JP2016177144A - Evaluation reference generation device and signing evaluation device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To give redundancy of evaluation to singing technique, while stabilizing reference of singing evaluation.SOLUTION: An evaluation reference generation device according to one embodiment of this invention includes: a reference acquisition unit configured to acquire reference data; a reference feature amount calculation unit configured to calculate feature amount data indicating time change of a feature amount of sound on the basis of the reference data; and a generation unit configured to generate evaluation reference data for defining a reference feature amount range that is set so as to contain a feature amount indicated by the feature amount data as reference, contains an allowable range from the reference, and is changed based on information in which the allowable range is contained in the reference data.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a technique for evaluating a song.

  Karaoke devices often have a function of analyzing and evaluating singing voices. Various methods are used for singing evaluation. As one of the methods, for example, Patent Document 1 discloses a technique for comparing a pitch of a singing voice with a pitch of a model voice to be sung and evaluating the singing according to a pitch difference. In addition, when it is detected that the vibrato is included in the singing voice, there is also a technology that loosens the evaluation criteria so that a high evaluation can be obtained even if the pitch difference is large in the section where the vibrato is included. It is disclosed in Document 1.

JP 2011-209654 A

  If the vibrato is not included in the singing voice where the vibrato is included in the model voice, it is a problem of the presence or absence of the singing technique, that is, the expression problem. is not. However, in the technique disclosed in Patent Document 1, the singing voice without vibrato does not follow the vibrato of the model voice, and the evaluation is lowered. In addition, the evaluation standard is determined depending on the singing voice. Therefore, for example, even in a section where vibrato is unnecessary, there is a problem that the evaluation standard becomes unstable because the evaluation standard of the singing is changed by vibrato included in the singing voice.

  One of the objects of the present invention is to provide evaluation redundancy for the singing technique while stabilizing the standard of singing evaluation.

  According to an embodiment of the present invention, a reference acquisition unit that acquires reference data, a reference feature amount calculation unit that calculates feature amount data indicating a temporal change in a sound feature amount from the reference data, and the feature amount data A reference feature amount range set to include the indicated feature amount as a reference, the reference feature amount range including the allowable range from the reference, and the allowable range being changed based on information included in the reference data And a generation unit that generates evaluation reference data for defining the evaluation reference data.

  According to one embodiment of the present invention, a singing sound acquisition unit that acquires a singing input sound, a singing feature amount calculation unit that calculates singing feature amount data indicating temporal changes in the characteristic amount from the singing input sound, and the feature amount A reference feature amount range that is set to include a reference feature amount that is a reference for evaluation of the reference feature amount, includes an allowable range from the reference feature amount, and the allowable range is changed by a section. An evaluation criterion acquisition unit that acquires evaluation criterion data to be defined; a comparison unit that compares the reference feature amount range indicated by the evaluation criterion data; and a temporal change of the feature amount indicated by the singing feature amount data; and the comparison And a valuation unit for calculating an evaluation value for the singing input sound based on the result. In addition, the allowable range of the reference feature amount range defined by the evaluation reference data acquired by the evaluation reference acquisition unit is set by the first rule except for the predetermined specific section, and the second rule in the specific section It may be set.

  According to one embodiment of the present invention, a singing sound acquisition unit that acquires a singing input sound, a singing feature amount calculation unit that calculates singing feature amount data indicating temporal changes in the characteristic amount from the singing input sound, and the feature amount An evaluation criterion acquisition unit that acquires evaluation criterion data that defines a reference feature amount range that is set by an upper limit value and a lower limit value as a criterion for evaluation, and the range from the upper limit value to the lower limit value is changed by a section; A comparison unit that compares the reference feature amount range indicated by the evaluation reference data with a temporal change of the feature amount indicated by the singing feature amount data, and an evaluation value for the singing input sound based on the comparison result A singing evaluation device comprising: an evaluation unit that calculates Further, the reference feature amount range defined by the evaluation reference data acquired by the evaluation reference acquisition unit is the first section of the allowable range in which the width between the upper limit value and the lower limit value is the first width, and the first width. It may include a second section of an allowable range having a different second width. Further, the reference feature amount range includes a low evaluation range having a large allowable range and a high evaluation range having a small allowable range, and the evaluation unit is adapted to a temporal change of the feature amount indicated by the singing feature amount data. The evaluation value for the singing input sound may be calculated based on the result compared with the low evaluation range and the result compared with the high evaluation range.

  According to an embodiment of the present invention, it is possible to provide evaluation redundancy for a singing technique while stabilizing a standard for singing evaluation.

It is a block diagram which shows the structure of the evaluation apparatus 1 in 1st Embodiment of this invention. It is a block diagram which shows the structure of the evaluation reference | standard production | generation function and evaluation function in 1st Embodiment of this invention. It is a figure explaining the 1st rule applied to determination of the standard pitch range in the usual section in a 1st embodiment of the present invention. It is a figure explaining the 2nd rule applied to the determination of the reference pitch range in the vibrato detection area in 1st Embodiment of this invention. It is a figure explaining the 2nd rule applied to determination of the reference pitch range in the Kobushi detection section in a 1st embodiment of the present invention. It is a figure explaining the 2nd rule applied to the determination of the reference pitch range in the fall detection area in 1st Embodiment of this invention. It is a figure explaining the 2nd rule applied to the determination of the reference | standard pitch range in the shackle detection area in 1st Embodiment of this invention. It is a figure explaining the 2nd rule applied to the determination of the reference pitch range in the long tone detection area in 1st Embodiment of this invention. It is a figure explaining the song evaluation method using the song pitch in 1st Embodiment of this invention. It is a figure explaining the reference | standard volume speed range in 2nd Embodiment of this invention. It is a figure explaining the determination method of the reference | standard pitch range in the normal area and vibrato detection area in 3rd Embodiment of this invention. It is a block diagram which shows the structure of the evaluation function in 4th Embodiment of this invention. It is a block diagram which shows the structure of the evaluation reference | standard production | generation function in 6th Embodiment of this invention.

