JP4573684B2 - Information search device, information search method, information search program, and recording medium - Google Patents

Description

  The present invention relates to an information search device, an information search method, an information search program, and a recording medium, and more specifically, an information search device, an information search method, and an information search for searching for one or more songs from a plurality of songs. The present invention relates to a computer program and a recording medium.

  In recent years, digital data for reproducing a large number of songs, such as an in-vehicle navigation device and a home server device, is stored, and a favorite song is selected and reproduced.

  In this case, Patent Document 1 is known as a music search method reflecting the user's subjectivity. The same document includes a correlation value between each search keyword (for example, “merry song”, “squat song”) and each feature word (for example, “brightness”, “genki”) and can be updated. In addition to preparing a table, a feature word list indicating whether or not there is a feature related to each feature word is prepared by “1” and “0”, and a user inputs a desired search keyword. Are searched based on the sensitivity table and the feature word list.

Japanese Patent Laid-Open No. 2003-132805 (FIGS. 1 to 3)

  However, since the music search method uses a sensitivity table only for the search keyword input by the user, the correlation with the music to be searched is not taken into consideration, and as a result it is appropriate. There is a problem that some songs are not searched.

  The present invention has been made in view of the above, and provides an information search device, an information search method, an information search program, and a recording medium that can search for a song that matches the subjectivity of the user. One of the purposes.

  In order to solve the above-described problems and achieve the object of the present invention, the present invention is an information search apparatus for searching for one or more songs from a plurality of songs, and stores music feature values for each song. Music feature quantity storage means, match history storage means for storing match music and non-match music for each search word, match music feature quantity, first non-match music feature quantity, and second for each search word Input by the input means, input means for inputting information such as a search word for searching for music in accordance with a user's operation, and the input means Based on the matching music and the non-matching music stored for each search word in the matching history storage means, the matching music feature for each search word is executed. Calculating the amount, the first non-match music feature, and the second non-match music feature Search feature value calculation means for executing a search feature value calculation process recorded in the search feature value storage means, wherein the search means is (A) stored in the music feature value storage means in the search process. The music feature quantities of all songs, the matched song feature quantities corresponding to the search words input by the search word input means, the first non-match song feature quantities, and the second Compare the distances with the non-matching music feature quantities, respectively, (B) search for a song having a music feature quantity that minimizes the distance to the matching music feature quantity among each distance as a candidate song, and (C) the input Based on the information input from the means, it is determined whether or not the candidate song matches the search word. If it matches the search word, it matches, and if it does not match the search word, it does not match Recorded in the match history storage means as a song, and the search In the search feature value calculation process, the collection amount calculation unit acquires, from the music feature value storage unit, (A) a music feature value corresponding to the matched music stored in the match history storage unit for each search word. Then, the statistical value of the acquired music feature value is calculated as the matched music feature value, and (B) the music feature value corresponding to the non-matched music stored in the match history storage unit is acquired from the music feature value storage unit. Then, a set of non-matching songs having a music feature amount smaller than the matching song feature value with the matching song feature value as a boundary is defined as a first non-matching song set, and a non-matching music feature value larger than the matching song feature value. The set of matching songs is classified as a second non-matching song set, and (C) the statistical value of the music feature value of the non-matching music belonging to the first non-matching song set is calculated as the first non-matching song feature value. And non-matching music belonging to the second non-matching music set And (D) the calculated matched music feature quantity, the first non-matching music feature quantity, and the second non-matching music feature quantity are calculated. It is recorded in the search feature amount storage means.

