WO2015115666A1

WO2015115666A1 - Musical composition analysis device and singing evaluation device

Info

Publication number: WO2015115666A1
Application number: PCT/JP2015/053016
Authority: WO
Inventors: 松本　秀一
Original assignee: ヤマハ株式会社
Priority date: 2014-02-03
Filing date: 2015-02-03
Publication date: 2015-08-06
Also published as: JP2015145955A

Abstract

In the present invention a lyrics analysis unit (32) identifies multiple words (W) included in the lyrics of a subject musical composition. An attribute estimation unit (34) estimates the attributes (A) of the subject musical composition from the multiple words (W) identified by the lyrics analysis unit (32). Specifically, the attribute estimation unit (34) uses reference information (R), which specifies a degree of affinity for each word with respect to each of multiple attributes, to estimate the attributes (A) of the subject musical composition from the multiple words (W) identified by the lyrics analysis unit (32).

Description

Music analysis device and singing evaluation device

The present invention relates to a technique for estimating music attributes.

Techniques that apply music attributes (for example, genre) to singing such as karaoke have been proposed. For example, in Patent Document 1, a classification code that specifies a musical genre such as enka or pop is set in advance for each song, and BGV (background video) displayed when the song is sung is set according to the classification code of the song. The configuration to be selected is disclosed. Patent Document 2 discloses a configuration in which the singing scoring standard is changed according to the genre of music.

Japanese Unexamined Patent Publication No. 2003-248489 Japanese Unexamined Patent Publication No. 2011-095437

However, in the techniques of Patent Document 1 and Patent Document 2, it is necessary to set the genre of music for each music in advance.
In view of the above circumstances, an object of the present invention is to estimate music attributes without requiring information for designating attributes for each music.

In order to solve the above-described problems, the music analysis apparatus of the present invention includes a lyrics analysis unit that specifies words included in the lyrics of the music, and an attribute estimation unit that estimates music attributes from the words specified by the lyrics analysis unit. It comprises. With the above configuration, the attributes of the music are estimated according to the words and phrases specified by the lyrics analysis unit from the lyrics of the music, so it is possible to estimate the attributes of the music without requiring information for specifying the attributes for each music It is.

In a preferred aspect of the present invention, the attribute estimation unit estimates the attribute of the music from the phrase specified by the lyrics analysis unit, using reference information that specifies the affinity for each of the plurality of attributes for each phrase. In the above aspect, there exists an advantage that the attribute of a music can be estimated simply by using the reference information which designates the affinity with respect to each of a some attribute for every phrase. “Phrase” implies a word that is the smallest unit of a meaningful language or a phrase (phrase) or sentence composed of a plurality of words. The language of the phrase is arbitrary and is not limited to Japanese.

In a preferred aspect of the present invention, the attribute estimation unit uses the recognition model for each attribute that represents the tendency of the words and phrases used in the lyrics of the music of each attribute, and determines the attributes of the music from the phrases specified by the lyrics analysis unit. presume. In the above aspect, since the attribute of a song is estimated using a recognition model for each attribute that represents the tendency of the phrase used for the lyrics of the song of each attribute, it is actually used for the lyrics of many songs of each attribute. Appropriate estimation of the attribute reflecting the tendency of the phrase to be performed is realized.

In a preferred aspect of the present invention, the attribute estimation unit estimates the attribute of each of the plurality of analysis sections into which the music is divided according to each phrase specified by the lyrics analysis unit from the lyrics in the analysis section. In the above aspect, since the attribute is estimated for each analysis section into which the music is divided, it is possible to estimate the attribute that appropriately reflects the temporal change of the tune and the subject in the music.

In a preferred aspect of the present invention, the attribute estimation unit estimates the attribute of the music according to the music information (for example, performance tempo and tone) of the music. In the above aspect, since the performance tempo of the music is added to the estimation of the attributes in addition to the respective words included in the lyrics of the music, for example, compared with the configuration in which only the words of the lyrics of the music are used for the estimation of the attributes There is an advantage that the attribute of the music can be estimated with high accuracy.