  Hereinafter, an evaluation apparatus according to an embodiment of the present invention will be described in detail with reference to the drawings. The following embodiments are examples of the embodiments of the present invention, and the present invention is not limited to these embodiments.

<First Embodiment>
The evaluation apparatus according to the first embodiment of the present invention will be described in detail with reference to the drawings. The evaluation apparatus which concerns on 1st Embodiment is an apparatus which produces | generates the evaluation reference | standard for evaluating the singing voice of the user who sings (it may be called a singer hereafter), and evaluates a singing voice by the evaluation reference | standard. This evaluation device generates a pitch range (hereinafter, also referred to as a reference pitch range) that serves as a reference for evaluation by inputting a voice (in this example, a singing voice that serves as a model (model voice)). At this time, the reference pitch range is set according to different rules in the section in which the predetermined voice technique is included in the model voice and the other sections. The model voice is not limited to a model singer, and may be a voice other than a human voice such as a synthesized voice, a voice generated by a computer, or a musical instrument.

  The evaluation device compares the reference pitch range generated in this way with the singing pitch and evaluates the singing voice. If the reference pitch range is set by the method described below, the evaluation of the singing can be prevented from being lowered even if the singing voice does not include the singing technique in the section in which the singing technique is included in the model voice. . Further, when the singing voice is included in the same section as the section in which the singing technique is included in the model voice, the evaluation can be prevented from being lowered even if the singing voice is not exactly the same as the model voice. Hereinafter, such an evaluation apparatus will be described.

[Hardware]
FIG. 1 is a block diagram showing a configuration of an evaluation apparatus 1 in the first embodiment of the present invention. The evaluation device 1 is, for example, a karaoke device. The evaluation device 1 may be a mobile device such as a smartphone, or may be configured by a plurality of devices that cooperate via a network. The evaluation device 1 includes a control unit 11, a storage unit 13, an operation unit 15, a display unit 17, a communication unit 19, and a signal processing unit 21. Each of these components is connected via a bus. In addition, a microphone 23 and a speaker 25 are connected to the signal processing unit 21.

  The control unit 11 includes an arithmetic processing circuit such as a CPU. The control unit 11 causes the CPU to execute the control program stored in the storage unit 13 to realize various functions in the evaluation device 1. The realized functions include a function for evaluating singing voice (evaluation function) and a function for generating a reference for the evaluation (evaluation reference generation function). Only one of the functions may be realized. In other words, it may be a device (evaluation device) that realizes the evaluation function without realizing the evaluation criterion generation function, or a device (evaluation criterion generation device) that realizes the evaluation criterion generation function without realizing the evaluation function. There may be.

  The storage unit 13 is a storage device such as a nonvolatile memory or a hard disk. The storage unit 13 stores a control program for realizing the evaluation function. The control program may be provided in a state stored in a computer-readable recording medium such as a magnetic recording medium, an optical recording medium, a magneto-optical recording medium, or a semiconductor memory. In this case, the evaluation device 1 only needs to include a device that reads the recording medium. The control program may be downloaded via a network.

  Moreover, the memory | storage part 13 memorize | stores music data, singing voice data, and evaluation reference data as data regarding a song. The music data includes data related to the karaoke song, for example, guide melody data, accompaniment data, and lyrics data. The guide melody data is data indicating the melody of the song. Accompaniment data is data indicating the accompaniment of a song. The guide melody data and accompaniment data may be data expressed in the MIDI format. The lyric data is data for displaying the lyrics of the song and data indicating the timing for changing the color of the displayed lyrics telop. The singing voice data is data indicating the singing voice input from the microphone 23 by the singer. In this example, the singing voice data is buffered in the storage unit 13 until the singing voice is evaluated by the evaluation function. In addition, the memory | storage part 13 may memorize | store the data which show the model audio | voice input in order to produce | generate evaluation criteria data.

  The evaluation reference data is information used as a reference for evaluation of the singing voice by the evaluation function, and is associated with music data indicating the singing song to be evaluated (the singing tune outputted when the singing voice is input). It has been. In this example, the evaluation standard data is generated based on the input of the model voice by the evaluation standard generation function, and data defining a reference pitch range (reference feature range) that changes with time as the song to be evaluated progresses. Is generated as The reference pitch range is defined by an upper limit pitch and a lower limit pitch. The reference pitch range is not limited to being defined as one range, but may be defined as a plurality of ranges. In this case, the plurality of ranges are defined as a first range between the first upper limit pitch and the first lower limit pitch and a second range between the second upper limit pitch and the second lower limit pitch. Further, the reference pitch range is determined so that the width from the reference pitch is different between a normal section and a specific section described later.

  Note that the evaluation criterion data is not limited to the case where the upper limit value and the lower limit value of the feature quantity are specified, and the normal section and the specific section are distinguished by identifiers and the like, and the reference feature quantity and the width of the feature quantity in the normal section , And the width of the feature amount in the specific section.