  In order to solve the above-described problems and achieve the object of the present invention, the present invention provides an information search method for searching for one or a plurality of songs from a plurality of songs in accordance with a user operation. An input process for inputting information such as a search word for searching for a song, a search process for searching for a song that matches the search word input in the input process, and a matching song and a non-match for each search word Referring to the match history storage means for storing songs, based on the stored matched songs and non-matched songs, for each search word, the matched song feature quantity, the first non-matched song feature quantity, and the second non-matched song quantity are stored. A search feature value that calculates a match song feature value and records a match song feature value, a first non-match song feature value, and a second non-match song feature value for each search word in the search feature value storage unit. In the search process, (A) stored in the music feature amount storage means The music feature quantities of all songs, the matched song feature quantities corresponding to the search words input by the search word input means, the first non-match song feature quantities, and the second Compare the distances with the non-matching music feature quantities, respectively, (B) search for a song having a music feature quantity that minimizes the distance to the matching music feature quantity among each distance as a candidate song, and (C) the input Based on the information input from the means, it is determined whether or not the candidate song matches the search word. If it matches the search word, it matches, and if it does not match the search word, it does not match As a song, it is recorded in the match history storage means, and in the search feature amount calculation step, for each search word, a music feature amount corresponding to the match song stored in the match history storage means is stored in the music feature. Statistics of music features acquired from quantity storage means And (B) acquiring a music feature amount corresponding to the non-matching song stored in the match history storage unit from the music feature amount storage unit, and setting the match song feature amount as a boundary A set of non-matching songs having a music feature amount smaller than the matching song feature amount is a first non-matching song set, and a set of non-matching songs having a music feature amount greater than the matching song feature amount is a second non-matching Categorized as a song set; (C) calculating a statistical value of a music feature amount of a non-matching song belonging to the first non-matching song set as a first non-matching song feature amount; and The statistical value of the music feature value of the non-matching music belonging to the music set is calculated as the second non-matching music feature value. (D) The calculated matching music feature value, the first non-matching music feature value, and the second The non-matching music feature quantity is recorded in the search feature quantity storage means.

  In order to solve the above-described problems and achieve the object of the present invention, the present invention provides an information search program for causing a computer to search one or more songs from among a plurality of songs. In accordance with the operation, an input process for inputting information such as a search word for searching for a song, a search process for searching for a song that matches the search word input in the input process, and a match for each search word Referring to the match history storage means for storing songs and non-match songs, based on the stored match songs and non-match songs, for each search word, the match song feature value, the first non-match song feature value, and The second non-matching music feature value is calculated, and the search feature value storage unit records the matching music feature value, the first non-matching music feature value, and the second non-matching music feature value for each search word. And a search feature amount calculating step for causing the computer to execute the search processing (A) The music feature values of all the songs stored in the music feature value storage means and the matched song features corresponding to the search word input by the search word input means stored in the search feature value storage means And a distance between the first non-matching music feature quantity and the second non-matching music feature quantity, and (B) a music feature quantity that minimizes the distance from the matching music feature quantity among the distances. (C) based on the information input from the input means, it is determined whether or not the candidate song matches the search word, and if it matches the search word Is recorded in the match history accumulating means as a matched song, and if it does not match the search word, it is recorded in the match history accumulating means. In the search feature value calculating step, (A) is stored in the match history accumulating means for each search word. The music feature value storage unit stores a music feature value corresponding to the matched music piece And a statistical value of the acquired music feature value is calculated as a matched music feature value, and (B) a music feature value corresponding to the non-matched music stored in the match history storage means is calculated as the music feature value. A set of non-matching songs having a music feature value smaller than the matching song feature value, with the matching song feature value as a boundary, obtained from the storage means is a first non-matching song set, a music feature larger than the matching song feature value. A set of non-matching songs having a quantity is classified as a second non-matching song set, and (C) a statistical value of a music feature value of the non-matching songs belonging to the first non-matching song set is calculated as a first non-matching song. A statistical value of a music feature value of a non-matching music belonging to the second non-matching music set is calculated as a second non-matching music feature value, and (D) the calculated matching music feature Search for a quantity, a first non-matching music feature quantity, and a second non-matching music feature quantity It is recorded in the collected amount storage means.

  The best mode of the present invention will be described in detail below. It should be noted that the present invention is not limited by this embodiment, and all combinations of features described in the embodiment are not necessarily required for the solution means of the invention. In addition, constituent elements in the following embodiments include those that can be easily assumed by those skilled in the art or that are substantially the same.

  FIG. 1 is a diagram showing a configuration example of an information search apparatus according to the present invention. The information search apparatus shown in FIG. 1 searches for music that matches a user's preference using music feature values. As shown in FIG. 1, the information search apparatus according to the present invention includes a song input unit 1, a song database 2, a music feature value generation unit 3, a music feature value database 4, a match history database 5, and a search feature. An amount calculation unit 6, a search processing unit 7, an operation unit 8, and a song output unit 9 are provided.

  The song input unit 1 records song data (audio file) in the song database 2 using the song ID as a key. The song database 2 stores song data (audio files) in association with song IDs. When new song data is recorded in the song database 2, the music feature value generating unit 3 generates the song basic information and the music feature value, and uses the song ID as a key to obtain the song basic information and the music feature value. Recorded in the music feature database 4. Here, the song basic information includes a title name, an artist name, and the like. In addition, the music feature amount is included in, for example, BPM (Song Speed: Beat Per Minutes), recording level, maximum volume (maximum output level of the song), average volume (average output level of the song), and the song. Parameters such as the chord to be played, the beat level of the song (ie, the signal level (magnitude) of the beat component of the song), the key of the song (key such as C major or B minor), and the like.