The present invention further provides a singing evaluation apparatus.
A singing evaluation apparatus according to an aspect of the present invention includes a lyrics analysis unit that identifies a phrase included in the lyrics of a song, an attribute estimation unit that estimates a song attribute from the phrase specified by the lyrics analysis unit, and an attribute estimation unit And a singing evaluation unit that evaluates the singing voice by an evaluation method according to the attribute estimated by.
A singing evaluation apparatus according to an aspect of the present invention includes a lyric analysis unit that identifies words included in the lyrics of a song, an attribute estimation unit that estimates the attributes of the song from the phrases specified by the lyric analysis unit, and a singing voice A singing evaluation unit that causes a display device to display a comment according to the evaluation result and the attribute estimated by the attribute estimation unit.
A singing evaluation apparatus according to an aspect of the present invention includes a lyrics analysis unit that identifies words included in the lyrics of a song, an attribute estimation unit that estimates attributes of the song from the phrases specified by the lyrics analysis unit, and attribute estimation A control unit that controls the operation of the lighting device or a display image by the display device according to the attribute estimated by the unit.

It is a block diagram of the song evaluation apparatus which concerns on 1st Embodiment of this invention. It is explanatory drawing of reference information. It is a block diagram of a music analysis part. It is a flowchart of an attribute estimation process. It is a flowchart of a song evaluation process. It is explanatory drawing of weight value information. It is a specific example of weight value information. It is a specific example of a basic value. It is a block diagram of the music analysis part in 2nd Embodiment. It is a flowchart of the attribute estimation process in 2nd Embodiment. It is explanatory drawing of the analysis area in 3rd Embodiment. It is explanatory drawing of operation | movement of 4th Embodiment.

<First Embodiment>
FIG. 1 is a configuration diagram of a singing evaluation apparatus 100 according to the first embodiment of the present invention. The singing evaluation device 100 is an information processing device that evaluates (scores) the skill of singing by a user, and is realized by a computer system including an arithmetic processing device 12, a storage device 14, a sound collection device 16, and a display device 18. The The singing evaluation apparatus 100 is suitably used as a karaoke apparatus that reproduces accompaniment sounds of music, for example.

The sound collection device 16 is a device (microphone) that collects ambient sounds. The sound collection device 16 of the first embodiment collects a singing voice V sung by a user on a specific music (hereinafter referred to as “target music”). The target music is, for example, a music selected by the user from a plurality of music. The display device 18 (for example, a liquid crystal display panel) displays an image instructed from the arithmetic processing device 12. For example, the evaluation result of the singing voice V is displayed on the display device 18. In addition, it is also possible to output the evaluation result of the singing voice V by voice from a sound emitting device (for example, a speaker).

The arithmetic processing device 12 controls each element of the song evaluation device 100 in an integrated manner by executing a program stored in the storage device 14. Specifically, as illustrated in FIG. 1, the arithmetic processing device 12 evaluates the skill of singing the target music according to the music analysis unit 22 that estimates the attribute of the target music and the estimation result by the music analysis unit 22. The singing evaluation unit 24 is realized. The attribute estimated by the music analysis unit 22 is, for example, information that directly or indirectly represents the tone (atmosphere) or the subject of the target music. For example, information such as “broken heart”, “sad”, “inseparable”, “happy”, “fun”, “hope”, “city”, “city”, “old”, “future”, “parting”, and “meeting” are estimated as attributes of the target music. Note that it is also possible to estimate musical genres such as “enka”, “pops”, and “love songs” as attributes. A configuration in which each function of the arithmetic processing device 12 is distributed to a plurality of devices, or a configuration in which a dedicated electronic circuit realizes a part of the function of the arithmetic processing device 12 may be employed.

The storage device 14 stores programs executed by the arithmetic processing device 12 and various data used by the arithmetic processing device 12. A known recording medium such as a semiconductor recording medium and a magnetic recording medium or a combination of a plurality of types of recording media is arbitrarily employed as the storage device 14. The storage device 14 of the first embodiment stores music information D for each of a plurality of music pieces. The music information D of an arbitrary piece of music includes melodic information DA that specifies the main melodic note sequence (a time series of a plurality of notes constituting the singing part of the music) of the music, and the lyrics of the music (at the time of the phrase) And lyric information DB designating a series).

