JP2010197738A - Tone pitch determination system, register determination system, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To properly determine tone pitch and register corresponding to a user. <P>SOLUTION: A transition pattern of singing tone pitch along a time axis when a user sings a song is compared with a transition pattern of a guide tone pitch k when the song is sung properly, thereby calculating an error d between the transition patterns and reflecting it in an error distribution of the user. Then, the guide tone pitch k where the error ratio obtained by comparing differences Δd from errors d in adjacent guide tone pitches k-1 and k+1 becomes minimum is determined to be the highest tone pitch kup in the register of the user corresponding to the error distribution (s310-s330). Similarly, the guide tone pitch k where the error ratio obtained by comparing differences from errors in adjacent guide tone pitches k-1 and k+1 becomes minimum is determined to be the lowest pitch klo in the register of the user corresponding to the error distribution (s410-s430). The register of the user is determined based on the highest pitch kup and the lowest pitch klo (s260). <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a system for determining a pitch in a user's range.

  In recent years, various techniques for specifying the range of a user who has sung music have been proposed.

  For example, while the user sings a song, the highest pitch (highest pitch) and lowest pitch (lowest pitch) in the sound input from the microphone is detected, and from the highest pitch thus detected to the lowest pitch Is determined to be the sound range of the user (see Patent Document 1).

JP 2002-73058 A

  However, since the above-mentioned technique simply determines the range of the highest pitch and the lowest pitch detected when singing as the corresponding user's range, it is detected regardless of whether or not the voice is properly spoken. The determined pitch is determined as a pitch that the user can properly utter, and furthermore, the range defined by the detected pitch is determined as the user's range.

  Specifically, assuming that a song having a pitch section that is difficult for a certain user to sing is sung, even if the song temporarily reaches an appropriate pitch in that section, the pitch in that section is The transition pattern is likely to be greatly different from the transition pattern when the section is appropriately sung.

  About such a pitch, it can be said that the pitch of the user is uttered forcibly, and it is difficult to say that the pitch can be uttered in a state suitable for singing. It is desirable to determine the user's range so that it is not determined and does not include this pitch.

  The present invention has been made to solve such a problem, and an object of the present invention is to provide a technique for determining a pitch and a sound range that a user can utter more appropriately.

In order to solve the above problem, the first configuration is as follows.
Singing data indicating the transition of the pitch along the time axis associated with the singing of the music by the user, and guide data indicating the transition of the pitch along the time axis when the singing music related to the singing data is appropriately sung , The transition patterns of the pitches (hereinafter referred to as “singing pitches”) in the respective unit sections (hereinafter referred to as “singing intervals”) on the time axis indicated by the song data are indicated by the guide data. Both transition patterns are compared with the transition pattern of the pitch (hereinafter referred to as “guide pitch”) in the guide section corresponding to the corresponding singing section in each unit section (hereinafter referred to as “guide section”) on the time axis. Error calculating means for calculating an error Δd [k] of a singing pitch with respect to a guide pitch k (= 1 to n) to be uttered in the guide section And,
Among the error distributions prepared for each of a plurality of users and distributing the error of the singing pitch with respect to the guide pitch of the user for each of the guide pitches, this is an opportunity for calculation by the error calculation means. Each of the calculated errors is additionally distributed in the error distribution corresponding to the user who performed the song as an error d [k] of the singing pitch with respect to the guide pitch k in the guide section referred to in the calculation. A distribution updating means for updating the error distribution;
Among the guide pitches k in the error distribution updated by the distribution updating means, the difference Δd [k−1] (= d [) from the error d [k−1] in the guide pitch k−1 adjacent to the lower side. k] −d [k−1]) is compared with the difference Δd [k] (= d [k + 1] −d [k]) from the error d [k + 1] at the guide pitch k + 1 adjacent to the higher side. A guide pitch k that minimizes | Δd [k−1] / Δd [k] | is extracted, and the extracted guide pitch k is determined as the highest tone kup in the user's range corresponding to the error component. And a highest sound determination means.

  In the pitch determination system in this configuration, first, the transition pattern of the pitch (singing pitch) along the time axis associated with the user's singing, and the transition of the pitch (guide pitch) when the song is properly sung By comparing with the pattern, the error d of these transition patterns is calculated and reflected in the error distribution of the user.

  Then, of the guide pitches 1 to n in the error distribution, the guide that minimizes the error ratio | Δd [k−1] / Δd [k] | that compares each difference Δd with the error d in the adjacent guide pitch. The pitch k is determined as the highest sound kup in the user's range corresponding to the error distribution.

