JP2000347689A - Method and system for retrieving sound data, and recording medium recorded with sound data retrieving program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce useless collation processing to enhance efficiency for retrieval, by determining discriminatingly a portion necessary to be retrieved from an unnecessary portion out of input sound data. SOLUTION: This system is provided with a separator detecting means 201 for detecting a data consistent with separator information stored in a separator information storing part 207 based on a sound data accumulated in an input sound data accumulating part 206 to divide the sound data in a coinciding point, and for storing addresses of the two separator informations existing in a front side and a rear side of the divided sound information data into a collation-objective sound data storing part 209, and a data length comparing means 202 for calculating a data lenght of the divided sound data based on the addresses of the two separator informations stored in the data storing part 209 so as to confirm the consistency with a resistered data length stored in sound data storing part 208. Collation processing is conducted when the consistency is confirmed as a result of the confirmation by the data length comparing means 202.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a sound data search method, a sound data search device, and a recording medium on which a sound data search program is recorded. Sound data search method to reduce unnecessary collation processing and improve search efficiency,
The present invention relates to a sound data search device and a recording medium storing a sound data search program. It is assumed that the sound data includes not only voices but also all sounds such as music and chime sounds.

[0002]

2. Description of the Related Art Conventionally, when collating and searching the same sound data as input sound data, the input sound data is divided into sound data at regular intervals (for example, every 1/100 second), and then divided. The divided sound data is converted into spectrum data and collated with all the candidate sound data stored in advance. In this case, when the number of candidate sound data increases, the matching processing amount also increases, so that the matching processing time also increases. For example, sound data of 1/100 second cannot be processed within 1/100 second, and processing delay is reduced. There has been a problem that the data may be accumulated and search may not be possible (hereinafter referred to as a first conventional technique).

[0003] As a technique for solving the above-mentioned problem in the first conventional technique, a speech recognition method and apparatus disclosed in Japanese Patent Laid-Open No. 5-204396 (hereinafter referred to as a second conventional technique). There is. According to the second conventional technique, preliminary selection of candidate sound data is performed by obtaining a feature vector of input sound data at predetermined time intervals, and a collation process is performed when the number of selected candidate sound data falls below a predetermined number. , The amount of collation processing is reduced.

[0004]

In the first and second prior arts described above, the collation processing is performed on all the input sound data. In the second prior art, the candidate sound data is compared. Although the amount of collation processing is reduced by narrowing down the number, the amount of collation processing cannot be expected to be so large in relation to the number of candidate sound data.

The present invention focuses on input sound data,
It is an object of the present invention to determine a part of the sound data that requires a search and a part that does not need a search, thereby reducing unnecessary collation processing and improving the efficiency of the search.

[0006]

According to a first sound data search method of the present invention, input sound data is compared with previously registered and stored sound data, and is compared with the inputted sound data. A sound data retrieval method for retrieving sound data from the registered and stored sound data, the method comprising: separating separator information serving as a division point for dividing the input sound data; and at least one designated time value. Registered and stored in advance, the input sound data is divided by the separator information, and when the length of the divided sound data matches one of the designated time values, the registered and stored sound data is stored. It is characterized in that comparison and collation with sound data are performed.

In a second sound data search method according to the present invention, a separator information serving as a dividing point for dividing inputted sound data and at least one designated time value are registered and stored in advance. Step, a second step of storing the input sound data, a third step of dividing the input sound data by the separator information, and a length of the sound data divided in the third step. A fourth step of calculating the length of the sound data, a fifth step of comparing and determining whether the length of the sound data calculated in the fourth step matches any of the designated time values, As a result of the above step, if they match, the divided sound data is converted to frequency component spectrum data, and the same sound data is collated with previously registered and stored spectrum data. A sixth step of search, the fifth step results if they do not match, the process ends without performing the matching of the divided sound data,
And a seventh step of returning to the third step.

[0008] A third sound data search method according to the present invention is characterized in that, in the first or second sound data search method according to the present invention, the separator information is a silent portion that has continued for a predetermined time or more.