In addition, the storage device 14 of the first embodiment stores reference information R used for estimating the attributes of the target music by the music analysis unit 22. FIG. 2 is a schematic diagram of the reference information R. As illustrated in FIG. 2, the reference information R includes the affinity H (H [1,1], H [1,2] for each word W for each of the K attributes a [1] to a [K]. , ...) is a data table for designating. For any one attribute a [k] (k = 1 to K), the affinity H of each of a plurality of words / phrases W that can be included in the lyrics of a large number of songs is designated. The affinity H of one word / phrase W for the attribute a [k] is an index of the degree to which the word / phrase W is compatible (harmonized) as the lyrics of the music of the attribute a [k]. Specifically, the affinity H increases so that the possibility that the word W appears in the lyrics of the music with the attribute a [k] (that is, the word W becomes more familiar with the musical atmosphere of the music). The affinity H for each word W is statistically set for each of the K attributes a [1] to a [K].

FIG. 3 is a configuration diagram of the music analysis unit 22. As illustrated in FIG. 3, the music analysis unit 22 of the first embodiment includes a lyrics analysis unit 32 and an attribute estimation unit 34. The lyrics analyzing unit 32 analyzes the lyrics designated by the lyrics information DB of the target music. Specifically, the lyrics analyzing unit 32 specifies a plurality of words / phrases W included in the lyrics of the target music. For specifying each word / phrase W by the lyric analysis unit 32, known natural language processing such as morphological analysis is arbitrarily employed. For example, the lyric analysis unit 32 selects a predetermined word / phrase W that can be used for the lyric of the music among a plurality of morphemes (the smallest unit having linguistic meaning) obtained by dividing the lyrics specified by the lyric information DB.

The attribute estimation unit 34 estimates the attribute A of the target song from the plurality of words W specified by the lyrics analysis unit 32 for the lyrics of the target song. The reference information R stored in the storage device 14 is used for attribute estimation by the attribute estimation unit 34.

FIG. 4 is a flowchart of a process in which the attribute estimation unit 34 estimates the attribute A of the target music (hereinafter referred to as “attribute estimation process”). When the attribute estimation process is started, the attribute estimation unit 34 counts the number of appearances C for each of the plurality of words W specified by the lyrics analysis unit 32 from the lyrics of the target music (SA1). That is, the appearance frequency C of each word / phrase W corresponds to the number (number of times) of the word / phrase W included in the lyrics of the target music.

The attribute estimation unit 34 calculates an estimated index X [k] (X [1] to X [K]) for each of the K attributes a [1] to a [K] registered in the reference information R ( SA2). Arbitrary one estimated index X [k] in the first embodiment represents the affinity H of each word / phrase W specified in the attribute a [k] corresponding to the estimated index X [k] in the reference information R. It is a weighted sum (X [k] = Σ {HC}) of the number of appearances C as a weighted value. As can be understood from the above description, among the K attributes a [1] to a [K], the attribute a [k] having a large number of appearances C of the word / phrase W having a high affinity H specified by the reference information R. ] Is set to a larger numerical value as the estimated index X [k]. In other words, the estimated index X [k] of the attribute a [k] increases as the number of appearances C of the word W (word W having a high affinity H) familiar with the musical atmosphere of the attribute a [k] in the lyrics of the target song increases. Tends to be set to a large number. Therefore, each estimation index X [1] can be used for estimating the attribute A of the target music piece.

The attribute estimation unit 34 estimates the attribute A of the target music in accordance with the K estimated indexes X [1] to X [K] calculated by the above procedure (SA3). Specifically, the attribute estimation unit 34 of the first embodiment selects one attribute a [k] having the maximum estimated index X [k] among the K attributes a [1] to a [K]. Confirmed as attribute A of the target song. The above is a specific example of attribute estimation processing.

1 evaluates the skill of the singing by the user by analyzing the singing voice V according to the attribute A of the target music estimated by the music analysis unit 22. Schematically, the singing evaluation unit 24 compares the musical note string specified by the melody information DA of the music information D of the target music with the singing voice V picked up by the sound pickup device 16, and matches the two. An evaluation index (score) S corresponding to the degree of (difference) is calculated.

