Next, a method for detecting time-series data and a device for detecting time-series data according to a second embodiment of the present invention will be described with reference to the drawings. FIG. 6 is a schematic block configuration diagram of a time-series data detection device according to the second embodiment of this invention.

FIG. 6 shows a time-series data detecting device 2 according to the present invention.
00 is shown as a computer constituting the apparatus, and the computer is an input device 210 such as a modem, a keyboard, a pointing device, etc .; an output device 220 such as a modem, a printer, a display; a data processing device 230; A medium 250 is provided.
The recording medium 250 stores a time-series data detection system control program of the present invention capable of controlling the operation of each unit, and uses an FD, a CD-ROM, a semiconductor memory, or the like.

The configuration of the time-series data detecting device and the method for detecting the time-series data are the same as those in the first embodiment, and therefore the description thereof is omitted.

The control program of the time series data detection system for searching for the time series data contained in the input broadcast signal is read from the recording medium 250 into the data processing device 230 and controls the operation of the data processing device 230. The data processing device 230 executes the following processing under the control of the control program.

That is, the time-series data (original data) to be searched is input, converted into a characteristic signal, the original data characteristic signal is stored, and all the original data characteristic signals are matched using the threshold value. Perform similarity calculation and store as original data similarity table, input and store digital watermark information of original data, complete processing of necessary original data related materials, input broadcast signal and perform feature signal processing The broadcast feature signal is stored, the time schedule of the broadcast signal is input and stored, the process of completing the necessary broadcast signal related materials is performed, the broadcast signal to be searched is selected, and the broadcast feature signal of the broadcast signal is selected. Read, read the time schedule, select the original data to be searched, read the original data feature signal,
Processing for reading the digital watermark, reading the similarity table of the original data, calculating the number of original data to be included in the broadcast signal from the time schedule, and using the matching threshold value for the broadcast characteristic signal and the original data characteristic signal And if the number of detected original data is less than the number of original data to be included in the broadcast signal calculated from the time schedule, the detected number of original data is calculated from the time schedule. A process in which the threshold value is changed to a low value until the number of original data to be included in the calculated broadcast signal is equal to or more than a predetermined value, and the matching process is repeated. By doing so, the matching threshold value corresponding to the matching process is appropriately set, and the number of data and the data position of the original data in the input data are detected. If there is no similar data in the cluster corresponding to the final matching threshold, the number of detected original data and the data position are output as matching data, and if there is similar data, the original data in the broadcast signal is output. A process of reading out the digital watermark embedded at the detected position and comparing it with the digital watermark of the original data, identifying that there is original data at the matched position, and outputting the number of original data and the data position as matching data; Execute

[0070]

As described above, the present invention has an effect that time-series data to be detected can be quickly and reliably detected in all regions of the time-series data.
That is, the similarity between all the original data is calculated in correspondence with the matching threshold value of the matching process, and if there is similar data at the matching threshold level of the matching process between the input data and the original data, it is detected. The matching between the digital watermark read from only the original data position in the input data and the digital watermark of the original data is performed to determine the matching data, thereby compensating for the uncertainty in the matching process and reducing the processing time in the digital watermarking process. This is because shortening can be performed.

[Brief description of the drawings]

FIG. 1 is a block diagram of a time-series data detection device according to a first embodiment of the present invention.

FIG. 2 is a flowchart of a preparation stage of the time-series data detection method according to the first embodiment of the present invention. (A) is a flowchart of an original data related material accumulation process. (B)
Is a flowchart of a broadcast signal input step.

FIG. 3 is a flowchart of a detection process of the time-series data detection method according to the first embodiment of the present invention.

FIG. 4 is a schematic diagram showing a flow of searching for a specific CM from a broadcast signal.

FIG. 5 is a schematic diagram showing a flow of searching for original music data from a broadcast signal.

FIG. 6 is a schematic block diagram of a time-series data detection device according to a second embodiment of the present invention.

FIG. 7 is a schematic diagram for explaining a problem of a detection method by matching in which a CM submission confirmation is used as an example. (A) shows a state before detection. (B) shows the state after detection.

[Explanation of symbols]

71 Set of input data 72 Set of CMs to be detected 73 CM detected by matching 74 CM detected correctly 75 CM not detected 76 CM detected erroneously 100, 200 Time-series data detection device 111, 211 Original Data input unit 112, 212 Digital watermark input unit 113, 213 Broadcast signal input unit 114, 214 Time schedule input unit 121, 221 Judgment result output unit 131, 231 Feature signal processing unit 132, 232 Similarity calculation unit 133, 233 Matching Processing units 134, 234 Appearance number determination unit 135, 235 Similarity determination unit 136, 236 Watermark extraction / determination unit 138, 238 Control unit 141, 241 Original data storage unit 142, 242 Digital watermark storage unit 143, 243 Broadcast signal storage unit 144, 244 Times Joule storage unit 145, 245 the original data, wherein the signal storage unit 146, 246 broadcast feature signal storage unit 147,247 raw data similarity storage unit 250 recording medium S101~S112, S201~S209, S301~
S321 Step

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Naozumi Tsuchida 3-19-2 Nishi-Shinjuku, Shinjuku-ku, Tokyo Japan Telegraph and Telephone Corporation (72) Inventor Yasuhiro Hirano 3-192-1, Nishi-Shinjuku, Shinjuku-ku, Tokyo No. Nippon Telegraph and Telephone Corporation F term (reference) 5B075 ND12 NK06 NK39 PQ02 PR06 QM01 QM08 5C063 AB03 AB07 CA23 CA34 DA07 5L096 BA18 DA01 GA51 HA08 JA03 JA11 MA07 9A001 EE03 FF03 HH35 KK43 KK60 KK03