In a fourth sound data search method according to the present invention, in the first or second sound data search method according to the present invention, the separator information includes a bell sound, a time signal, a chime, a game start pistol sound, It is characterized by being sound data having a certain characteristic such as a part of a music piece.

The first sound data search device of the present invention compares and compares input sound data with previously registered and stored sound data, and registers the same sound data as the inputted sound data. What is claimed is: 1. A sound data search device for searching among stored sound data, wherein means for pre-registering and storing at least one designated time value and separator information serving as a dividing point for dividing the inputted sound data. Means for dividing the input sound data according to the separator information; means for determining whether the length of the divided sound data matches any of the designated time values. And means for comparing and collating with the stored sound data.

A second sound data retrieval apparatus according to the present invention pre-registers and stores an input sound data storage section for storing sound data inputted from the outside, and separator information serving as division points for extracting sound data. A separator information storage unit, a sound data length storage unit that pre-registers and stores the length of input sound data to be subjected to the matching processing, and a sound data length stored in the sound data length storage unit. A collation target sound data storage unit for storing divided sound data having a matching length and addresses of two pieces of separator information before and after the sound data, and sound data stored in the collation target sound data storage unit A spectrum data storage unit that stores spectrum data obtained by converting the spectrum data into frequency components as extracted spectrum data, and a data group to be compared is stored in a frequency component candidate spectrum. A candidate spectrum data storage unit stored in advance in association with an identification number as data, and search result information such as a title and sound data corresponding to each candidate spectrum data stored in the candidate spectrum data storage unit. A search result information storage unit stored in advance in association with the identification number of the candidate spectrum data storage unit, and sound data stored in the input sound data storage unit, based on the sound data stored in the separator information storage unit. The sound data is divided at the coincidence point by detecting the data that matches the separator information, and the addresses of two pieces of separator information before and after the divided sound data are stored in the matching target sound data storage unit. Separator detecting means for notifying the length comparing means that separator information has been detected, and the separator detecting means The data length of the divided sound data is calculated from the addresses of the two pieces of separator information stored in the comparison target sound data storage unit, and whether the registered data length matches the registered data length stored in the sound data length storage unit is determined. Confirming, if matched, storing the sound data in the matching target sound data storage unit, and a data length comparing means for notifying the spectrum converting means that there is the matching target sound data; and The sound data stored in the storage unit is converted into spectrum data of frequency components and stored as extracted spectrum data in the spectrum data storage unit.
04, and a search for data matching the extracted spectrum data stored in the spectrum data storage unit from the candidate spectrum data stored in the candidate spectrum data storage unit. Sound data matching means for notifying the search result management means of the identification number of the candidate spectrum data, and search result information such as sound data or commercial title corresponding to one candidate spectrum data searched by the sound data matching means, A search result management unit that searches by the identification number from the search result information storage unit and outputs the search result to an external search result output device.

According to a third sound data search device of the present invention, in the first or second sound data search device of the present invention, the separator information is a silent portion that has continued for a predetermined time or more.

A fourth sound data search device according to the present invention is the first or second sound data search device according to the present invention, wherein the separator information includes a bell sound, a time signal, a chime, a start pistol sound of a game, It is characterized by being sound data having a certain characteristic such as a part of a music piece.

According to the recording medium of the present invention, first, separator information as a division point for dividing inputted sound data and at least one designated time value are registered and stored in advance.
Processing, a second processing for accumulating the input sound data, a third processing for dividing the input sound data by the separator information, and a processing of the sound data divided in the third processing. A fourth process for calculating the length, a fifth process for comparing and determining whether the length of the sound data calculated in the fourth process matches any of the designated time values, If the result of the processing in step 5 is a coincidence, the divided sound data is converted into spectrum data of frequency components, and the same sound data is searched by collating with the previously registered and stored spectrum data. And, if the result of the fifth process does not match, the seventh process that ends the process without collating the divided sound data and returns to the third process. Sound data detection to be executed by the computer And characterized by recording a program.