FIG. 5 is a flowchart of an operation in which the singing evaluation unit 24 evaluates the singing voice V (hereinafter referred to as “singing evaluation processing”). When the singing evaluation process is started, the singing evaluation unit 24 calculates basic values s [1] to s [N] for each of a plurality (N) of different musical evaluation items (SB1). The basic value is calculated regardless of the attribute of the target music when the target music is sung. N evaluation items are musical viewpoints to evaluate the skill of singing, such as accuracy of pitch, propriety of inflection, singing technique (bar, fall, vibrato, kobushi, shakuri, long tone) Includes presence or absence. A known technique is arbitrarily adopted for detection of the evaluation item. For example, Japanese Patent Application Laid-Open No. 2008-268370 discloses a technique for detecting a sung section using the Kobushi technique. Japanese Unexamined Patent Application Publication No. 2004-102146 discloses a technique for detecting vibrato, and Japanese Unexamined Patent Application Publication No. 2012-8596 discloses a technique for detecting a long tone. Furthermore, Japanese Unexamined Patent Application Publication No. 2007-334364 discloses a technique for detecting vibrato, intonation, voice quality, timing, and sneezing, respectively. A known technique is arbitrarily employed for singing evaluation for each evaluation item (calculation of basic values s [1] to s [N]). As a known technique, there is, for example, Japanese Patent No. 5585320. Japanese Patent No. 5585320 discloses singing evaluation based on singing voice and music information (music data).

The singing evaluation unit 24 specifies N weighted values w [1] to w [N] corresponding to different evaluation items (SB2). The weight value information E shown in FIG. 6 is used to calculate each weight value w [n] (n = 1 to N). As illustrated in FIG. 6, the weight information E is a data table that specifies N weight values w [1] to w [N] for each of the K attributes a [1] to a [K]. is there. Each of the N weighted values w [1] to w [N] is individually set for each attribute a [k] according to the importance of each evaluation item in the singing of the music of each attribute a [k]. . For example, for the attribute a [k] of songs such as enka, where singing techniques such as measures and vibrato are regarded as important, the weight value w [n] of the evaluation item of each singing technique is set to a relatively large value. For a song attribute a [k] where high accuracy is important, the weight value w [n] of the evaluation item of pitch accuracy is set to a relatively large numerical value. As described above, the weight value information E reflecting the weight of the evaluation item for each attribute a [k] is prepared in advance and stored in the storage device 14.
A specific example of the load value information E in FIG. 6 is illustrated in FIG. In FIG. 7, musical genres such as enka, pops, singing, etc. are shown as the attribute a [k], and evaluation items such as shaku, kobushi, vibrato, pitch, rhythm, and long tone are shown. . The weight value w [n] of these evaluation items is a numerical value of 1 to 10. In FIG. 7, in enka, numerical values such as sneezing, kobushi and vibrato are set large, and numerical values of rhythm are set small. In pops, numerical values such as sneezing, vibrato, and pitch are set large, and numerical values for kobushi are set small. In the singing song, numerical values such as pitch and long tone are set large, and numerical values of sneezing, kobushi, and inflection are set small.
The singing evaluation unit 24 stores N weight values w [1] to w [N] corresponding to the attribute A estimated by the attribute estimation unit 34 among the K attributes a [1] to a [K]. It is specified from the 14 weight value information E.

The singing evaluation unit 24 sings according to the N basic values s [1] to s [N] calculated in step SB1 and the N weighted values w [1] to w [N] specified in step SB2. The evaluation index S of the voice V is calculated (SB3). Specifically, the singing evaluation unit 24 uses the weighted sum of the basic value s [n] to which the weighted value w [n] for each evaluation item is applied as the final evaluation index S (S = Σ {w [n] s [n]}). Therefore, the evaluation index S is set to a larger numerical value as the evaluation regarding each evaluation item regarded as important for the attribute A of the target musical piece among the N evaluation items is higher (the basic value s [n] is larger).
A specific example of the basic value s [n] calculated in step SB1 is shown in FIG. The basic value s [n] shown in FIG. 8 is a numerical value of 1 to 5. If the estimated attribute of the target music is enka, the final evaluation index S is calculated as follows using the weighted value w [n] in FIG. 7 and the basic value s [n] in FIG. The
S = 8 × 5 + 10 × 3 + 10 × 4 + 5 × 3 + 3 × 5 + 5 × 5 + 9 × 2 +...
As understood from the above description, the singing evaluation unit 24 evaluates the singing voice V by a variable evaluation method (evaluation standard) corresponding to the attribute A estimated for the target music piece. The singing evaluation unit 24 displays the evaluation index S calculated by the above procedure on the display device 18 (SB4).