  The music feature database 4 stores the song ID, the song basic information, and the music feature in association with each other. FIG. 2 is a diagram illustrating a configuration example of the music feature amount database 4. In the example shown in the figure, data of song basic information (song title,...) And music feature values (BPM, recording level,...) Are stored in association with the song ID.

  The match history database 5 stores a match song ID, a non-match song ID, a match song feature value, a first non-match song feature value, and a second non-match song feature value for each search word. Here, as the search terms, for example, emotion words such as “warm up”, “encouraged”, “sad”, “smiling”, “bright”, “fun”, and the like can be used. FIG. 3 is a diagram illustrating a configuration example of the match history database 5. As shown in FIG. 3, the match history database 5 includes a match history list 5a including a match song list and a non-match song list, and a search feature value list 5b for each search.

  In the match history list 5a, the user's determination result (match music ID, non-match music ID) whether or not the candidate music searched in the search processing (see FIG. 6) of the search processing unit 7 matches the search word. ) Is recorded as history. In the search feature quantity list 5b, the search feature quantity (matched music feature quantity which is a statistical value of the music feature quantity of the matched music, calculated by the search feature quantity calculation processing (see FIG. 7) of the search feature quantity calculation unit 6 is included. A first non-matching music feature value that is a statistical value of the music feature value of the first non-matching music set, and a second non-matching music feature value that is a statistical value of the music feature value of the second non-matching music set ) Is recorded. Here, for example, an average value or a variance value can be used as the statistical value, but in the following description, a case where the average value is used as the statistical value will be described.

  The search feature value calculation unit 6 executes a search feature value calculation process, which will be described later, and obtains a music feature value from the music feature value database 4 using the matched song ID recorded in the match history list 5a of the match history database 5 as a key. The average value is read out as the matched music feature value and recorded in the search feature value list 5 b of the matched history database 5. For example, in the example shown in FIG. 3, in the case of the search word A, the music feature quantities are read from the music feature quantity database 4 using the matching song IDs “0001” and “0004” as keys, and two music feature quantities are obtained. The average value of is calculated. When the music feature amount is a plurality of parameters, an average value is calculated for each parameter.

  Further, the search feature quantity calculation unit 6 reads out the music feature quantity from the music feature quantity database 4 using the non-matching song ID of the match history database 5 as a key, and uses the matching song feature quantity calculated above to Classify as a non-matching song set. Specifically, a set of songs having a music feature amount smaller than the match song feature amount with the match song feature amount as a boundary is defined as a first non-match song set, a song set having a music feature amount larger than the match song feature amount. Let the set be the second non-matching song set. Then, the average value of the music feature values of the songs belonging to the first and second non-matching song sets is calculated, and the search feature value list of the match history database 5 is calculated as the first and second non-matching song feature values. Record in 5b.

  The search processing unit 7 executes a search process described later to search for candidate songs that match the search word input from the operation unit 8. Specifically, the music feature values of all songs stored in the music feature database 4, the matched song feature values corresponding to the search words stored in the match history database 5, the first non-match song feature values, Then, the distances to the second non-matching music feature quantity are compared, and the music having the music feature quantity that minimizes the distance from the matching music feature quantity among the distances is searched for as a candidate song.

  FIG. 4 is a diagram illustrating an example of music feature amounts and music distribution. In the figure, the horizontal axis indicates music feature (BPM), and the vertical axis indicates the existence ratio of songs. In FIG. 4, when a music feature quantity inappropriate for a search word is represented by one non-matching music feature quantity, for example, music features of non-matching music exist in both poles so that the matching music feature quantity is sandwiched between them. Is averaged with large and small values and does not become an appropriate feature amount. In the present embodiment, the matched music feature value is sandwiched between the first non-matched music set and the second non-matched music set by using the matched music feature value as a boundary, thereby sandwiching the matched music feature value. Thus, even if there are music feature values of non-matching songs in both poles, the feature value is not averaged with the magnitude value, and the impression of the user's search word like music has a certain degree of music feature value Even when the user concentrates on the music and deviates from the preference as the user moves away from the value, the music feature amount of the non-matching music set does not become ambiguous, so that the matching candidate music can be searched by the search word.