[0015]

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

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention. A sound data input device 1 such as a TV tuner, a sound data search device 2, and a search for a speaker, a sound board, a CRT, etc. The result output device 3 is shown. Here, the sound data includes not only voices but also all sounds such as music and chime sounds.

The sound data search device 2 includes a separator detection means 201, a data length comparison means 202, a spectrum conversion means 203, a sound data comparison means 204, a search result management means 205, an input sound data storage section 206, , Separator information storage unit 207 and sound data length storage unit 208
, A matching target sound data storage unit 209, a spectrum data storage unit 210, and a candidate spectrum data storage unit 211.
And a search result information storage unit 212.

20 provided in the sound data search device 2
Of the means 1-22, the storage unit and the storage unit,
The storage units and storage units 6 to 212 will be described.

The input sound data storage unit 206 stores the sound data input from the sound data input device 1, and has a capacity for storing at least the longest sound data registered in the sound data length storage unit 208. (As shown in FIG. 4, “14.5 seconds to 15.
In the case where two types of sound data lengths of “5 seconds” and “29.5 seconds to 30.5 seconds” are registered and stored, for example, a capacity of about 40 seconds is considered in consideration of silence parts and errors. ) Is stored as the sound data is sequentially input from the sound data input device 1. Since the storage capacity is limited, the storage time is limited within the range of the storage capacity. Newly input sound data is sequentially stored while deleting previously stored sound data.

The separator information storage unit 207 pre-registers and stores separator information serving as division points for extracting sound data. For example, as shown in FIG.
If “silence for more than seconds” is registered as separator information, it means “silence for more than A seconds is a division point.” That is, a silent part that continues for more than A seconds in the input sound data. Is detected, sound data is divided and extracted with the silent part as a delimiter.

The sound data length storage unit 208 previously registers and stores the length of input sound data to be subjected to the collation processing. For example, as shown in FIG.
Two sound data lengths of “5 seconds” and “29.5 to 30.5 seconds” are stored.
13. Since the data length of the sound data of the commercial to be extracted is 15 seconds or 30 seconds, some errors are taken into account.
5 seconds to 15.5 seconds, and 29.5 seconds to 30.5 seconds. Therefore, the data length of the sound data is from 14.5 seconds to 15.
Only in the case of 5 seconds or 29.5 to 30.5 seconds, the sound data is recognized as the sound data to be subjected to the matching processing.

The collation target sound data storage unit 209 stores the divided sound data (the sound data to be subjected to the collation processing) having the length corresponding to the sound data length stored in the sound data length storage unit 208 and this sound. The addresses of two pieces of separator information before and after the data are stored.

The spectrum data storage unit 210 applies Fourier transform to the sound data stored in the comparison target sound data storage unit 209, and stores the spectrum data converted into frequency components as extracted spectrum data.

The candidate spectrum data storage unit 211 stores in advance data groups to be collated as candidate spectrum data of frequency components. For example, as shown in FIG. 5, candidate spectrum data obtained by converting sound data of first to fourth television commercials into frequency components are stored in association with identification numbers 01 to 04, respectively.

The search result information storage unit 212 stores search result information corresponding to each spectrum data stored in the candidate spectrum data storage unit 211 in advance. For example, as shown in FIG. 6, titles and sound data of the first to fourth commercials are stored in association with identification numbers 01 to 04 of the candidate spectrum data storage unit 211, respectively.

Next, 201 to 20 of the sound data search device 2
5 will be described.

The separator detecting means 201 detects data matching the separator information stored in the separator information storage section 207 based on the sound data stored in the input sound data storage section 206, and detects the sound data at the matching point. Split. In the present embodiment, a silent part is detected, and sound data is divided using the silent part as a boundary.