As described above, in the first embodiment, the attribute A of the target music is estimated according to the plurality of words W specified by the lyrics analysis unit 32 from the lyrics of the target music, so that information specifying the attribute for each music is specified. There is an advantage that it is possible to estimate the attribute A of the target music without the need for. In the first embodiment, in particular, the attribute A of the target musical piece is simply estimated by using the reference information R that specifies the affinity H of each word / phrase W for each of the K attributes a [1] to a [K]. There is an advantage that you can.

Moreover, in 1st Embodiment, since a user's song is evaluated by the variable evaluation method according to the attribute A estimated about the object music, appropriate evaluation (for example, attribute A of attribute A) is evaluated. It is possible to achieve an evaluation that emphasizes particularly important evaluation items in singing music.

Second Embodiment
A second embodiment of the present invention will be described below. In addition, about the element which an effect | action and function are the same as that of 1st Embodiment in each form illustrated below, the reference | standard referred by description of 1st Embodiment is diverted, and each detailed description is abbreviate | omitted suitably.

FIG. 9 is a configuration diagram of the music analysis unit 22 in the second embodiment. The attribute estimation unit 34 of the music analysis unit 22 of the second embodiment uses the K recognition models M [1] to M [K] corresponding to the different attributes a [k] to determine the attribute A of the target music. presume. One recognition model M [k] corresponding to an arbitrary attribute a [k] is a statistical model representing the tendency of the word / phrase W used for the lyrics of a song having the attribute a [k]. For example, a statistical model using a mixed distribution such as GMM (Gaussian Mixture Model) is preferably used as the recognition model M [k].

The K recognition models M [1] to M [K] are generated in advance by machine learning using a lot of learning information L corresponding to different music pieces. The learning information L includes attribute information (label) LA that specifies the attribute a [k] of the music, and lyrics information LB that specifies the lyrics of the music. A plurality of attributes a [k] can be specified for an arbitrary piece of music. A large number of learning information L is classified for each attribute a [k] specified by the attribute information LA, and the tendency of the word / phrase W extracted from the lyrics information LB of each learning information L classified into the attribute a [k] is statistically calculated. Through the analysis, a recognition model M [k] for each attribute a [k] is generated. The K recognition models M [1] to M [K] (actually variables that define each recognition model M [k]) prepared in advance by the machine learning exemplified above are stored in the storage device 14, This is used for estimating the attribute A of the target music piece by the attribute estimation unit 34. A known machine learning technique is arbitrarily employed for generating the recognition model M [k]. As a known technique, for example, Japanese Patent Laid-Open No. 2006-139185 is available. Japanese Unexamined Patent Publication No. 2006-139185, paragraph [0011] discloses a technique related to generation of a recognition model.

Each recognition model M [k] of the second embodiment is used to calculate the estimated index Y [k] of the attribute a [k] according to the plurality of words W specified by the lyrics analysis unit 32 from the lyrics of the target music. It is a statistical model. The estimated index Y [k] calculated by any one recognition model M [k] is an index of the probability (likelihood) that the target music corresponds to the attribute a [k]. In other words, the higher the degree that the plurality of phrases W of the target song match the tendency expressed by the recognition model M [k] (the tendency of the phrase W used in the lyrics of the song with the attribute a [k]), the higher the estimated index. Y [k] is set to a large value. That is, the estimated index (likelihood) Y [k] is also considered as the affinity of the word / phrase W for the attribute a [k].

FIG. 10 is a flowchart of attribute estimation processing in which the attribute estimation unit 34 of the second embodiment estimates the attribute A of the target music piece. When the attribute estimation process is started, the attribute estimation unit 34 applies the plurality of words W specified by the lyrics analysis unit 32 to each of the K recognition models M [1] to M [K], thereby different attributes. K estimated indices Y [1] to Y [K] corresponding to a [k] are calculated (SC1). As described above, the degree to which a plurality of phrases W used in the lyrics of the target music matches the tendency of the recognition model M [k] is high (the target music is attribute a [k from the viewpoint of the tendency of the words used in the lyrics). ] Is higher, the estimated index Y [k] is set to a larger numerical value.