  The error distribution is such that an error d [k] as a transition pattern of the singing pitch with respect to the guide pitch k in the specific guide section is distributed for each guide pitch k, and therefore the guide having a small error d [k] is distributed. The pitch k can be said to be a pitch that the corresponding user can properly utter in a similar transition pattern, whereas a guide pitch whose error d [k] is large is uttered by the corresponding user forcibly. It can be said that the pitch.

  Then, in the inflection region on the high pitch side from the guide pitch k that the user can properly utter to the guide pitch k + 1 that is forcibly uttered, the error d in the guide pitch k−1 adjacent to the low side. The difference Δd [k] from the error d [k] at the guide pitch k + 1 adjacent to the higher side tends to increase rapidly with respect to the difference Δd [k−1] from [k−1]. This tendency is expected to be most noticeable in the vicinity of the highest guide pitch k at which the user can speak properly. In this case, the error ratio | Δd [k−1] / Δd [ k] | indicates a minimum value when the difference Δd [k], which is the denominator, is significantly increased.

  For this reason, in the above configuration, the guide pitch k at which the error ratio | Δd [k−1] / Δd [k] | is the minimum is the guide pitch k that the user can properly utter. Of these, the highest pitch kup can be determined.

  In this configuration, it is desirable from the viewpoint of processing load and determination accuracy to limit the range of the guide pitch k referred to when determining the highest sound kup to some extent. As a configuration for this purpose, for example, it is conceivable to make the above configuration as a second configuration (claim 2) shown below.

  In this configuration, the highest sound determination unit has the difference Δd [k] larger than 0 in the guide pitch k in the error distribution updated by the distribution update unit, and the error ratio | Δd [k− 1] / Δd [k] |, the guide pitch that minimizes is extracted.

  In this configuration, the range of the guide pitch k that is referred to when determining the highest sound kup is limited to a range in which the difference Δd [k] is greater than zero.

  In the inflection region described above, it is expected that the error d [k] in the guide pitch k + 1 adjacent to the higher side will increase as the guide pitch k becomes higher than the highest sound that can be properly spoken. Therefore, the difference Δd [k] (= d [k + 1] −d [k] between the error d [k] and the error d [k + 1] is naturally larger than 0.

  Then, even if the target guide pitch k is limited to a difference Δd [k] larger than 0, the guide pitch k to be extracted as the highest sound kup is not excluded from the extraction target. The processing load can be reduced.

  Furthermore, since only the appropriate guide pitch k is targeted, it is possible to prevent the determination accuracy of the highest sound kup from being lowered due to extraction of an inappropriate guide pitch k. be able to.

  Further, in each of the above configurations, the guide pitch k that is the highest tone kup is in the inflection region on the higher tone side of all the guide pitches in the error distribution, and therefore the guide pitch that is referred to when determining the highest tone kup. The range of k may be limited to a band that forms a high-frequency inflection region. As a configuration for this purpose, for example, it is conceivable to make each of the above configurations as a third configuration (claim 3) shown below.

  In this configuration, among a plurality of guide pitches in the error distribution updated by the distribution updating means, a band composed of a predetermined number of guide pitches located on the higher pitch side is extracted as a high frequency error inflection band. High frequency error inflection band extracting means. Then, the highest sound determination means extracts the guide pitch that becomes the highest sound kup from the guide pitches in the high frequency error inflection band extracted by the high frequency error inflection band extraction means.

  In this configuration, the processing load required for the determination can be suppressed by limiting the range of the guide pitch k referred to when determining the highest sound kup to the error inflection band on the high sound side in advance.

  Further, in this configuration, the specific configuration for extracting the treble error inflection band is not particularly limited. For example, the fourth configuration (claim 4) described below may be used.

  In this configuration, the high frequency error inflection band extracting means is located on the higher pitch side among the plurality of guide pitches in the error distribution updated by the distribution updating means, and is adjacent to each other. A treble error inflection band made up of guide pitches each having a certain rate of change in error with respect to the guide pitch is extracted.

  In the band on the high pitch side in the error distribution, the error is expected to increase as the guide pitch k becomes higher than the pitch that can be properly spoken. The degree of change in error between adjacent guide pitches also increases.

  Therefore, in the above configuration, as a rate of change when extracting the band, by setting a value assumed as a rate of change in the pitch above the guide pitch that can be appropriately spoken by the user as a “constant rate of change”, An appropriate band can be extracted as the treble error inflection band.