  The operation unit 15 is a device such as an operation button, a keyboard, or a mouse provided on an operation panel and a remote controller, and outputs a signal corresponding to the input operation to the control unit 11. The display unit 17 is a display device such as a liquid crystal display or an organic EL display, and displays a screen based on control by the control unit 11. A touch panel in which the operation unit 15 and the display unit 17 are integrated may be used. The communication unit 19 is connected to a communication line such as the Internet based on the control of the control unit 11 and transmits / receives information to / from an external device such as a server. The function of the storage unit 13 may be realized by an external device that can communicate with the communication unit 19.

  The signal processing unit 21 includes a sound source that generates an audio signal from a MIDI format signal, an A / D converter, a D / A converter, and the like. The singing voice is converted into an electric signal by the microphone 23 and input to the signal processing unit 21, and A / D converted by the signal processing unit 21 and output to the control unit 11. As described above, the singing voice is buffered in the storage unit 13 as singing voice data. The accompaniment data is read out by the control unit 11, D / A converted by the signal processing unit 21, and output from the speaker 25 as an accompaniment of the song. At this time, a guide melody may be output from the speaker 25.

[Function of Evaluation Apparatus 1]
An evaluation standard generation function and an evaluation function realized by the control unit 11 of the evaluation apparatus 1 executing a control program will be described. Part or all of the configuration for realizing the functions described below may be realized by hardware or may be realized on a server via a network.

  FIG. 2 is a block diagram showing the configuration of the evaluation reference generation function and the evaluation function in the first embodiment of the present invention. The evaluation criterion generation function 100 generates evaluation criterion data using the input model voice and stores it in the storage unit 13. The evaluation function 200 compares the input singing voice with the evaluation reference data and evaluates based on the comparison result. Each function will be specifically described below.

[Evaluation criteria generation function]
The evaluation reference generation function 100 includes a reference acquisition unit 101, a reference feature amount calculation unit 103, a specific section detection unit 105, a generation unit 107, and an evaluation accuracy setting unit 109. The reference acquisition unit 101 acquires data (reference sound data) indicating the reference sound (in this example, model voice) input to the microphone 23. At this time, music data to be associated with the generated evaluation reference data is designated. Note that the reference acquisition unit 101 may output the accompaniment sound of the song of the designated song data from the speaker 25 and may acquire the reference sound data in real time in conjunction with the progress of the song, or may be stored in advance. The obtained reference sound data may be acquired by reading from a storage device such as the storage unit 13. Note that the acquired reference sound data only needs to indicate a sound serving as a reference for evaluation, and therefore does not necessarily indicate a voice that serves as a model for singing.

  The reference feature value calculation unit 103 analyzes the model voice (reference input sound) indicated by the reference sound data acquired by the reference acquisition unit 101, and calculates a temporal change in the feature value of the model voice. In this example, the feature amount is a pitch. Therefore, the reference feature amount calculation unit 103 calculates a temporal change in the pitch of the model voice (hereinafter sometimes referred to as a model pitch), that is, a model pitch waveform. Specifically, the model pitch waveform (feature data) is calculated by a known method such as a method using the zero cross of the waveform of the model voice or a method using FFT (Fast Fourier Transform).

The specific section detection unit 105 analyzes the model pitch waveform and detects a section satisfying a predetermined condition (including a singing technique in this example) in the input period of the model voice. The detected section is called a specific section. The specific section detected at this time is associated with each type of singing technique. The detected singing technique includes, for example, the following singing techniques. In addition, if it sets beforehand as an area which satisfy | fills the predetermined specific condition, it is good also as a detection target also about pitch changes (specific transition state) other than these.
(1) Vibrato: The pitch changes finely (within a predetermined period or less). A specific example of vibrato detection is disclosed in Japanese Patent Application Laid-Open No. 2005-107087.
(2) Kobushi: The pitch temporarily increases (within a predetermined time), and then returns to the original pitch. A specific example of Kobushi detection is disclosed in Japanese Patent Laid-Open No. 2008-268370.
(3) Shaking: The pitch increases over a predetermined time and then stabilizes. A specific example of shackle detection is disclosed in Japanese Patent Laid-Open No. 2005-107334.
(4) Fall: The pitch is lowered over a predetermined time, and then the singing is interrupted. A specific example of fall detection is disclosed in Japanese Patent Laid-Open No. 2008-225115.
(5) Long tone: The pitch continues for a certain period or more in a narrow range. A specific example of long tone detection is disclosed in Japanese Patent Laid-Open No. 2008-225114.

  The generation unit 107 determines a reference pitch range based on the model pitch waveform. In this example, the reference pitch range is determined according to different rules (first rule, second rule) between the normal section (the section in which the singing voice is input and the section excluding the specific section) and the specific section. First, the first rule applied for determining the reference pitch range in the normal section will be described.

  FIG. 3 is a diagram illustrating a first rule applied to determination of a reference pitch range in a normal section according to the first embodiment of the present invention. The case where the exemplary pitch waveform is the waveform PS shown in FIG. 3 will be described as an example. First, a window CS having a width PW in the pitch direction and a width TW in the time direction is generated. Note that the width TW does not have to exist.