The attribute estimation unit 34 estimates the attribute A of the target music according to the K estimation indexes Y [1] to Y [K] (SC2). Specifically, the attribute estimation unit 34 of the second embodiment selects one attribute a [k] having the maximum estimated index Y [k] among the K attributes a [1] to a [K]. Confirmed as attribute A of the target song. The above is a specific example of the attribute estimation process in the second embodiment. The content of the song evaluation process by the song evaluation unit 24 is the same as that of the first embodiment.

In the second embodiment, the same effect as in the first embodiment is realized. In the second embodiment, the attribute A of the target song is determined by using the recognition model M [k] for each attribute a [k] representing the tendency of the words used in the lyrics of the song of each attribute a [k]. Since it is estimated, it is possible to realize an appropriate estimation of the attribute A reflecting the actual tendency of the word / phrase W used in the lyrics of a large number of music pieces with the attribute a [k].

<Third Embodiment>
FIG. 11 is an explanatory diagram of the operation of the song evaluation apparatus 100 according to the third embodiment. As illustrated in FIG. 11, the attribute estimation unit 34 individually estimates the attribute A for each of a plurality of analysis sections P (in the example of FIG. 11, a total of four analysis sections P1 to P4 in FIG. 11). Specifically, the attribute estimation unit 34 applies each word / phrase W identified by the lyric analysis unit 32 from the lyrics in one arbitrary analysis interval P to the attribute estimation process, so that the attribute A of the analysis interval is determined. A method similar to that of the first embodiment or the second embodiment may be adopted for the estimation of the attribute A.

The analysis section P is, for example, each section (number 1, number 2, etc. of the music) in which the target music is divided on the time axis with an interlude as a boundary. In FIG. 11, the attribute A “exciting” is estimated for the analysis section P1 of the target music, the attribute A “parting” is estimated for the analysis section P2, and the attribute A “toned” is estimated for the analysis section P3. The case where the attribute A of “sad” is estimated for the section P4 is illustrated. As described above, since the attribute A is estimated for each analysis section P in the third embodiment, the evaluation method of the singing voice V can be changed for each analysis section P according to the estimation result of the attribute A.

In the third embodiment, the same effect as in the first embodiment is realized. Further, in the third embodiment, since the attribute A is estimated for each analysis section P, it is possible to estimate the attribute A that appropriately reflects the change in the tune and the subject in the music over time.

<Fourth embodiment>
The attribute estimation unit 34 of the fourth embodiment estimates the attribute A of the target song according to the performance tempo of the target song. Specifically, the attribute estimation unit 34 variably controls the range of the attribute a [k] that can be selected as the attribute A of the target music (the candidate range of the attribute A) according to the performance tempo.

For example, music with negative (passive) attribute A such as “sad” and “farewell” tends to have a relatively slow performance tempo. The attribute estimation unit 34 determines whether the performance tempo of the target music is high or low, and as shown in FIG. 12, when the performance tempo is low, K attributes a [1] including a negative attribute a [k] are included. While the attribute A is selected from ~ a [K], the negative attribute a [k] is excluded from the selection target when the performance tempo is high. That is, the negative attribute a [k] is not selected as the attribute A for the target music piece having a high performance tempo.

In the fourth embodiment, the same effect as in the first embodiment is realized. Further, in the fourth embodiment, since the performance tempo of the target music is added to the estimation of the attribute A of the target music in addition to each word W of the lyrics of the target music, for example, only each word W is used for the estimation of the attribute A. There is an advantage that the attribute A of the target musical piece can be estimated with high accuracy compared to the configuration to be performed.

<Modification>
Each of the above forms can be variously modified. Specific modifications are exemplified below. Two or more aspects arbitrarily selected from the following examples may be appropriately combined.

(1) In each of the above-described embodiments, one attribute a [k] corresponding to the estimated index (X [k], Y [k]) among the K attributes a [1] to a [K] is targeted. Although selected as the attribute A of the music piece, a combination of a plurality of attributes a [k] can be estimated as the attribute A of the target music piece. For example, a configuration in which a combination of a predetermined number of attributes a [k] positioned in descending order of the estimated index X [k] or the estimated index Y [k] is selected as the attribute A of the target music, or the estimated index X [k] Alternatively, a configuration in which one or more attributes a [k] whose estimated index Y [k] exceeds a predetermined threshold is selected as the attribute A of the target music is adopted.