  More specifically, this configuration is preferably a fifth configuration (claim 5) described below.

  In this configuration, the high frequency error inflection band extracting means predetermines a guide pitch positioned at least on the higher pitch side among a plurality of guide pitches in the error distribution updated by the distribution updating means. In addition to classifying into guide groups for each guide pitch, and averaging the guide pitch errors classified into the same group for each group, the average error in each group is adjacent to the treble side. For one or more groups including the group having the maximum adjacent error ratio excluding the average error in the group, a band composed of the guide pitches classified into these groups is extracted.

  With this configuration, the guide pitch errors for each group consisting of a predetermined number of guide pitches are averaged, and one or more groups in which the adjacent error ratio between adjacent groups is maximized are included in this group. It is possible to extract a treble error inflection region consisting of each guide pitch.

  In this configuration, the “guide pitch located on the higher pitch side” to be classified as a group only needs to be located on the higher pitch side in the error distribution. It is conceivable to set a certain ratio (for example, several tens of percent) on the high pitch side in the entire pitch range.

In order to solve the above problem, the sixth configuration is:
Singing data indicating the transition of the pitch along the time axis associated with the singing of the music by the user, and guide data indicating the transition of the pitch along the time axis when the singing music related to the singing data is appropriately sung , The transition patterns of the pitches (hereinafter referred to as “singing pitches”) in the respective unit sections (hereinafter referred to as “singing intervals”) on the time axis indicated by the song data are indicated by the guide data. Both transition patterns are compared with the transition pattern of the pitch (hereinafter referred to as “guide pitch”) in the guide section corresponding to the corresponding singing section in each unit section (hereinafter referred to as “guide section”) on the time axis. Error calculating means for calculating an error Δd [k] of a singing pitch with respect to a guide pitch k (= 1 to n) to be uttered in the guide section And,
Among the error distributions prepared for each of a plurality of users and distributing the error of the singing pitch with respect to the guide pitch of the user for each of the guide pitches, this is an opportunity for calculation by the error calculation means. Each of the calculated errors is additionally distributed in the error distribution corresponding to the user who performed the song as an error d [k] of the singing pitch with respect to the guide pitch k in the guide section referred to in the calculation. A distribution updating means for updating the error distribution;
Among the guide pitches k in the error distribution updated by the distribution updating means, the difference Δd [k] (= d [k + 1] −d [k] from the error d [k + 1] in the guide pitch k + 1 adjacent to the higher side. ]) In comparison with the difference Δd [k−1] (= d [k] −d [k−1]) from the error d [k−1] in the guide pitch k−1 adjacent to the lower side. A guide pitch k that minimizes | Δd [k] / Δd [k−1] | is extracted, and the extracted guide pitch k is determined as the lowest tone klo in the user's range corresponding to the error component. Minimum sound determination means.

  In the pitch determination system in this configuration, as in the above configuration, after updating the error distribution in the user, each of the differences from the error in the adjacent guide pitch among the guide pitches 1 to n in this error distribution is calculated. The guide pitch k at which the compared error ratio | Δd [k] / Δd [k−1] | is minimized is determined as the lowest sound klo in the user's range corresponding to the error component.

  As described above, the guide pitch k having a small error Δd [k] can be said to be a pitch at which the corresponding user can properly utter with a similar transition pattern, whereas the guide pitch having a large error Δd [k]. High can be said to be the pitch that the corresponding user is forcing.

  Then, in the inflection region on the low pitch side from the guide pitch that the user has forcibly uttered to the guide pitch that can be properly uttered, the error d [k + 1] in the guide pitch k + 1 adjacent to the high side is The difference Δd [k−1] from the error d [k−1] in the guide pitch k−1 adjacent to the lower side tends to increase rapidly with respect to the difference Δd [k]. This tendency is expected to be most noticeable in the vicinity of the lowest guide pitch k at which the user can properly speak. In this case, the error ratio | Δd [k] / Δd [k− 1] | indicates the minimum value when the difference Δd [k−1], which is the denominator, is significantly increased.

  For this reason, in the above configuration, the guide pitch k at which the error ratio | Δd [k] / Δd [k−1] | Of these, the lowest pitch klo can be determined.

  In this configuration, it is desirable from the viewpoint of processing load and determination accuracy that the range of the guide pitch k referred to when determining the lowest sound klo is limited to some extent. As a configuration for this purpose, for example, the above configuration can be considered as a seventh configuration (claim 7) described below.