The data length comparing means 202 checks whether the data length of the sound data divided by the separator detecting means 201 matches the registered data length stored in the sound data length storage section 208, and if the data lengths match. Stores the sound data in the matching target sound data storage unit 209. In the present embodiment, when the data length of the divided sound data is 14.5 seconds to 15.5 seconds or 29.5 seconds to 30.5 seconds, the matching sound data is stored in the matching target sound data storage unit 209. It is stored as target sound data.

The spectrum converting means 203 converts the sound data stored in the matching target sound data storage section 209 into frequency component spectrum data, and stores it in the spectrum data storage section 210 as extracted spectrum data.

The sound data collating means 204 stores data matching the extracted spectrum data stored in the spectrum data storage section 210 into the candidate spectrum data storage section 2.
A search is performed from the candidate spectrum data stored in No. 11.

The search result management unit 205 searches the search result information storage unit 212 for the sound data or commercial title corresponding to one candidate spectrum data searched by the sound data matching unit 204 by using the identification number. Output to the output device 3.

FIG. 2 is a flowchart showing the operation of the embodiment of the present invention. FIG. 3 shows an example of the separator information registered in the separator information storage unit 207, and FIG. 4 shows an example of the data length registered in the sound data length storage unit 208.
FIG. 5 is a diagram illustrating an example of candidate spectrum data registered in the candidate spectrum data storage unit 211, and FIG. 6 is a diagram illustrating an example of search result information registered in the search result information storage unit 212, respectively.

FIG. 7 is a diagram showing an example of input sound data. The input sound data includes CM1 (first television commercial) for 15 seconds, ETC1 (non-CM part) for 300 seconds, and CM2. (Second television commercial) includes sound data for 30 seconds, and a boundary of each sound data includes a silent portion 1, a silent portion 2, a silent portion 3, and a silent portion 4 serving as separator information. ing. FIG. 8 is a conceptual diagram of a waveform of input sound data.

Next, the operation of the embodiment of the present invention will be described in detail with reference to FIGS.

The present embodiment will be described on the assumption that the time setting and management means (not shown) is activated at a preset first time and ends at a preset second time.

First, at the start time, sound data is supplied from the sound data input device 1, for example, a television tuner, to the input sound data storage unit 206, and the separator detecting means 201 is activated.
From the separator information storage unit 207, separator information, that is, delimiter information relating to sound data division points is obtained. In the present embodiment, as shown in FIG. 3, "silence for more than A seconds" is registered and stored as separator information in the separator information storage unit 207. Is the separator information. Then, the separator detection unit 201 searches for a silent part in the sound data stored in the input sound data storage unit 206. At this time, a search is made from the beginning of the data, and a silent portion (silent portion 1 in FIG. 7) in which silence continues for A seconds or more is detected. (Step A1 in FIG. 2).

As a result of the search in step A1, if two pieces of separator information, that is, two silent sections are detected, the separator detecting means 201 determines whether the two detected silent sections (for example, the silent section 1 and the silent section 2 in FIG. (1 set) are stored in the comparison target sound data storage unit 209, and the data length comparing unit 202 is notified that a silent section has been detected. In addition,
At this time, if the same set of addresses has already been stored in the collation target sound data storage unit 209, the separator detection unit 201 stores the collation target sound data storage unit until a silence part having a different set of addresses is detected. The address storage to the section 209 and the notification to the data length comparing means 202 that the silent section has been detected are not performed.

The address of the silent section may be the start address, the end address, or the intermediate address of the silent section. There is no particular limitation, but it is assumed that which section is to be set in the separator detecting means 201 in advance. In addition, the two (one set) silence parts, for example, the silence part 1 in FIG.
And the address of the silent section 2 are stored in the comparison target sound data storage section 209.
When the search is completed and the next silent part detection is performed to detect the silent part 3 and the silent part 4 in FIG. 7, a set of addresses of the silent part 3 and the silent part 4 is As a result of overwriting the set of the silent part 1 and the silent part 2 stored at the time of the previous detection of the silent part, the address stored in the matching target sound data storage unit 209 becomes the latest set (step A2). → A3).