  The center C of the generated window CS is moved along the model pitch waveform PS. At this time, a region through which the window CS has passed is defined as a reference pitch range AW1. The upper limit of the reference pitch range AW1 is shown as the upper limit pitch UL, and the lower limit is shown as the lower limit pitch LL. For the exemplary pitch waveform PS, the range up to the upper limit pitch UL and the range up to the lower limit pitch LL are allowable ranges for the exemplary pitch. When there is a section NT where the reference pitch range AW1 does not exist due to the existence of a section where the exemplary pitch waveform PS does not exist, an identifier or the like indicating that the reference pitch range does not exist in the section NT may be included. .

  In this manner, the reference pitch range in the normal section is determined according to the first rule based on the model pitch waveform. Subsequently, the second rule applied to the determination of the reference pitch range in the specific section will be described. The second rule determines the reference pitch range by a method different from the first rule. In addition, the reference pitch range is set differently depending on the type of singing technique associated with the specific section (in this example, vibrato detection section, kobushi detection section, fall detection section, shackle detection section, and long tone detection section). decide. Hereinafter, the second rule applied to the determination of the reference pitch range in each type will be described.

  FIG. 4 is a diagram for explaining the second rule applied to the determination of the reference pitch range in the vibrato detection section in the first embodiment of the present invention. As shown in FIG. 4, the reference pitch range AW2 determined by applying the second rule in the vibrato detection section VP has a predetermined width (for example, the width of the window CS) based on the maximum value and the minimum value of the model pitch waveform PS. The pitch is half the pitch of PW. At this time, the upper limit pitch UL and the lower limit pitch LL are determined to be constant values in the vibrato detection section. In this way, the singing pitch can be included in the reference pitch range as long as the vibrato period, peak timing, etc. do not match, as long as the vibrato is close to some extent.

  Such a determination method is an example, and the reference pitch range AW2 in the vibrato detection section VP includes the entire exemplary pitch waveform PS, and is determined by applying different rules from the determination method of the reference pitch range AW1 in the normal section NP. It only has to be done.

  FIG. 5 is a diagram for explaining the second rule applied to the determination of the reference pitch range in the bump detection section in the first embodiment of the present invention. As shown in FIG. 5, the reference pitch range AW3 determined by applying the second rule in the Kobushi detection section KP is determined in the same manner as the first rule with respect to the upper limit pitch UL based on the exemplary pitch waveform PS. ing. On the other hand, the lower limit pitch LL is determined in the same manner as the first rule with reference to the waveform VS assuming that the model voice does not include the Kobushi technique. In this way, when singing that does not include the Kobushi technique, it is out of the standard pitch range, that is, the Kobushi technique must be performed according to the model voice, and it will be disliked. can do.

  Such a determination method is an example, and the reference pitch range AW3 in the Kobushi detection section KP includes the entire exemplary pitch waveform PS, and is determined by applying different rules from the determination method of the reference pitch range AW1 in the normal section NP. It only has to be done.

  FIG. 6 is a diagram illustrating a second rule applied to determination of the reference pitch range in the fall detection section according to the first embodiment of the present invention. As shown in FIG. 6, the reference pitch range determined by applying the second rule in the fall detection section FP includes a first range AW4 and a second range AW5. The upper limit pitch UL1 and the lower limit pitch LL1 of the first range AW4 are determined in the same manner as the first rule with the model pitch waveform PS as a reference. On the other hand, the upper limit pitch UL2 and the lower limit pitch LL2 of the second range AW5 are determined in the same manner as the first rule on the basis of the waveform VS assuming that the exemplary voice does not include the fall technique. In this way, when you sing a song that does not include the fall technique, you will be excluded from the reference pitch range, that is, if you do not necessarily execute the fall technique as in the model voice, it will be excluded. can do.

  Note that the entire pitch range included between the upper limit pitch UL2 and the lower limit pitch LL1 may be determined as the reference pitch range without considering the upper limit pitch UL1 and the lower limit pitch LL2. In this way, it is possible to exclude the low evaluation even when the fall technique has a small number of pitch drops. Such a determination method is an example, and the reference pitch range (the first range AW4 and the second range AW5) in the fall detection section FP includes the entire exemplary pitch waveform PS in a part of either range, It is only necessary to apply a rule different from the method for determining the reference pitch range AW1 in the normal section NP.

  FIG. 7 is a diagram illustrating a second rule applied to determination of the reference pitch range in the shackle detection section in the first embodiment of the present invention. As shown in FIG. 7, the reference pitch range determined by applying the second rule in the shackle detection section SP includes a first range AW6 and a second range AW7. The upper limit pitch UL1 and the lower limit pitch LL1 of the first range AW6 are determined in the same manner as the first rule with the model pitch waveform PS as a reference. On the other hand, the upper limit pitch UL2 and the lower limit pitch LL2 of the second range AW7 are determined in the same manner as the first rule on the basis of the waveform VS assuming that the model voice does not include the brushing technique. In this way, when singing that does not include the Shakuri technique, it is out of the standard pitch range, that is, it is excluded that the Shakuri technique must be performed according to the model voice, and it will be disliked. can do.

  Note that the entire pitch range included between the upper limit pitch UL2 and the lower limit pitch LL1 may be determined as the reference pitch range without considering the upper limit pitch UL1 and the lower limit pitch LL2. In this way, even if there are few ways to raise the pitch of the shackle technique, it can be excluded that the evaluation is low. Such a determination method is an example, and the reference pitch range (the first range AW6 and the second range AW7) in the shackle detection section SP includes the entire exemplary pitch waveform PS in a part of either range, It is only necessary to apply a rule different from the method for determining the reference pitch range AW1 in the normal section NP.