  In this configuration, the lowest sound determination means has the difference Δd [k] smaller than 0 in the guide pitch k in the error distribution updated by the distribution update means, and an error ratio | Δd [k] / The guide pitch k that minimizes Δd [k−1] | is extracted.

  In this configuration, the range of the guide pitch k that is referred to when determining the lowest sound klo is limited to a range in which the difference Δd [k] is smaller than zero.

  In the inflection region described above, the error d [k] in the guide pitch k + 1 adjacent to the higher side is expected to increase as the guide pitch k becomes lower than the lowest sound that can be properly spoken. Therefore, the difference Δd [k] (= d [k + 1] −d [k]) between the error d [k] and the error d [k + 1] is naturally smaller than 0.

  Then, even if the target guide pitch k is limited to a difference Δd [k] smaller than 0, the guide pitch k to be extracted as the lowest sound klo is not excluded from the extraction target. The processing load can be reduced.

  Furthermore, since only the appropriate guide pitch k is the target, it is possible to prevent the determination accuracy of the lowest sound klo from being lowered due to the extraction of the inappropriate guide pitch k. be able to.

  Further, in each of the above configurations, the guide pitch k that becomes the lowest sound klo is in the inflection region on the lower sound side among all the guide pitches in the error distribution, so that the guide pitch that is referred to when determining the lowest sound klo. The range of k may be limited to a band that forms a low-frequency inflection region. As a configuration for this purpose, for example, it is conceivable to make each of the above configurations as shown in an eighth configuration (claim 8).

  In this configuration, a band composed of a predetermined number of guide pitches located on the lower pitch side among a plurality of guide pitches in the error distribution updated by the distribution updating means is extracted as a low frequency error inflection band. Low-frequency error inflection band extraction means. Then, the lowest sound determination means extracts a guide pitch that becomes the lowest sound klo from the guide pitches in the low frequency error inflection band extracted by the low frequency error inflection band extraction means.

  In this configuration, the processing load required for the extraction and determination can be reduced by limiting the range of the guide pitch k referred to when determining the lowest sound klo to the error inflection band on the low sound side in advance.

  Further, in this configuration, a specific configuration for extracting the high frequency band is not particularly limited, but for example, a ninth configuration (claim 9) described below may be used.

  In this configuration, the low frequency error inflection band extracting means is located on the low pitch side among a plurality of guide pitches in the error distribution updated by the distribution updating means, and is adjacent to each other. A bass band composed of guide pitches whose error change rate with respect to the guide pitch is equal to or greater than a certain level is extracted.

  In the band on the low frequency side in the error distribution, the error is expected to increase as the guide pitch k becomes lower than the pitch that can be properly spoken. The degree of change in error between adjacent guide pitches also increases.

  Therefore, in the above configuration, as a rate of change when extracting the band, by setting a value assumed as a rate of change in the pitch below the guide pitch that can be appropriately spoken by the user as a “constant rate of change”, An appropriate band can be extracted as the treble error inflection band.

  More specifically, this configuration may be a tenth configuration (claim 10) described below.

  In this configuration, the low-frequency error inflection band extracting unit predetermines a guide pitch located at least on the low pitch side among a plurality of guide pitches in the error distribution updated by the distribution updating unit. In addition to classifying into guide groups for each guide pitch, and averaging the guide pitch errors classified into the same group for each group, the average error in each group is adjacent to the treble side. For one or more groups including a group having a minimum adjacent error ratio excluding the average error in the group, a band made up of each guide pitch classified into these groups is extracted.

  With this configuration, the guide pitch error is averaged for each group composed of a predetermined number of guide pitches, and one or more groups in which the adjacent error ratio between the adjacent groups is minimum are determined in this group. A low-frequency error inflection region consisting of each guide pitch can be extracted.

  In this configuration, the “guide pitch located on the low pitch side” to be classified as a group only needs to be located on the low pitch side in the error distribution. It is conceivable to use a certain ratio (for example, several tens of percent) on the low sound side in the entire pitch range.

  Further, in each of the above configurations, the error of the singing pitch with respect to the guide pitch in the specific guide section may be any value, for example, the singing pitch transition pattern and the guide pitch transition pattern. Employing a quantified difference in shape as a pattern, a quantified difference in the pitch of the final pitch reached in the singing pitch transition pattern and the guide pitch transition pattern be able to.

  Moreover, when acquiring guide data in the said structure, what is necessary is just to acquire what corresponds to the music which the user sang in the song data out of several guide data prepared beforehand. Here, “corresponding to the song sung by the user” may be specified by associating the song data sung by the user with the song data, and specifying the song based on the correspondence.