In the configuration in which a plurality of attributes a [k] are selected as illustrated above, the song evaluation process by the song evaluation unit 24 is executed according to the plurality of attributes a [k]. For example, assuming that a total of three combinations of the attribute a [k1], the attribute a [k2], and the attribute a [k3] are estimated as the attribute A of the target music, the singing evaluation unit 24 uses the attribute a [ k1], a weight value w [n] corresponding to the attribute a [k2], a weight value w [n] corresponding to the attribute a [k3], and a weight value w [n] corresponding to the attribute a [k3]. The evaluation index S is calculated as a definite weight value w [n] of the base value s [n].

(2) In each of the above-described embodiments, each weight value w [n] applied to the calculation of the evaluation index S (the weighted sum of the basic values s [n]) is changed according to the attribute A of the target song. The configuration for changing the singing evaluation method according to the attribute A of the target song is not limited to the above examples. For example, the evaluation index S is calculated by adopting an evaluation method corresponding to the attribute A of the target music among different types of evaluation methods (singing evaluation processing), and the basic value s [n] for each evaluation item A configuration in which a variable applied to the calculation is variably set according to the attribute A of the target music, and a combination of the basic value s [n] applied to the calculation of the evaluation index S among the N basic values s [n] ( In other words, a configuration in which the evaluation item added to the singing evaluation) is changed according to the attribute A of the target music piece may be employed.

(3) In each form mentioned above, although the structure which applies the attribute A which the music analysis part 22 estimated to the song evaluation process was illustrated, the usage method of the attribute A estimated about the object music is not limited to the above illustration. Specifically, the following aspects are preferably employed.

[A] The arithmetic processing unit 12 selects an image (BGV) displayed on the display device 18 when the target song is sung according to the attribute A of the target song. For example, when the attribute A “sad” is estimated for the target song, a sad impression BGV is selected and displayed. Further, in the configuration in which the BGV (moving image) displayed on the display device 18 is sequentially selected from a plurality of candidates at each time point in the music, the selection target at each time is controlled according to the attribute A of the target music. Is also possible. A configuration in which an image displayed on a terminal device such as a mobile phone or a smartphone owned by each user is controlled according to the attribute A may be employed.

[B] The arithmetic processing unit 12 (the singing evaluation unit 24) causes the display device 18 to display a comment (an evaluation comment or a guidance comment) according to the attribute A of the target song. For example, a configuration in which an evaluation comment related to the evaluation result of the song of the target music is presented to the user is suitable. Specifically, the singing evaluation unit 24 presents an evaluation comment selected according to the attribute A from a plurality of candidates prepared in advance together with the evaluation index S to the user. For example, when the evaluation index S exceeds a predetermined value and the attribute A “satisfied” is estimated for the target song, an evaluation comment “Nice songs are also good!” Is displayed on the display device 18 together with the evaluation index S. Is displayed.

In addition, a configuration in which guidance comments related to guidance of singing the target music are presented to the user is also adopted. Specifically, the singing evaluation unit 24 presents, together with the evaluation index S, the evaluation comment selected according to the attribute A from the plurality of candidates prepared in advance to the user, in the same manner as the presentation of the evaluation comment described above. For example, the evaluation index S is lower than a predetermined value, and the target music is attribute A such as “not cut” or “sad” (especially emotional to be sung with particular attention to inflection compared to other elements such as pitch) When the music attribute is estimated, a guidance comment such as “Let's sing with emotion while paying attention to inflection” is displayed on the display device 18 together with the evaluation index S.

[C] The arithmetic processing unit 12 (the singing evaluation unit 24) is used for the production of various lighting devices (for example, mirror balls or lasers that realize various visual effects) that are installed in the acoustic space together with the singing evaluation unit 100. It is also possible to control the operation of the device) according to the attribute A. For example, when the attribute A “not cut” is estimated, warm color (red, etc.) illumination light is emitted, and when the attribute A “fresh” is estimated, cold color (blue, etc.) illumination light is emitted. When the attribute A of “violent” is estimated, the brightness or hue of the illumination light is changed.

(4) In the fourth embodiment, the attribute A of the target music is estimated according to the performance tempo, but the music information (musical element) added to the estimation of the attribute A is not limited to the performance tempo. For example, music information such as the performance part configuration (musical instrument combination), rhythm pattern, music tone, music length, music composition, etc. of the target music can be added to the estimation of the attribute A.