  FIG. 8 is a diagram illustrating the second rule applied to the determination of the reference pitch range in the long tone detection section in the first embodiment of the present invention. As shown in FIG. 8, the reference pitch range AW8 determined by applying the second rule in the long tone detection section LP has a predetermined width (for example, a pitch half of the width PW of the window CS) based on the model pitch waveform PS. ) Is an expanded range. In the long tone, if the pitch is deviated, the deviation becomes conspicuous. Therefore, it is possible to tighten the criterion in the long tone detection section.

  At this time, the width may be narrower in the section after the long tone detection section LP. Such a determination method is an example, and the reference pitch range AW8 in the long tone detection section LP includes the entire exemplary pitch waveform PS, and rules different from the determination method of the reference pitch range AW1 in the normal section NP are applied. It only has to be decided.

  In this way, the generation unit 107 determines the reference pitch range, and generates evaluation reference data defined by the parameters indicating the reference pitch range (for example, the upper limit pitch and the lower limit pitch).

  Returning to FIG. 2, the description will be continued. The evaluation accuracy setting unit 109 sets the accuracy of singing evaluation. The evaluation accuracy is set by the user. The evaluation accuracy is a parameter for determining an allowable range in at least one of the first rule and the second rule in the generation unit 107. For example, when the evaluation accuracy increases, the generation unit 107 generates the evaluation reference data so that the reference pitch range (AW1, AW2,...) Is narrowed. Specifically, as the evaluation accuracy is set higher, the reference pitch range is determined so as to reduce the area (in particular, the width PW) of the window CS described above. Note that the evaluation accuracy setting unit 109 may not exist. If it does not exist, the evaluation accuracy may be set to a predetermined value.

[Evaluation function]
Next, the evaluation function 200 will be described. The evaluation function 200 includes an accompaniment output unit 201, a singing sound acquisition unit 203, a singing feature amount calculation unit 205, an evaluation criterion acquisition unit 207, a feature amount comparison unit 209, and a voice evaluation unit 211. The accompaniment output unit 201 reads the accompaniment data corresponding to the song tune designated by the singer, and outputs the accompaniment sound from the speaker 25 via the signal processing unit 21. The singing sound acquisition unit 203 acquires singing sound data indicating the singing sound input from the microphone 23. In this example, the input sound to the microphone 23 during the period in which the accompaniment sound is output is recognized as the singing sound to be evaluated. The singing sound acquisition unit 203 acquires the singing voice data buffered in the storage unit 13, but may acquire the singing voice data of the entire song stored in the storage unit 13, or the signal processing unit. You may acquire directly from 21.

  The singing feature amount calculation unit 205 analyzes the singing voice data acquired by the singing sound acquisition unit 203, and calculates a temporal change in the singing voice feature amount. In this example, as described above, the feature amount is a pitch. Accordingly, the singing feature amount calculation unit 205 calculates a temporal change in the pitch of the singing voice (hereinafter, sometimes referred to as a singing pitch), that is, a singing pitch waveform. The calculation method is the same as the above-described exemplary pitch waveform calculation method.

  The evaluation reference acquisition unit 207 reads out and acquires evaluation reference data associated with the song data of the song to be evaluated (accompaniment sound output from the accompaniment output unit 201) from the storage unit 13. The feature amount comparison unit 209 compares the reference pitch range indicated by the evaluation reference data with the singing pitch.

  FIG. 9 is a diagram for explaining a song evaluation method using a song pitch in the first embodiment of the present invention. The waveform IS (singing pitch waveform) indicating the time change of the singing pitch calculated by the singing feature amount calculation unit 205 is compared with a reference pitch range determined by the upper limit pitch UL and the lower limit pitch LL. FIG. 9 shows an example in which the singing pitch is calculated until a timing slightly exceeding the vibrato detection section VP. At this time point, the singing pitch waveform IS has a pitch lower than the lower limit pitch LL in two outliers NG and is out of the reference pitch range. As one parameter in the evaluation result of the singing voice, for example, the higher the proportion of outliers NG in all the sections, the higher the mismatch degree.

  Also, in the vibrato detection section VP, even if the vibrato period, peak timing, etc. in the singing pitch waveform IS are shifted, the reference pitch range is determined by the second rule described above, so that the outlier section NG does not increase. can do. As a result, the degree of inconsistency can be prevented from becoming high. Thus, according to this singing evaluation method, by defining the allowable range for the model pitch waveform and preliminarily setting it as the reference pitch range, the singing evaluation standard is stabilized and the evaluation redundancy for the singing technique is increased. You can have it. In addition, when the singing pitch waveform is included in the section NT, it may not be considered as an evaluation target, or may be treated as an outlier section NG.

  Returning to FIG. 2, the description will be continued. The voice evaluation unit 211 calculates an evaluation value serving as an index for evaluating the singing voice based on the comparison result in the feature amount comparison unit 209. In this example, the higher the mismatch degree calculated by the feature amount comparison unit 209, the lower the evaluation value is calculated, and the evaluation of the singing voice becomes worse. Note that the voice evaluation unit 211 may calculate the evaluation value based on another element, instead of calculating the evaluation value based only on the degree of mismatch. As other elements, a singing technique and other parameters that can be extracted from the singing voice data are assumed. The evaluation result by the voice evaluation unit 211 may be presented on the display unit 17. The above is the description of the evaluation reference generation function 100 and the evaluation function 200.