  In the above configuration, when updating the error distribution corresponding to the corresponding user, the user who performed the singing is associated with the song data, and the error distribution of the user specified based on the correspondence relationship is updated. And it is sufficient.

  In addition, when acquiring singing data in the above configuration, it is only necessary to acquire from the outside of the system what was separately generated along with the singing of the music by the user, and it is generated each time the user sings the music. It is good also as acquiring this.

  In generating the singing data like the latter, it is sufficient that the user generates the singing data based on the voice data formed by singing the music. The configuration of A is good.

  In this configuration, for voice data formed by a user singing a song, the pitch of each position along the time axis in the voice data is calculated, and data indicating the transition of the pitch along the time axis is used as song data. Get as.

  If it is this structure, song data can be produced | generated from the audio | voice data which concern on the song of a music by a user, and a pitch can be determined based on this.

  By the way, the singing data described above shows the transition of the pitch along the time axis associated with the singing, but when viewed in each singing section, the timing at which the pitch change actually started is the song. May be out of sync with the timing when singing properly.

  In this case, if the error is simply calculated by comparing the transition patterns as described above, the error increases due to such a timing shift. However, in light of the objective of determining the user's pitch, it can be said that the transition pattern itself in which the pitch actually shifts is more important than such timing. It is desirable.

  As described above, in order to compensate for the timing shift, it is conceivable to make each of the above configurations as a Bth configuration shown below.

  In this configuration, for each singing section indicated by the singing data, the degree of approximation between the singing pitch transition pattern in the singing section and the guide pitch transition pattern in the corresponding guide section indicated by the guide data is The position on the time axis in the singing section is corrected so as to be maximum. Then, the error calculation means, each of the transition patterns of the singing pitch in each unit section indicated by the singing data corrected in this way, guide corresponding to the corresponding singing section in each guide section indicated by the guide data. By comparing with the transition pattern of the guide pitch in the section, the error of both transition patterns is calculated as the error of the singing pitch with respect to the guide pitch to be uttered in the guide section.

  If it is this structure, the position on the time axis in a song pitch will be corrected so that the transition pattern of a song pitch and the transition pattern of a guide pitch may be approximated most. When the transition pattern is approximated in this way, it means that the timing at which the pitch changes is also approximated. Thus, the timing deviation at which the pitch change is started can be compensated. .

  In this configuration, any method may be employed to correct the position on the time axis so that the transition pattern approximates. For example, as described in JP-A-2005-107330 It is conceivable to adopt such a method.

  In order to solve the above problem, an eleventh configuration (claim 11) is a sound range provided with all means of any one of the first to fifth configurations and any of the sixth to tenth configurations. A determination system, wherein a range of pitches from a highest sound determined by the highest sound determination means to a lowest sound determined by the lowest sound determination means corresponds to an error distribution referred to in the determination A sound range determination means for determining a sound range of

  With this configuration, as described above, the user's range is appropriately set based on the highest tone kup and the lowest tone klo in the user's range determined as the pitch excluding the pitch that is forcibly uttered. Can be determined.

  In order to solve the above problem, the configuration of claim 12 (claim 12) causes a computer to execute various processing procedures for causing all the means according to any one of the first to fifth configurations to function. It is a program.

  A computer controlled by this program can function as a part of any one of the first to fifth configurations.

  In order to solve the above problem, the configuration of claim 13 (claim 13) causes the computer to execute various processing procedures for causing all the means according to any of the sixth to twelfth configurations to function. It is a program.

  The computer controlled by this program can function as a part of any of the sixth to twelfth configurations.

  In order to solve the above problem, a configuration of claim 14 (claim 14) is a program for causing a computer to execute various processing procedures for causing all the means according to the 111th configuration to function.

  The computer controlled by this program can function as a part of the eleventh configuration.

  Note that the above-described program is composed of an ordered sequence of instructions suitable for processing by a computer system, and can be used for a pitch determination system via various recording media and communication lines, and for users who use these. It is provided.

Block diagram showing the overall configuration of the range determination system Flow chart showing error tabulation process The figure which shows the transition pattern of the pitch shown by song data and guide data The figure which shows the transition pattern of the sound waveform (a) shown by audio | voice data, and the pitch shown by song data (b) Diagram showing error distribution Flow chart showing the range determination process (1/2) Flow chart showing the range determination process (2/2)

Embodiments of the present invention will be described below with reference to the drawings.
(1) Hardware Configuration The sound range determination system 1 is realized by mounting a program on a terminal device or a karaoke device composed of a known computer system.