  In addition, the search processing unit 7 puts the user's determination result (matched song ID, non-matched song ID) on whether or not the searched candidate song matches the search word in the match history list 5a corresponding to the search word. Record. The candidate song memory 7a is used as a work memory when the search processing unit 7 performs a search process.

  The operation unit 8 includes, for example, a touch panel display unit. The user inputs a search term, inputs a search start instruction, inputs a match / non-match with the search word of the searched candidate song, and plays a song. This is a user interface for giving instructions and the like. FIG. 5A is a diagram illustrating an example of the search start screen of the operation unit 8, and FIG. 5B is a diagram illustrating an example of the search result screen of the operation unit 8.

  The search start screen shown in FIG. 5A includes a search word selection field 11 for selecting a search word and a search button 12 for instructing start of music search. Operate the cursor 11a in the search word selection field 11 to select a desired search word (for example, “warm up”, “encouraged”, “sad”, “sad”, “bright”, “fun”, etc.) When the search button 12 is pressed, a search for music that matches the search word selected by the search processing unit 7 is executed. In the example shown in the figure, “bright” is selected as the search term.

  The search result screen of the search processing unit 7 is displayed on the search result screen shown in FIG. The search result screen includes a search result display column (order, title name, artist name) 21, a match button 22, a non-match button 23, a play button 24, a rewind button 25, a fast-forward button 26, a play song display column (title name, Artist name) 27 is provided.

  In the search result display field 21, search results of candidate songs that match the search word in the search processing unit 7 are displayed in descending order of match. Candidate songs displayed in the search result display field 21 can be selected with the cursor 21a. The selected candidate song can be played back by the song output unit 9 by pressing the play button 24. The title and artist name of the song being played are displayed in the playback song display column (title name, artist name) 27. Further, whether or not the selected candidate song matches the impression of the search word can be input using the match button 22 / non-match button 23. When the selected candidate song matches the impression of the search word, the match button 22 is pressed, and when it does not match the impression of the search word, the non-match button 23 is pressed.

  The song output unit 9 reads out song data (audio file) from the song database 2 using the song ID as a key in accordance with an instruction from the search processing unit 7, and performs playback or external output.

  FIG. 6 is a flowchart for explaining the search processing of the search processing unit 7. In FIG. 6, when a search word is selected on the search start screen (see FIG. 5A) of the operation unit 8 and the search button 12 is pressed, the search processing unit 7 starts the search. First, the search processing unit 7 searches the search feature quantity (match music feature quantity T0, first non-match music feature quantity T1, second data) from the search feature quantity list 5b corresponding to the selected search word in the match history database 5. A non-matching music feature amount T2) is acquired (step S1).

  Next, the search processing unit 7 executes candidate song search processing (steps S2 to S8), checks all songs in ascending order of the song IDs in the music feature amount database 4, and searches for candidate songs. Specifically, the search processing unit 7 first acquires the song ID and the music feature amount T from the music feature amount database 4 (step S2), and the music feature amount T and the search feature amount (matched song feature amount). D0 = | T−T0 |, D1 = | T−T1 |, D2 = the distances D0, D1, and D2 with respect to T0, the first non-matching music feature amount T1, and the second non-matching music feature amount T2). | T−T2 | is calculated respectively (step S3).

  The search processing unit 7 compares the calculated distances D0, D1, and D2 to determine whether or not the distance D0 is the minimum (step S4). If the distance D0 is the minimum (“Yes” in step S4) ”), The song ID as a candidate song and the calculated distance D0 are stored in the candidate song memory 7a (step S5). When the distance D0 is not the minimum (“No” in step S4), the process proceeds to step S6. In step S6, the search processing unit 7 determines whether or not all songs have been checked. If all songs have not been checked, the search processing unit 7 returns to step S2 and executes the same processing for the next song ID. Then, it is repeatedly executed until the check of all the songs is completed (step 2 to step S6). In this way, the candidate song memory 7a stores a list of song IDs and distances D0 of candidate songs that match the search word.

  When a search is performed by combining a plurality of music feature quantities (BPM and recording level, etc.), the processing from step S2 to step S6 is performed for each music feature quantity. From the candidate songs detected for each of the music feature quantities, a song that has become a candidate song in common (taking the and condition) can be made the final candidate song. In addition, the distances D0, D1, and D2 are calculated for each music feature amount, and the distances D0, D1, and D2 for each music feature amount are added. May be a candidate song. In the case of the addition, each distance may be weighted by the effectiveness of the music feature amount. As described above, by performing a search using a plurality of music feature amounts, a more accurate search can be performed.