Second Embodiment
In the first embodiment described above, the pitch is used as the feature amount. In the second embodiment, an example will be described in which a first-order differential value (volume speed) is used as the feature amount. It should be noted that since the contents of the feature values calculated by the reference feature value calculation unit 103 and the singing feature value calculation unit 205 are different, detailed description thereof is omitted.

  FIG. 10 is a diagram for explaining a reference volume speed range in the second embodiment of the present invention. FIG. 10A shows a waveform AS1 of the model volume calculated from the model sound. FIG. 10B shows a volume velocity waveform AS2 obtained by first-order differentiation of the model volume waveform AS1 with respect to time. Further, with reference to this volume speed waveform AS2, reference volume speed ranges AW10 and AW11 (upper limit volume) determined using rules corresponding to the plurality of rules (first rule and second rule) described in the first embodiment. Also shown is between the speed UL and the lower limit volume speed LL. The reference volume speed range AW10 is a volume speed range in the normal section NP. The reference volume speed range AW11 is a volume speed range in the vibrato detection section VP.

  Evaluation reference data indicating the reference volume speed range determined in this manner is stored in the storage unit 13. Then, when evaluating the singing sound, the volume speed calculated as the feature amount from the singing voice is compared with the reference volume speed range. In this way, the direct current component of the volume is removed by performing the first-order differentiation, and the tendency of the volume change is compared. Thereby, even if the input level of the singing sound is greatly different from the input level of the model voice, the comparison using the tendency of change can be performed. When the volume speed is used, for example, techniques such as singing inflection and breathing can be detected in addition to the exemplified vibrato.

  In addition, the feature amount may be a frequency distribution. For example, by using a power ratio between a fundamental tone and a harmonic overtone or a parameter generated based on the power ratio as a feature amount, it is possible to perform an evaluation based on a temporal change in voice quality. Regardless of which feature quantity is adopted, in the specific section in which a special song included in the model voice is detected, the second rule is the reference feature quantity range even if the special voice is not included in the song voice. Is set to be included.

<Third Embodiment>
In the third embodiment, an example in which the reference pitch range included in the evaluation reference data is determined in two stages will be described. In the following description, the reference pitch range is defined in two steps, but it may be defined in a plurality of more steps.

  FIG. 11 is a diagram illustrating a method for determining a reference pitch range in the normal section and the vibrato detection section in the third embodiment of the present invention. FIG. 11 is a diagram corresponding to FIG. As shown in this figure, in the normal section NP, the reference pitch range including the exemplary pitch waveform PS includes a high evaluation range AW1 and a low evaluation range AW1a. In the vibrato detection section VP, the reference pitch range including the exemplary pitch waveform PS includes a high evaluation range AW2 and a low evaluation range AW2a. In this example, the low evaluation ranges AW1a and AW2a are between an upper limit pitch ULa that is increased by a predetermined pitch with respect to the upper limit pitch UL and a lower limit pitch LLa that is decreased by a predetermined pitch with respect to the lower limit pitch LL. The pitch width.

  Thus, the evaluation reference data may indicate a reference pitch range divided into a plurality of stages such as a high evaluation range and a low evaluation range. The reference pitch range divided into a plurality of stages in this way can be applied in various ways in the evaluation of singing voice. First, in the first embodiment, determination is made based on whether or not the singing pitch waveform is included in the reference pitch range, and the degree of deviation is not considered. On the other hand, as in this example, by providing a plurality of steps in the reference pitch range, it is also possible to specify a section in which the singing pitch waveform is included between the upper limit pitches ULa and UL or between the lower limit pitches LLa and LL. it can. According to this, a more detailed singing voice can be evaluated by weighting the calculation of the degree of inconsistency by distinguishing between a case where the singing pitch waveform is slightly separated from the reference pitch range and a case where it is completely separated from the reference pitch range. Can do.

  Alternatively, the upper limit pitches ULa and UL and the lower limit pitches LLa and LL may be distinguished from each other, and it may be determined between which the singing pitch waveform is included. According to this, it can be evaluated whether the singing pitch is shifted to the sharp side (high pitch side) or the flat side (low pitch side).

  Further, the reference pitch width AW1 between the upper limit pitch UL and the lower limit pitch LL, a mismatch degree 1 when comparing the AW2 and the singing pitch waveform, and the reference pitch width AW1a between the upper limit pitch ULa and the lower limit pitch LLa, When the AW 2a is compared with the singing pitch waveform, the feature quantity comparison unit 209 calculates the mismatch degree 2 and uses the relative relationship (eg, ratio) between the mismatch degree 1 and the mismatch degree 2 as a parameter for evaluation. You can also.

  Further, the pitch width AW1Ua and the pitch width AW1La may be different from each other instead of the same width. For example, by narrowing the pitch width AW1Ua and making the pitch width AW1La larger than the pitch width AW1Ua, it is possible to make it stricter for a song that touches the sharp side.

<Fourth embodiment>
In the fourth embodiment, an evaluation function 200A that detects the singing technique from the singing voice and changes the processing content of the feature amount comparison unit 209A or the voice evaluation unit 211A will be described.

  FIG. 12 is a block diagram showing the configuration of the evaluation function in the fourth embodiment of the present invention. Of the configuration of the evaluation function 200A, a configuration different from the evaluation function 200 in the first embodiment described above will be described. The specific section detection unit 213 has the same function as the specific section detection unit 105 of the evaluation reference generation function 100 in the first embodiment, and detects a second specific section including a specific singing technique from the singing pitch waveform. In the following description, the specific section detected by the specific section detection unit 105 is described as a first specific section and is distinguished from the second specific section.