Claims (6)

[Claims] 1. The time series data to be searched (hereinafter referred to as
A method for detecting time-series data, which detects original data) from input time-series data (hereinafter referred to as input data), wherein the original data and the input data can be uniquely identified. The original data to be searched is input, the feature signal is processed and stored as an original data feature signal, and all the original data feature signals are determined using a predetermined threshold. Perform matching processing, calculate similarity, store as original data similarity table, input and store digital watermark information of the original data, complete necessary original data related materials, and input the input data A feature signal is processed and stored as an input data feature signal to complete necessary input data related data. A predetermined operation is performed between the input data feature signal and the original data feature signal. Performing a matching process using a threshold to detect the number of data and the data position of the original data in the input data; and when there is similar data in a cluster corresponding to the threshold in the original data similarity table. Reads out the digital watermark embedded in the detected original data position in the input data, compares it with the digital watermark information of the original data, eliminates similar data, and removes the original data in the input data. A method for detecting time-series data, wherein the number of data and the data position are determined. 2. A time schedule indicating a planned positional relationship of the original data in the input data is stored in the input data-related material, and the number of original data detected in the matching process is calculated from the time schedule. If the number of original data to be included is smaller than the number of original data to be included in the input data, the matching is performed until the number of detected original data is equal to or greater than the number of original data to be included in the input data calculated from the time schedule. 2. The time-series data detection method according to claim 1, wherein a threshold value is changed to a low value, and a matching process is repeatedly performed to detect a data number and a data position of the original data in the input data. Further, by accumulating and analyzing the result of the matching processing and the data of the threshold value,
The method of detecting time-series data according to claim 1, wherein the matching threshold value corresponding to the matching process is appropriately set, and the number of data and the data position of the original data in the input data are detected. 4. A time-series data detecting device for detecting time-series data (original data) to be searched from input time-series data (input data), comprising: an original data input unit, an original data storage unit, Watermark input unit and digital watermark storage unit, input data input unit and input data storage unit, time schedule input unit and time schedule storage unit, feature signal processing unit, original data feature signal storage unit, input data feature signal storage unit, and similar A degree calculation unit, an original data similarity storage unit, a matching processing unit, an appearance number determination unit, a similarity determination unit, a watermark extraction / determination unit, a determination result output unit, and a control unit. Inputting data, storing the data in the original data storage unit, and inputting the digital watermark data embedded in the original data into the digital watermark data input unit, and embedding the digital watermark data The input data input unit inputs the input data and stores it in the input data storage unit, and the time schedule input unit inputs the time schedule of the input data. The feature signal processing unit performs feature signal processing on the original data to generate an original data feature signal, stores the original data feature signal in the original data feature signal storage unit, and stores the input data. The input data feature signal is generated by performing a feature signal conversion process on the input data feature signal storage unit, and stored in the input data feature signal storage unit. Calculate the degree of similarity between them, and perform clustering processing so that the original data with similar characteristics are in the same cluster. The original data characteristic signal and the input data characteristic signal are stored in the original data similarity storage unit using a predetermined threshold, and the matching processing unit performs a matching process. The scheduled number of appearances on the time schedule of the original data is compared with the number of appearances of the original data by the matching process, and the threshold value of the matching processing unit is changed as necessary. The clustering information of the original data corresponding to the finally selected threshold value is read, and whether there is only one original data in the cluster is determined.
It is determined whether or not similar data exists, and the result is sent to the watermark extraction determining unit. In the watermark extraction determining unit, if there is only one original data, the matching processing result of the matching processing unit is output to the determination result output unit. And if similar data exists, read out the digital watermark of the original data, read out and verify the digital watermark embedded in the position of the original data as a result of the matching processing of the input data, Outputting the number of times of data matching and position information to the judgment result output unit, the judgment result output unit outputting the judgment result to the outside, and the control unit controlling and executing the processing of each unit. Sequence data detection device. 5. The time-series data detection device according to claim 4, further comprising a recording medium, wherein the operation of the control unit can be controlled by a time-series data detection system control program recorded on the recording medium. 6. A recording medium on which a control program for detecting time-series data (original data) to be searched from input time-series data (input data) is recorded, wherein the original data to be searched is recorded. Is input, processed as a feature signal, stored as an original data feature signal, and all original data feature signals are subjected to a matching process using a predetermined threshold to calculate similarity, and stored as an original data similarity table. Inputting and storing the digital watermark information of the original data; and inputting the input data, converting the input data into a characteristic signal, storing the input data as a characteristic signal, and the expected positional relationship of the original data in the input data. And storing a time schedule indicating the following, and performing a matching process between the input data feature signal and the original data feature signal using a predetermined threshold value. Performing the detection of the number of data and the data position of the original data in the input data, and the number of the original data detected in the matching process is to be included in the input data calculated from the time schedule. If the number of original data is less than the number of original data, the matching threshold is changed to a lower value until the number of detected original data is equal to or more than the number of original data to be included in the input data calculated from the time schedule. A procedure of repeatedly performing a matching process to detect a data number and a data position of the original data in the input data, and accumulating and analyzing a result of the matching process and data of the threshold value, The matching threshold value corresponding to the matching process is appropriately set, and the number of data and the data position of the original data in the input data are detected. If there is similar data in the cluster corresponding to the threshold value of the original data similarity table, the electronic watermark embedded in the detected original data position in the input data is read and the original data is read. A step of determining the number of data and the data position of the original data in the input data by excluding similar data by collating with the digital watermark information, and a machine-readable recording recording a program for executing Medium.