As a result of the search in step A1, if no separator information, that is, two (one set) of silence parts are not detected, the separator detection means 201 continues to search for silence parts. For example, when the ETC1 shown in FIG. 7 is input to the input sound data storage unit 206, the memory capacity of the input sound data storage unit 206 is about 40 seconds as described above, and the ETC1 shown in FIG. Since the sound data length is 300 seconds, the silent part 2 and the silent part 3 cannot be detected as one set. In such a case, the separator detecting means 201 continues the search for the silent part. I do. That is, in the present embodiment, in order to extract and retrieve a commercial of 15 seconds or 30 seconds from the input sound data, the sound data length storage unit 208 stores "14.5 to 1".
“5.5 seconds” and “29.5 seconds to 30.5 seconds” are registered and stored, and a data length exceeding 30.5 seconds is not extracted (step A2 → A1).

Upon receiving a notification from the separator detecting means 201 that a silent section has been detected, the data length comparing means 202 causes the data length registered and stored in the sound data length storing section 208, ie, "14. 5 seconds to 15.5
Second "," 29.5 to 30.5 seconds ", and the 1 stored in the matching target sound data storage unit 209.
A set of addresses (for example, silence 1 and silence 2 shown in FIG. 7)
And the data length of the sound data portion (CM1 shown in FIG. 7) sandwiched between both addresses is calculated based on both the obtained addresses. Then, the calculated sound data length and “14.5 seconds to 15.5 seconds”, “29.5 seconds to 3 seconds”
0.5 seconds ". If the calculated sound data length is 15 seconds," 14.5 seconds to 15.5 "registered and stored in the sound data length storage unit 208. Seconds ”, the data length comparison unit 202 extracts the corresponding sound data from the input sound data storage unit, stores it in the comparison target sound data storage unit 209, and sends the comparison target sound data to the spectrum conversion unit 203. A notification that there is data is provided (step A4).

As a result of the comparison in step A4, if the calculated sound data length does not match "14.5 seconds to 15.5 seconds" or "29.5 seconds to 30.5 seconds", Returning to step A1, the separator detecting means 201 continues to search for a silent part (step A4 → A1).

Upon receiving a notification from the data length comparing means 202 that there is sound data to be compared, the spectrum converting means 203 acquires the extracted sound data stored in the sound data storing section 209 and performs Fourier transform. Is converted into spectrum data which is data of frequency components (spectral data 1 in FIG. 8), and this is used as extracted spectrum data as a spectrum data storage unit 2.
10 and activate the sound data collating means 204 (step A5).

The sound data collating means 204 acquires the extracted spectrum data stored in the spectrum data storage section 210, and stores the first candidate spectrum data (1 in FIG. 5) stored in the candidate spectrum data storage section 211 in advance. (The spectrum data of the commercial) (step A6).

In step A6, when the spectrum data matches, information indicating that the first candidate spectrum data matches, that is, the identification number shown in FIG.
"01" is supplied to the search result management means 205. The search result management means 205 stores the sound data and title (identification number "01" in FIG. 6) stored in the search result information storage unit 212 and having the identification number "01" shown in FIG. 1 and outputs the title to the search result output device 3. The search result output device 3 displays the supplied title on a CRT screen or the like, and reproduces the sound data (step A7 → A8).

If the data does not match the first candidate spectrum data at step A6, it is compared with the second candidate spectrum data (the second commercial spectrum data in FIG. 5). If it does not match the second candidate spectrum data, the steps A7 → A9 → A10 are repeated sequentially from the third and fourth steps, and all the candidate spectrum data stored in the candidate spectrum data storage unit 211 (FIG. 5 does not match even if compared with the first to fourth commercial spectrum data shown in FIG.
1 continues the search for the silent part (step A9 → A
1) If a matching candidate spectrum data is found on the way, the same operation as the above-described step A7 → A8 is performed (step A7 → A8).

Next, another embodiment of the present invention will be described in detail with reference to FIG.