  The feature amount comparison unit 209A calculates the degree of inconsistency by substantially the same processing as the feature amount comparison unit 209 in the first embodiment, but differs in the following points. The feature amount comparison unit 209A determines whether there is a section in which the normal section that is not the first specific section in the evaluation reference data and the second specific section corresponding to the specific singing technique overlap in the singing pitch waveform. . If there are overlapping sections, the overlapping sections are not used for calculating the degree of mismatch. When the singer uses the singing technique in a section different from the model voice, the reference pitch range may be deviated. By such processing, an increase in the degree of inconsistency can be suppressed. In addition, in 1st Embodiment which does not perform such a process, what is necessary is just to be used when the song close | similar to a model voice is calculated | required.

  In this example, the voice evaluation unit 211A calculates the evaluation value by substantially the same process as the voice evaluation unit 211 in the first embodiment, but the second specific section overlaps with the normal section in the evaluation reference data as described above. If so, the evaluation value is calculated so that the evaluation is high. Note that the voice evaluation unit 211A does not increase the evaluation in such a case, but calculates the evaluation value so that the evaluation increases when the first specific section and the second specific section match. May be.

<Fifth Embodiment>
In the embodiment described above, the specific section detection unit 105 detects a specific section (evaluation target section) in which a waveform corresponding to the singing technique is included in the exemplary pitch waveform. In the fifth embodiment, the specific section detection unit 105 detects a specific section (evaluation exclusion section) for excluding evaluation when the exemplary pitch waveform includes a waveform that satisfies a specific condition. . Therefore, the specific section includes an evaluation target section and an evaluation exclusion section that are determined according to conditions that the exemplary pitch waveform satisfies. The evaluation target section is a section including the singing technique as described above. On the other hand, the evaluation exclusion section may be, for example, a section where the model pitch changes steeply or a section where the model pitch cannot be calculated and dispersed. Then, according to the second rule applied in the generation unit 107, the same processing as that of the first embodiment is executed for the evaluation target section, whereas the evaluation exclusion section is not evaluated. It is only necessary to execute processing for adding an identifier indicating that to the evaluation reference data. That is, it can be said that the evaluation reference data in which the allowable range is set to “0” by the second rule is generated in the specific section that is not the evaluation target.

  In addition, it is good also considering all the specific areas as an evaluation exclusion area. In this case, a section including the singing technique may be set as an evaluation exclusion section. And when the 2nd specific area demonstrated in 4th Embodiment overlaps with an evaluation exclusion area (or when included in an evaluation exclusion area), an evaluation value may be calculated so that evaluation may become high.

<Sixth Embodiment>
In the embodiment described above, the reference feature value calculation unit 103 calculates a feature value (for example, a model pitch) from the model voice acquired as the reference input sound. As described above, it has been described that the model voice may be a voice other than a human voice, such as a synthesized voice, a voice generated by a computer, or a musical instrument sound, as described above. Several examples will be described for the case where data other than the waveform data of simple sound is used as the reference data for feature amount calculation.

  The reference data may be sequence data that defines a sound, such as MIDI format data. 3 may be data obtained by handwriting input of a waveform PS as shown in FIG. 3 (scanning a waveform drawn on paper and inputting it as image data into a computer or using the operation unit 15). Such handwriting input may be performed while listening to a song as a model. As described above, the reference data may not be data indicating the sound itself (exemplary sound in the first embodiment) as long as it is data that can be converted into a temporal change in the sound feature amount.

  In addition, when the reference data is model voice as in the first embodiment, the specific section is defined as a section in which the temporal change of the feature amount satisfies a specific condition. As a specific condition, the condition that a predetermined singing technique exists has been described as a specific example. On the other hand, when the feature value is outside the predetermined range (if the feature value is pitch, the sound is high or low), whether or not a specific condition is satisfied is determined by the time of the feature value. The value of the feature value at each timing is not limited to the case where the change is a target, and may be the target.

  Moreover, when the information of song music is contained in reference | standard data, the specific area changed from the tolerance | permissible_range of a normal area may be detected based on the information of this song music. The information on the song includes, for example, composition information of the song (singing, rust, A melody, B melody, head of A melody, ending, etc.), range information (treble, bass, etc.), and sound information (chord constituent sounds, Progress sound, etc.). Note that the feature amount described above is also information obtained by calculation from reference data (model voice), and thus is a kind of information included in the reference data.

  The change amount of the allowable range may be set by the user via the operation unit 15 or may be set in advance. At this time, the change amount of the allowable range may be set to be different depending on the type of information included in the reference data. The allowable range is not only changed between the normal section and the specific section, but may be changed without distinguishing between sections. The evaluation reference generation function 100B in such a case will be described.

  FIG. 13 is a block diagram showing the configuration of the evaluation reference generation function in the sixth embodiment of the present invention. As illustrated in FIG. 13, the evaluation criterion generation function 100B may not include the specific section detection unit 105 in order to execute the process without distinguishing the specific section. On the other hand, the generation unit 107B acquires information included in the reference data from the reference acquisition unit 101 or the reference feature amount calculation unit 103, and determines the reference data range while changing the allowable range based on this information. Thus, the configuration for detecting the specific section is not necessarily included.