  Subsequently, the search processing unit 7 sorts the candidate songs stored in the search candidate memory 7a in ascending order by the distance D0 and displays the results as search results on the search result display screen (see FIG. 5-2) of the operation unit 8. (Step S7).

  Next, the search processing unit 7 performs a search word match / non-match determination process (steps S8 to S12), determines whether or not all candidate songs match the search word, and determines the determination result. The match history database 5 stores the match history list (match song list, non-match song list) 5a.

  Specifically, the user operates the search result screen of the operation unit 8 to play back all candidate songs, determine whether or not the search words match during playback of each candidate song, and search words If it matches, the match button 22 is pressed, and if it does not match the search word, the non-match button 23 is pressed. When the match button 22 is pressed, the search processing unit 7 determines that the song matches the search word, and matches the song ID (match song ID) in the match history list 5 a of the match history database 5. When it is stored in the song list and the non-match button 23 is pressed, it is determined that the song does not match the search word, and the song ID (non-match song ID) is stored in the non-match song list.

  First, when the user selects a candidate song displayed in the search result display field 21 of the search result screen of the operation unit 8 by operating the cursor 21 and presses the play button 24 for the selected candidate song, The search processing unit 7 causes the song output unit 9 to play the selected candidate song (step S38). The user determines whether or not the candidate song being played matches the search word, and if it matches, presses the match button 22; otherwise, presses the non-match button 23 To do.

  The search processing unit 7 determines whether or not there is an input from the match button 22 or the non-match button 23 during reproduction of the candidate song (step S9). If the match button 22 is pressed (“Yes” in step S9), it is determined that the song matches the search word, and the song ID of the song being played is stored in the match history list 5a of the match history database 5. While adding to the match song list and proceeding to step S12 (step S10), if the non-match button 22 is pressed (“No” in step S9), it is determined that the song does not match the search word. Then, the song ID of the song being reproduced is added to the non-matching song list in the matching history list 5a of the matching history database 5 and the process proceeds to step 12 (step S11). The search processing unit 7 determines whether or not the reproduction of all candidate songs has been completed (step S12), and repeatedly executes the same processing until the reproduction of all candidate songs is completed (steps S8 to S12). In this way, each time the search process is performed, the matched music list and the non-matched music list in the matched history list 5a of the matched history database 5 are updated.

  In this example, all candidate songs are played, the user determines match / non-match, and presses the match button 22 / non-match button 23. However, the user does not play the candidate song. Thus, the match button 22 / non-match button 23 may be pressed for each candidate song to determine match / non-match. In addition, among the candidate songs, a song played by the user may be matched, and a song that is not played may be determined as a non-match. This is because the user can determine whether or not it is a search term only with the title if it is a known song, so usually only the songs that match the search term are played, and all candidate songs are played. This is because even a song that you do not want to listen to is played, which may make the user feel annoying.

  FIG. 7 is a flowchart for explaining the search feature value calculation processing of the search feature value calculation unit 6. Such search feature value processing includes (1) when the search processing unit 7 starts a search, (2) when the search processing unit 7 finishes the search, and (3) a matching song in the match history list 5a of the match history database 5. It is executed in any case when a song is added to the list / non-matching song list. Thereby, when the match history list 5a is updated by the search processing unit 7, the search feature list 5b can be updated.

  In FIG. 7, the search feature quantity calculation unit 6 reads the music feature quantity from the music feature quantity database 4 using the song ID (match song ID) included in the match song list of the match history list 5 a of the match history database 5 as a key. Then, the average value is calculated as the matched music feature value and recorded in the search feature value list 5b of the matched history database 5 (step S21).

  Next, the search feature amount calculation unit 6 uses the music ID (non-match song ID) included in the non-match song list of the match history list 5a of the match history database 5 as a key to obtain the music feature value from the music feature value database 4. The music pieces that are read out and included in the non-matching music list are classified into two non-matching music sets using the matching music feature values calculated in step S21 (step S22). Specifically, a set of songs having a music feature amount smaller than the matched song feature amount with a matched song feature amount as a boundary is a first non-matched song set, and a set of songs having a large music feature amount is a second non-matched song feature amount. A set of matching songs. Then, the search feature quantity calculation unit 6 calculates the average value of the music feature quantities of the songs included in each of the first non-matching music set and the second non-matching music set as the first non-matching music feature quantity, Is calculated as a non-matching music feature amount (step S23) and recorded in the search feature amount list 5b of the matching history database 5. As described above, when the search processing unit 7 performs the search process and the match history list 5a is updated, the search feature amount calculation unit 6 updates the search feature amount list 5b and performs a search closer to the user's subjectivity. Feature quantities (matching song feature quantity, first non-matching song feature quantity, second non-matching song feature quantity) are generated, and it becomes possible to search for a matching song according to the user's subjectivity.