FIG. 9 is a block diagram showing the configuration of another embodiment of the present invention, which comprises a sound data input device 1, a search result output device 3, a computer 4, and a recording medium 5. . The configuration of the computer 4 is basically the same as the sound data search device 2 described in the embodiment of the present invention in FIG. The recording medium 5 stores a sound data search program. The recording medium 5 may be a magnetic disk, an optical recording disk, a semiconductor memory, or another recording medium. The sound data search program is read from the recording medium 5 into the computer 4, and controls the same operation as in the embodiment of the present invention.

In the above description, the case where the TV tuner is used as the sound data input device has been described. However, a microphone, a sound board, a radio tuner may be used, or a network, a telephone line, or the like may be used. It is clear that the present invention can be applied even if the sound data is used, and there is no restriction on the type of the sound data.

In the above description, a silent portion is used as separator information. However, the present invention is not limited to this. The sound of a bell, a time signal, a chime, the sound of a start pistol of a competition,
Any sound data, such as a part of a music piece, having certain characteristics can be used.

Further, in the above description, the memory capacity of the input sound data storage unit 206 has been described as about 40 seconds, but the value of this memory capacity may be any value.
Assuming that the memory capacity is, for example, 600 seconds, E shown in FIG.
The data of TC1 (300 seconds) is also stored in the input sound data storage unit 20.
6, a set of silence portions 2 and 3 is detected, and it is determined whether or not the set value matches the time value registered in the sound data length storage portion 208.

[0051]

According to the present invention, the length of the input sound data divided by the separator information is set in advance by setting the separator information as the division point of the input sound data and the length of the sound data to be searched. Since the matching process is performed by determining that the sound data needs to be searched when the value matches the searched value, it is possible to reduce unnecessary use of the matching process and improve the efficiency of the search.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of an embodiment of the present invention.

FIG. 2 is a flowchart showing the operation of the embodiment of the present invention.

FIG. 3 is a diagram illustrating an example of separator information registered in a separator information storage unit 207.

FIG. 4 is a diagram showing an example of a data length registered in a sound data length storage unit 208.

FIG. 5 is a diagram showing an example of candidate spectrum data registered in a candidate spectrum data storage unit 211.

FIG. 6 is a diagram illustrating an example of search result information registered in a search result information storage unit 212;

FIG. 7 is a diagram showing an example of input sound data.

FIG. 8 is a conceptual diagram of a waveform of input sound data.

FIG. 9 is a block diagram showing a configuration of another embodiment of the present invention.

[Explanation of symbols]

 Reference Signs List 1 sound data input device 2 sound data search device 201 separator detection means 202 data length comparison means 203 spectrum conversion means 204 sound data collation means 205 search result management means 206 input sound data storage section 207 separator information storage section 208 sound data length storage section 209 Reference target sound data storage unit 210 Spectrum data storage unit 211 Candidate spectrum data storage unit 212 Search result information storage unit 3 Search result output device 4 Computer 5 Recording medium

Claims (9)