<Seventh embodiment>
In the embodiment described above, the generation unit 107 determines the reference pitch range according to different rules for the normal section and the specific section. In the seventh embodiment, when determining the reference pitch range, the normal interval and the specific interval are not divided, and the reference pitch range is determined according to one rule. The one rule may be a rule corresponding to the first rule described above as long as the reference pitch range is determined based on the model pitch waveform, and other rules, for example, the lower pitch side A rule for increasing the allowable range, a rule for changing the allowable range in accordance with the amount of change in the exemplary pitch waveform, and the like may be applied. In such a case, since it is not necessary to detect a specific section to which the second rule in the above-described embodiment is applied, the specific section detection unit 105 may not exist.

  When the second rule is not used in this way, the reference feature value calculation unit 103 performs a filter process such as LPF (low-pass filter) on the waveform indicating the temporal change in the feature value of the model voice, and uses this as a model. It may be a pitch waveform (feature data). For example, the model pitch waveform obtained through the LPF is obtained by removing the vibrato waveform, for example. In this state, if the allowable range is set wider, the evaluation redundancy for the singing technique is improved. Even when the second rule is used, the reference feature value calculation unit 103 may calculate an exemplary pitch waveform subjected to the filter process.

DESCRIPTION OF SYMBOLS 1 ... Evaluation apparatus, 11 ... Control part, 13 ... Memory | storage part, 15 ... Operation part, 17 ... Display part, 19 ... Communication part, 21 ... Signal processing part, 23 ... Microphone, 25 ... Speaker, 100 ... Evaluation criteria production | generation function DESCRIPTION OF SYMBOLS 101 ... Standard acquisition part 103 ... Reference | standard feature-value calculation part 105 ... Specific area detection part 107 ... Production | generation part 109 ... Evaluation precision setting part 200, 200A ... Evaluation function, 201 ... Accompaniment output part, 203 ... Signal Acquiring unit, 205 ... Singing feature value calculating unit, 207 ... Evaluation standard acquiring unit, 209, 209A ... Feature value comparing unit, 211, 211A ... Voice evaluating unit, 213 ... Specific section detecting unit

Claims (11)

A reference acquisition unit for acquiring reference data;
A reference feature amount calculation unit for calculating feature amount data indicating a temporal change in the sound feature amount from the reference data;
A reference feature amount range set to include the feature amount indicated by the feature amount data as a reference, including an allowable range from the reference, and the allowable range is changed based on information included in the reference data A generation unit that generates evaluation reference data that defines a reference feature amount range;
An evaluation criterion generating device comprising: From the feature amount data, further comprising a detection unit for detecting a specific section in which the feature amount satisfies a specific condition,
The allowable range of the reference feature amount range defined by the evaluation reference data generated by the generation unit is set by the first rule outside the specific section, and is set by the second rule in the specific section. 2. The evaluation reference generation device according to claim 1, wherein the allowable range is changed based on information included in the reference data. An evaluation accuracy setting unit for setting the evaluation accuracy is further provided.
The evaluation reference generation device according to claim 1, wherein the generation unit changes the allowable range based on the evaluation accuracy set by the evaluation accuracy setting unit.   The evaluation criterion generation apparatus according to claim 1, wherein the allowable range is divided into a plurality of evaluation ranges different in evaluation method. The reference data indicates an input sound,
The feature amount includes a pitch of the input sound,
The allowable range set by the first rule is a pitch width from the pitch of the input sound,
The evaluation criterion generation apparatus according to claim 2, wherein the allowable range set by the second rule is a range wider than the allowable range set by the first rule.   The second rule includes a rule for setting an allowable range including a first pitch width including the pitch of the reference sound and a second pitch width which does not include the pitch and is not continuous with the first pitch width. 6. The evaluation reference generation device according to claim 5, wherein   The evaluation reference generation device according to claim 1, wherein the feature amount includes a differential value of a volume of the reference sound. The specific section includes an evaluation target section and an evaluation exclusion section that are determined according to a condition that the feature amount satisfies,
The said 2nd rule sets the said tolerance | permissible_range in the said evaluation object area, and adds the identifier which shows that it is not an evaluation object in the said evaluation exclusion area to the said evaluation reference data. Evaluation standard generator.   The reference feature amount calculation unit calculates the feature amount data indicating a waveform obtained by performing a predetermined filtering process on a waveform indicating a temporal change in a sound feature amount from the reference data. Item 2. The evaluation criterion generating device according to Item 1. The evaluation reference generation device according to any one of claims 1 to 9,
A singing sound acquisition unit for acquiring a singing input sound;
A singing feature value calculation unit for calculating singing feature value data indicating a temporal change in the feature value from the singing input sound;
A comparison unit that compares the reference feature amount range indicated by the evaluation reference data with a temporal change in the feature value indicated by the singing feature amount data;
Based on the compared results, an evaluation unit that calculates an evaluation value for the singing input sound;
A singing evaluation apparatus comprising: A singing sound acquisition unit for acquiring a singing input sound;
A singing feature amount calculation unit for calculating singing feature amount data indicating a temporal change of the feature amount from the singing input sound;
A reference feature amount range that is set to include a reference feature amount that serves as a reference for evaluation of the feature amount, includes an allowable range from the reference feature amount, and the allowable feature is changed according to a section. An evaluation criteria acquisition unit for acquiring evaluation criteria data defining a quantity range;
A comparison unit that compares the reference feature amount range indicated by the evaluation reference data with a temporal change in the feature value indicated by the singing feature amount data;
Based on the compared results, an evaluation unit that calculates an evaluation value for the singing input sound;
A singing evaluation apparatus comprising:

Images