  As described above, according to the present embodiment, when the search word and the search instruction are input from the operation unit 8, the search processing unit 7 calculates the music feature values of all songs stored in the music feature value database 4. , Comparing the distance between the matched song feature quantity corresponding to the search word stored in the match history database 5, the first non-match song feature quantity, and the second non-match song feature quantity, Of these, a song having a music feature amount that minimizes the distance to the matching song feature amount is searched as a candidate song, and the user's determination result as to whether or not the searched candidate song matches the search word (match song) ID, non-matching song ID) is recorded in the matching history list 5a corresponding to the search word, while the search feature quantity calculation unit 6 uses the matching song ID recorded in the matching history list 5a of the matching history database 5 as a key. , Music feature database The music feature value is read out from the memory, the average value is calculated as the matching song feature value, recorded in the search feature value list 5b of the matching history database 5, and the non-matching song ID of the matching history database 5 is used as a key. The music feature quantity is read from the music feature quantity database 4, and a set of songs having a music feature quantity smaller than the matched song feature quantity is larger than the first non-matched song set and the matched song feature quantity with the matched song feature quantity as a boundary. Since a set of songs having music feature values is recorded as a second non-matching song set as first and second non-matching song feature values in the search feature value list 5b of the match history database 5, music Even if the user's impression of the search term concentrates on a certain amount of features and deviates from the preference as it departs from that value, the features of the non-matching song set are ambiguous Since there al, it is possible to search for candidate songs that match the search term, it is possible to search for songs that meet the subjectivity of the user.

  Although one embodiment of the present invention has been described in detail with reference to the drawings, a specific configuration example is not limited to this one embodiment, and a design change or the like within a scope not departing from the gist of the present invention. Is included in the present invention. For example, a program for realizing the function of the information retrieval apparatus of FIG. 1 is recorded on the computer-readable recording medium 50 shown in FIG. 8, and the program recorded on the recording medium 50 is shown in FIG. Each function may be realized by being read by the computer 40 and executed.

  The computer 40 shown in the figure includes a CPU (Central Processing Unit) 41 that executes the above-described program, an input device 42 such as a keyboard and a mouse, a ROM (Read Only Memory) 43 that stores various data, an operation parameter, and the like. RAM (Random Access Memory) 44, a reading device 45 for reading a program from a recording medium 50, and an output device 46 such as a display or a printer.

  The CPU 41 implements the above-described functions by reading a program recorded on the recording medium 50 via the reading device 45 and then executing the program. Examples of the recording medium 50 include an optical disk, a flexible disk, and a hard disk.

  The information retrieval apparatus may be configured with a home or business music server and client. FIG. 9 is a diagram showing a configuration example in the case where the information search apparatus is configured by a music server and a client. In FIG. 9, parts having the same functions as those in FIG. The music server 100 and the client 200 are connected to a network via communication means 101 and 102, respectively.

  The information search device, information search method, information search program, and recording medium according to the present invention can be used for a personal computer having a music server function, a car navigation system, a portable music player, a server client type network service such as music distribution, and the like. is there.

It is a figure which shows the structural example of the information search device which concerns on this invention. It is a figure which shows the structural example of a music feature-value database. It is a figure which shows the structural example of a coincidence history database. It is a figure which shows an example of distribution of music feature-value and music. It is a figure which shows an example of the search start screen of an operation part. It is a figure which shows an example of the search result screen of an operation part. It is a flowchart for demonstrating the music search process of a search process part. It is a flowchart for demonstrating the search feature-value calculation process of a search feature-value calculation part. It is a figure which shows the structural example of the information search device which concerns on the other Example of this invention. It is a figure which shows the structural example of the information search device which concerns on the other Example of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Song input part 2 Music database 3 Music feature-value production | generation part 4 Music feature-value database 5 Match history database 6 Search feature-value calculation part 7 Search processing part 8 Operation part 9 Song output part 10 Search word selection column 11 Search button 21 Search result Display field 22 Match button 23 Non-match button 24 Play button 25 Rewind button 26 Fast forward button 27 Play song display field 40 Computer 41 CPU
42 Input device 43 ROM
44 RAM
45 Reading device 46 Output device 50 Recording medium 100 Music server 101 Communication means 200 Client 201 Communication means