(5) In each above-mentioned form, although the song evaluation apparatus 100 which applies the attribute A which the music analysis part 22 estimated to the song evaluation process was illustrated, this invention is also used as a music analysis apparatus which estimates the attribute A of object music. Can be realized. The music analysis apparatus is an information processing apparatus including the music analysis unit 22 (the lyrics analysis unit 32 and the attribute estimation unit 34) exemplified in the above-described embodiments. The singing evaluation apparatus 100 of each form mentioned above is equivalent to the structure which added the singing evaluation part 24 to the music analysis apparatus.

The music analysis apparatus according to each of the above aspects is realized by hardware (electronic circuit) such as DSP (Digital Signal Processor) dedicated to analysis of music attributes, and a general-purpose such as CPU (Central Processing Unit). This is also realized by cooperation between the arithmetic processing unit and the program. A program according to a preferred aspect of the present invention causes a computer to function as a lyric analysis unit that identifies a phrase included in the lyrics of a song, and an attribute estimation unit that estimates a song attribute from the phrase specified by the lyrics analysis unit . The program of the present invention can be provided in a form stored in a computer-readable recording medium and installed in the computer. The recording medium is, for example, a non-transitory recording medium, and an optical recording medium (optical disk) such as a CD-ROM is a good example, but a known arbitrary one such as a semiconductor recording medium or a magnetic recording medium This type of recording medium can be included. For example, the program of the present invention can be provided in the form of distribution via a communication network and installed in a computer.
The present invention is also specified as an operation method (music analysis method) of the music analysis apparatus according to each aspect exemplified above. A music analysis method according to a preferred aspect of the present invention includes a lyric analysis process for specifying a phrase included in the lyric of the music, and an attribute estimation process for estimating the attribute of the music from the phrase specified in the lyric analysis process. .

This application is based on a Japanese patent application filed on February 3, 2014 (Japanese Patent Application No. 2014-018518), the contents of which are incorporated herein by reference.

According to the present invention, it is possible to estimate the attributes of a song without requiring information for specifying the attribute for each song.

DESCRIPTION OF SYMBOLS 100 ... Singing evaluation apparatus, 12 ... Arithmetic processing apparatus, 14 ... Memory | storage device, 16 ... Sound collection apparatus, 18 ... Display apparatus, 22 ... Music analysis part, 24 ... Singing evaluation part, 32 ... Lyric analysis unit, 34 …… Attribute estimation unit.

Claims

A lyric analyzer that identifies words included in the lyrics of the song;
A music analysis apparatus comprising: an attribute estimation unit that estimates the attribute of the music from the phrase specified by the lyrics analysis unit using reference information that specifies the affinity for each of the plurality of attributes for each phrase.
A lyric analyzer that identifies words included in the lyrics of the song;
A music analysis comprising: an attribute estimation unit that estimates an attribute of the music from the phrase specified by the lyrics analysis unit using a recognition model for each attribute that represents a tendency of the word used for the lyrics of the music of each attribute apparatus.
The attribute estimation unit estimates an attribute according to each phrase specified by the lyrics analysis unit from lyrics in the analysis section for each of a plurality of analysis sections into which the music is divided. Music analysis device.
A music analysis apparatus according to claim 1 or 2,
A singing evaluation device comprising: a singing evaluation unit that evaluates a singing voice by an evaluation method according to an attribute estimated by the attribute estimation unit.
A music analysis apparatus according to claim 1 or 2,
A singing evaluation device comprising: a singing evaluation unit that causes a display device to display a comment according to a result of evaluating a singing voice and an attribute estimated by the attribute estimation unit.
A music analysis apparatus according to claim 1 or 2,
A singing evaluation apparatus comprising: a control unit that controls an operation of a lighting device or a display image by a display device according to the attribute estimated by the attribute estimation unit.
Identify words and phrases included in the song lyrics,
A music analysis method for estimating an attribute of a music piece from a specified word / phrase using reference information for designating affinity for each of a plurality of attributes for each word / phrase.
Identify words and phrases included in the song lyrics,
A music analysis method for estimating an attribute of the music from a specified word / phrase using a recognition model for each attribute representing a tendency of the word / phrase used for lyrics of the music of each attribute.