  First, as shown in FIG. 1A, the hardware configuration when the program is installed in the “terminal device” includes a control unit 11 that controls the entire system, a storage unit 13 that stores various information, and a network 2. A communication unit 15 for controlling the communication, a user interface (U / I) unit 17 including a keyboard and a display, a media drive 19 for inputting and outputting information via a recording medium, and the like.

  In this configuration, when a predetermined command is received from the outside via the user interface unit 17 or the communication unit 15, the control unit 11 executes various processes according to the program stored in the storage unit 13. Exhibits the function as a sound range judgment system.

  In addition, as shown in FIG. 1B, the hardware configuration when the program is installed in the “karaoke apparatus” includes a control unit 11 that controls the entire system, music data and video indicating the accompaniment content and lyrics of the performance music. A storage unit 13 for storing data, a communication unit 15 for controlling communication via the network 2, a display unit 21 for displaying various images, an operation unit 23 including a plurality of keys and switches, and a voice from a microphone 25 The audio input / output unit 29 that controls input and output of audio from the speaker 27 is provided.

  In this configuration, when receiving a predetermined command from the outside via the operation unit 23 or the communication unit 15, the control unit 11 executes various processes according to the program stored in the storage unit 13, thereby Demonstrates the function of a range determination system.

In the present embodiment, the sound range determination system 1 is constituted by a single device (terminal device, karaoke device), but it goes without saying that it can be constituted by a plurality of devices that operate in cooperation with each other.
(2) Processing by Control Unit 11 Hereinafter, various processing procedures executed by the control unit 11 according to a program stored in the storage unit 13 will be described.
(2-1) Error Aggregation Process First, the processing procedure of the error aggregation process will be described with reference to FIG.

  This error totaling process is performed when a predetermined operation is performed through the user interface unit 17 or from the network 2 through the communication unit 15 if the present sound range determination system 1 is configured by a terminal device. It is activated when a command is received. On the other hand, if the real-range determination system 1 is realized by a karaoke device, the user sings a song (reproduction of a song based on song data) while the operation mode of the karaoke device is switched to the range determination mode. It is activated every time.

  When this error counting process is activated, first, singing data indicating the transition of the pitch along the time axis associated with the singing of music by the user is acquired (s110). This singing data shows the transition of the pitch (hereinafter referred to as “singing pitch”) when the user sings the music along the time axis. Specifically, the singing data includes the voice related to the user's singing. A transition pattern in which the included fundamental frequency is shifted along the time axis is shown (see FIG. 3A).

  In this s110, if the real sound range determination system 1 is realized by a terminal device, the song data designated through the operation to the user interface unit 17 is transmitted from the storage unit 13 or the media drive 19 (that is, the recording medium). The singing data acquired or received via the communication unit 15 is acquired. In addition, the song data acquired in this way are added with user identification information and song identification information for identifying the user and song associated with the song, respectively.

  Moreover, if the real range determination system 1 is the structure implement | achieved by the karaoke apparatus, the song data produced | generated at the time of the song of a music by a user are acquired. In this case, the user's voice data (see FIG. 4A) input from the voice input / output unit 29 along with the singing of the music is acquired, and each position along the time axis in this voice data is acquired. The pitch is calculated, and data indicating the transition of the pitch along the time axis (see FIG. 4B) is generated as song data. In addition, the song data acquired in this way is added with user identification information for identifying a user who is logged in to the karaoke apparatus at the time of singing, user identification information for identifying a song related to the song, and song identification information. It has been made.

  Next, for the music related to the singing in the singing data acquired in s110, guide data indicating the transition of the pitch along the time axis when the music is appropriately sung is acquired (s120). This guide data shows the transition of the pitch along the time axis when a song is properly sung. Specifically, the basic frequency that should be included in the sound when singing properly is the time. FIG. 3A shows a transition pattern that is shifted along the axis (see FIG. 3A).

  In this embodiment, for each of a plurality of music pieces, a plurality of guide data indicating the transition of the pitch along the time axis when the music is appropriately sung is stored in the storage unit 13, and in this s120, Among these guide data, guide data corresponding to the music identified by the music identification information added to the singing data acquired in s110 is read out and acquired. The guide data may be obtained by receiving provision from a server device or the like connected via the network 2.