Claims (7)

In an information retrieval apparatus for retrieving one or more songs from a plurality of songs,
Music feature storage means for storing the music feature T for each song;
Match history storage means for storing matching songs and non-matching songs for each search term;
Search feature value storage means for storing a matching music feature value T0 , a first non-matching music feature value T1 , and a second non-matching music feature value T2 for each search word;
Depending on the user's operation, an input means for inputting information,
Search means for searching for a song that matches a search word for searching for a song input from the input means;
Based on the match music and non-match music stored for each search word in the match history storage means, for each search word, the match music feature value T0 , the first non-match music feature value T1 , and the second non-match music A search feature quantity calculating means for calculating a song feature quantity T2 and recording it in the search feature quantity storage means;
With
The search means includes
(A) For each music feature amount T of a song stored in the music feature amount storage unit, a matching song feature amount corresponding to a search word input by the input unit stored in the search feature amount storage unit D0 = | T−T0 |, D1 = | T−T1 |, D2 = the distances D0, D1, and D2 with T0, the first non-matching music feature quantity T1, and the second non-matching music feature quantity T2. | T−T2 |
(B) For each music feature amount T, it is determined whether or not the distance D0 is the minimum among the calculated distances D0, D1, and D2, and if the distance D0 is the minimum, the music feature amount T corresponds. Search for songs as candidate songs,
(C) Based on the information input from the input means whether the candidate song matches the search word, it is determined whether the candidate song matches the search word, and the search word Is recorded in the match history storage means as a matched song if it matches, and as a non-matched song if it does not match the search term,
The search feature amount calculating means includes:
(A) For each search word, a music feature quantity T corresponding to the matched music stored in the match history storage means is acquired from the music feature quantity storage means, and a statistical value of the acquired music feature quantity T is obtained Calculate as the matching song feature T0 ,
(B) The music feature quantity T corresponding to the non-matched music stored in the match history storage means is acquired from the music feature quantity storage means, and the matched music feature quantity is defined using the matched music feature quantity T0 as a boundary. A set of non-matching music having a music feature value smaller than T0 is classified as a first non-matching music set, and a set of non-matching music having a music feature value larger than the matching music feature value is classified as a second non-matching music set. ,
(C) The statistical value of the music feature quantity of the non-matching music belonging to the first non-matching music set is calculated as the first non-matching music feature quantity T1 , and the non-matching music belonging to the second non-matching music set is calculated. Calculating the statistical value of the music feature quantity of the matching song as the second non-matching song feature quantity T2 ,
(D) An information retrieval apparatus that records the calculated matched music feature value T0 , the first non-matched music feature value T1 , and the second non-matched music feature value T2 in the search feature value storage means. . 2. The information search apparatus according to claim 1, wherein the music feature amount T is BPM, recording level, maximum volume, average volume, chord, beat level, key, and the like. When searching for candidate songs using a plurality of music feature quantities, the search means searches for candidate songs for each music feature, and the songs that have become candidate songs in common are the final candidate songs. The information search device according to claim 1 or 2 , wherein: The search feature amount calculation means is configured to perform the above (A) when the search means starts the search for the candidate song, after the search ends, or when a match song and a non-match song are added to the match history storage means. The information search device according to any one of claims 1 to 3 , wherein the processing (D) is executed. In an information search method for searching for one or more songs from a plurality of songs,
An input process for inputting information according to a user operation,
A search step for searching for a song that matches a search word for searching for a song input in the input step;
The search feature quantity calculation means refers to the match history storage means for storing the matched music and the non-matched music for each search word, and based on the stored matched music and the non-matched music, the matched music feature for each search word the amount T0, the first non-matching song characteristic quantity T1, and calculates the second nonmatching song feature amount T2, the search feature accumulating means, for each search term, consistent song feature quantities T0, the first non A search feature value calculating step for recording the matched music feature value T1 and the second non-matched music feature value T2 ,
Including
In the search step,
(A) In the input step, the music feature amount storage means for storing the music feature amount T for each song is stored in the search feature amount storage means for each music feature amount T of the song. The distances D0, D1, and D2 between the matched music feature value T0, the first non-matched music feature value T1, and the second non-matched music feature value T2 corresponding to the input search word are expressed as D0 = | T−T0. |, D1 = | T−T1 |, D2 = | T−T2 |