[Claims] 1. A method of comparing input sound data with previously registered and stored sound data and comparing the same to search for the same sound data from the registered and stored sound data. In the sound data search method, separator information serving as a division point for dividing the input sound data and at least one designated time value are registered and stored in advance, and the input sound data is stored in the sound data search method. Sound data divided according to separator information, and when the length of the divided sound data matches any of the designated time values, comparison and collation with the registered and stored sound data are performed. retrieval method. 2. A first step of pre-registering and storing separator information serving as a division point for dividing input sound data and at least one designated time value; A second step of accumulating; a third step of dividing the input sound data based on the separator information; a fourth step of calculating a length of the sound data divided in the third step; A fifth step of comparing and determining whether the length of the sound data calculated in the fourth step matches one of the designated time values; and, if the result of the fifth step is a match, A sixth step of converting the divided sound data into spectrum data of frequency components, searching for the same sound data by comparing the spectrum data with pre-registered and stored spectrum data, If the result of the fifth step does not match,
And a seventh step of terminating the process without checking the divided sound data and returning to the third step. 3. The apparatus according to claim 1, wherein the separator information is a silent portion that has continued for a predetermined time or more.
The described sound data search method. 4. The apparatus according to claim 1, wherein the separator information is sound data having certain characteristics such as a bell sound, a time signal, a chime, a start pistol sound of a game, and a part of a music piece. The described sound data search method. 5. The registered sound data is compared with the previously registered and stored sound data, and the same sound data as the inputted sound data is searched from the registered and stored sound data. A sound data search device for pre-registering and storing separator information serving as a division point for dividing the input sound data and at least one designated time value; Means for dividing according to the separator information, means for determining whether or not the length of the divided sound data matches any of the designated time values. A sound data search device comprising: means for performing collation. 6. An input sound data storage unit for storing sound data input from the outside, a separator information storage unit for pre-registering and storing separator information as a division point for extracting sound data, The length of the input sound data to be processed is registered in advance,
A stored sound data length storage unit, divided sound data having a length corresponding to the sound data length stored in the sound data length storage unit, and addresses of two separator information before and after the sound data. A comparison target sound data storage unit that stores the spectrum data obtained by converting the sound data stored in the comparison target sound data storage unit into frequency components as extracted spectrum data; A candidate spectrum data storage unit in which a target data group is stored in advance as frequency component candidate spectrum data in association with an identification number, and corresponding to each candidate spectrum data stored in the candidate spectrum data storage unit. Search result information such as title and sound data to be assigned to the identification number of the candidate spectrum data storage unit, respectively. Based on the search result information storage unit pre-stored in association with the sound data stored in the input sound data storage unit,
Data matching the separator information stored in the separator information storage unit is detected, sound data is divided at the coincidence point, and the addresses of two pieces of separator information before and after the divided sound data are used as the matching target sound. Along with storing the data length in the data storage unit, the separator length detection unit that notifies the data length comparison unit that the separator information has been detected, and the data length of the sound data divided by the separator detection unit is stored in the matching target sound data storage unit. It is calculated from the addresses of the two stored pieces of separator information, and it is checked whether the registered data length matches the registered data length stored in the sound data length storage section. Data length comparing means for notifying the spectrum conversion means that there is sound data to be compared, A spectrum conversion unit that converts sound data stored in the sound data storage unit into spectrum data of frequency components, stores the extracted spectrum data in the spectrum data storage unit, and activates a sound data comparison unit 204; Data matching the extracted spectrum data stored in the data storage unit is searched from the candidate spectrum data stored in the candidate spectrum data storage unit, and the identification number of the searched candidate spectrum data is used as a search result management unit. A search result information such as a sound data or a commercial title corresponding to one candidate spectrum data searched by the sound data matching means by an identification number from the search result information storage unit. Search output to an external search result output device The sound data retrieval apparatus characterized by having a fruit management means. 7. The apparatus according to claim 5, wherein the separator information is a silent portion that has continued for a predetermined time or more.
The described sound data search device. 8. The apparatus according to claim 5, wherein the separator information is sound data having certain characteristics such as a bell sound, a time signal, a chime, a start pistol sound of a competition, and a part of a music piece. The described sound data search device. 9. A first process of pre-registering and storing separator information serving as a dividing point for dividing inputted sound data and at least one designated time value, and A second processing to accumulate; a third processing to divide the input sound data by the separator information; a fourth processing to calculate the length of the sound data divided in the third processing; A fifth process of comparing and determining whether the length of the sound data calculated in the fourth process matches any of the designated time values, and a result of the fifth process, A sixth process of converting the divided sound data into spectrum data of frequency components, collating the spectrum data with pre-registered and stored spectrum data to search for the same sound data, and a result of the fifth process. , If not matched Ends the processing without performing the comparison of the divided sound data,
A recording medium on which a sound data search program for causing a computer to execute a seventh process returning to the third process is recorded.