  Next, the position of each unit section (hereinafter referred to as “singing section”) along the time axis in the transition pattern indicated by the singing data acquired in s110 is based on the guide data acquired in s120. Is corrected (s130).

  Here, for each singing section indicated by the singing data, the singing pitch transition pattern in this singing section is the singing of each unit section (hereinafter referred to as “guide section”) on the time axis indicated by the guide data. The position on the time axis in the singing section is corrected (time axis) so that it closely approximates the transition pattern of the pitch in the guide section corresponding to the section (hereinafter referred to as “guide pitch”). (Refer to FIG. 3B).

  As a method for correcting the position on the time axis so that the transition pattern is approximated in this way, any method may be adopted. For example, as described in JP-A-2005-107330 It is conceivable to adopt a method.

  Each of the above-described singing section and guide section indicates each section obtained by dividing the whole of the music along the time axis into two or more, but one singing section as a whole without dividing the entire music, It may be handled as a guide section.

  Next, a transition pattern of the singing pitch indicated by the singing data whose position on the time axis is corrected in s130, and a transition pattern of the guide pitch indicated by the guide data acquired in s120. By contrast, the error of the singing pitch with respect to the guide pitch is calculated (s140).

  Here, each of the transition patterns of the singing pitch in the singing section indicated by the singing data is compared with the transition pattern in the guide section corresponding to the corresponding singing section among the guide sections indicated by the guide data, and the utterance in the guide section Each of the errors d [k] of the singing pitch with respect to the guide pitch k (= 1 to n) to be calculated is calculated.

  “Error” here means, for example, a numerical difference between the pattern of the transition pattern of the singing pitch and the transition pattern of the guide pitch, or the transition pattern of the singing pitch and the guide pitch. For example, the difference (or the difference in the period during which the singing pitch is the same pitch as the guide pitch) in the transition pattern is quantified.

  Next, it is information prepared for each user, and among the error distribution information indicating the error distribution obtained by distributing the error of the singing pitch with respect to the guide pitch for each guide pitch, the singing acquired in s110 above. Error distribution information corresponding to the user related to the song in the data is acquired (s150). Here, based on the user identification information added to the singing data, error distribution information corresponding to the user identified thereby is acquired.

  As shown in FIG. 5A, the “error distribution” indicated by the error distribution information takes the accumulated value of errors on the vertical axis and the guide pitch k defined by the fundamental frequency on the horizontal axis. The error of the singing pitch with respect to the guide pitch k is distributed. FIG. 5 shows an error distribution by an envelope connecting the accumulated error values for each guide pitch k.

  Next, the miscalculation distribution information acquired in s150 is updated so as to indicate that each error calculated in s140 is additionally distributed in the error distribution (s160). Here, the errors for each guide pitch calculated in s140 are integrated as errors of the corresponding guide pitch in the error distribution indicated by the miscalculation distribution information acquired in s150, and thus integrated. The error distribution information indicating the error distribution is updated.

  Next, based on the error distribution information updated in s160, the user's range corresponding to the error distribution indicated by the error distribution information is determined (s170). Here, the range determination process described later is performed using the error distribution information updated in s160 as an argument, and the range of the user is determined here.

Then, the user's range determined in s180 is notified or stored (s180). Here, if the operation setting of the apparatus is set to notify the user's sound range determined in s170, the sound range is displayed on the user interface unit 17 or the display unit 21. Alternatively, the data is transmitted to an external device via the communication unit 15. Also, if the setting is such that the user's sound range should be accumulated, information indicating the sound range is stored in the storage unit 13 or the recording medium.
(2-2) Sound Range Determination Process Next, the processing procedure of the sound range determination process which is s170 of the error totaling process will be described based on FIG.

  In this sound range determination process, first, the error distribution indicated by the error distribution information passed from the error totaling process is classified into a group for each predetermined number of guide pitches (s210). Here, as shown in FIG. 5B, the horizontal axis of the error distribution is divided for each band corresponding to a predetermined number of guide pitches k, and the guide pitches in the same band thus divided are the same. It is classified as a group m (m is an index indicating a group).

  Next, for each group classified in s210, an average error d [m] obtained by averaging the error d [k] of the guide pitch k classified in the same group is calculated (s220). Here, as shown in FIG. 5C, the error d [k] of all the guide pitches k in the error distribution is set as an average error d [m] averaged for each group m.

  Next, for each of the groups classified in s210, an error change rate with the guide pitch adjacent to the group m is calculated (s230). In the present embodiment, the adjacent error ratio obtained by dividing the average error of the group adjacent to the group m by the average error d [m] of the group m is calculated as a parameter representing the change rate.