(B) For each music feature amount T, it is determined whether or not the distance D0 is the minimum among the calculated distances D0, D1, and D2, and if the distance D0 is the minimum, the music feature amount T corresponds. Search for songs as candidate songs,
(C) Based on information on whether or not the candidate song matches the search word, which is input in the input step, it is determined whether or not the candidate song matches the search word, and the search word Is recorded in the match history storage means as a matched song if it matches, and as a non-matched song if it does not match the search term,
In the search feature amount calculation step,
(A) For each search word, a music feature quantity T corresponding to the matched music stored in the match history storage means is acquired from the music feature quantity storage means, and a statistical value of the acquired music feature quantity T is obtained Calculate as the matching song feature T0 ,
(B) The music feature quantity T corresponding to the non-matched music stored in the match history storage means is acquired from the music feature quantity storage means, and the matched music feature quantity is defined using the matched music feature quantity T0 as a boundary. A set of non-matching music having a music feature value smaller than T0 is classified as a first non-matching music set, and a set of non-matching music having a music feature value larger than the matching music feature value is classified as a second non-matching music set. ,
(C) The statistical value of the music feature quantity of the non-matching music belonging to the first non-matching music set is calculated as the first non-matching music feature quantity T1 , and the non-matching music belonging to the second non-matching music set is calculated. Calculating the statistical value of the music feature quantity of the matching song as the second non-matching song feature quantity T2 ,
(D) An information retrieval method characterized by recording the calculated matched music feature value T0 , the first non-matched music feature value T1 , and the second non-matched music feature value T2 in the search feature value storage means. . In an information search program for causing a computer to search for one or more songs from a plurality of songs,
An input process for inputting information according to a user's operation,
Referring to the search step for searching for a song that matches the search word for searching for a song input in the input step, and the matching history storage means for storing a matching song and a non-matching song for each search word, Based on the matched music and the non-matched music, for each search word, the matched music feature value T0 , the first non-matched music feature value T1 , and the second non-matched music feature value T2 are calculated, and the search feature value is calculated. A search feature value calculating step of recording, for each search word, a matching song feature value T0 , a first non-matching song feature value T1 , and a second non-matching song feature value T2 in the storage means;
To the computer,
In the search step,
(A) In the input step, the music feature amount storage means for storing the music feature amount T for each song is stored in the search feature amount storage means for each music feature amount T of the song. The distances D0, D1, and D2 between the matched music feature value T0, the first non-matched music feature value T1, and the second non-matched music feature value T2 corresponding to the input search word are expressed as D0 = | T−T0. |, D1 = | T−T1 |, D2 = | T−T2 |
(B) For each music feature amount T, it is determined whether or not the distance D0 is the minimum among the calculated distances D0, D1, and D2, and if the distance D0 is the minimum, the music feature amount T corresponds. Search for songs as candidate songs,
(C) Based on information on whether or not the candidate song matches the search word, which is input in the input step, it is determined whether or not the candidate song matches the search word, and the search word Is recorded in the match history storage means as a matched song if it matches, and as a non-matched song if it does not match the search term,
In the search feature amount calculation step,
(A) For each search word, a music feature quantity T corresponding to the matched music stored in the match history storage means is acquired from the music feature quantity storage means, and a statistical value of the acquired music feature quantity T is obtained Calculate as the matching song feature T0 ,
(B) The music feature quantity T corresponding to the non-matched music stored in the match history storage means is acquired from the music feature quantity storage means, and the matched music feature quantity is defined using the matched music feature quantity T0 as a boundary. A set of non-matching music having a music feature value smaller than T0 is classified as a first non-matching music set, and a set of non-matching music having a music feature value larger than the matching music feature value is classified as a second non-matching music set. ,
(C) The statistical value of the music feature quantity of the non-matching music belonging to the first non-matching music set is calculated as the first non-matching music feature quantity T1 , and the non-matching music belonging to the second non-matching music set is calculated. Calculating the statistical value of the music feature quantity of the matching song as the second non-matching song feature quantity T2 ,
(D) For information retrieval, wherein the calculated matched music feature value T0 , the first non-matched music feature value T1 , and the second non-matched music feature value T2 are recorded in the search feature value storage means. program. A computer-readable recording medium storing the information search program according to claim 6 .

Images