  More specifically, as shown in FIG. 5D, the average error d [m−1] of the group m−1 adjacent to the corresponding group m on the bass side is changed to the average error d [m] of the corresponding group m. The value d [m−1] / d [m] divided by is calculated, and this is the adjacent error ratio of the corresponding group m in the error distribution. The adjacent error ratio becomes a value farther from “1” as the change rate increases. Here, the adjacent error ratio is calculated based on the relationship between the corresponding group m and the group m−1 adjacent to the bass side, but the adjacent error ratio is calculated based on the relationship between the group m + 1 adjacent to the high frequency side. It is good.

  Next, among the groups classified in s210, a band composed of one or more groups including a group in which the rate of change calculated in s230 is equal to or higher than a certain level is the highest tone in the range in the subsequent processing. Is extracted as a target band for specifying (S240). In the present embodiment, as shown in FIG. 5E, the adjacent error ratio calculated as the rate of change is the maximum (the furthest away from “1” in the positive direction) on the treble side. , And a group m−1 corresponding to the average error d [m−1] referred to when calculating the adjacent error ratio, and a high frequency band (high frequency error inflection band), It is extracted as a target band (see the shaded part A in the figure).

  Here, on the assumption that the adjacent error ratio is calculated based on the relationship between the group m and the group m-1, the group m and the group m-1 having the maximum adjacent error ratio are configured. Bands are extracted. However, when the adjacent error ratio is calculated based on the relationship between the group m and the group m + 1, the band formed by the group m and the group m + 1 having the maximum adjacent error ratio is changed to the high frequency side error variable. It may be extracted as a music band.

  Thus, after the target band is extracted in s240, the difference Δd between the guide pitches of the group constituting the target band and the error in the adjacent guide pitch is calculated (s310).

  In the present embodiment, for each guide pitch k, first, an error d [k + 1] in the guide pitch k + 1 adjacent to the high pitch side of the corresponding guide pitch k, an error d [k] in the corresponding guide pitch k, The difference Δd [k] (= d [k + 1] −d [k]) is calculated. Then, the difference Δd [k] (=) between the error d [k] at the corresponding guide pitch k and the error d [k−1] at the guide pitch k−1 adjacent to the low pitch side of the corresponding guide pitch k. d [k] -d [k-1]) is calculated.

  Next, for each guide pitch of the group that constitutes the target band, an error ratio obtained by comparing the difference Δd [k−1] and the difference Δd [k] calculated in s310 with respect to the corresponding guide pitch k | Δd [k−1] / Δd [k] | is calculated (s320).

  Next, among the guide pitches of the groups constituting the target band, the difference Δd [k] calculated in s310 is greater than 0, and the error ratio | Δd [k−1] calculated in s320 is set. ] / Δd [k] | is extracted, and this is determined as the highest sound kup in the user's range (s330).

  The error distribution indicated by the error distribution information is obtained by distributing the error d [k] as a transition pattern of the singing pitch with respect to the guide pitch k for each guide pitch. Therefore, the guide having a small error d [k]. The pitch k can be said to be a pitch that the corresponding user can properly utter in a similar transition pattern, whereas a guide pitch whose error d [k] is large is uttered by the corresponding user forcibly. It can be said that the pitch.

  Then, in the high-side error inflection band from the guide pitch that the user can properly utter to the guide pitch that is forcibly uttered, the highest guide pitch k and low-side that can be properly uttered The error d [k] at the guide pitch k + 1 adjacent to the high pitch side tends to increase rapidly with respect to the error d [k-1] at the guide pitch k-1 adjacent to (see FIG. 7A).

  In this case, the difference Δd [k] between the error d [k] for the guide pitch k and the error d [k + 1] for the guide pitch k + 1 adjacent to the high pitch side is naturally greater than zero. Further, the error ratio | Δd [k−1] / Δd [k] |, which is a ratio with the guide pitch k−1 adjacent to the low tone side, shows a constant small value. More specifically, the error ratio is such that Δd [k], which is the denominator, becomes extremely large with the guide pitch k of the highest sound that the user can properly utter in the error inflection band as the inflection point. Indicates the minimum value.

  Thus, the guide pitch k where the difference Δd [k] is greater than 0 and the error ratio | Δd [k−1] / Δd [k] | is minimum can be determined as the highest tone kup (FIG. 7 ( (See shaded part A in b)).

  Next, among the groups classified in s210, a band composed of one or more groups whose rate of change calculated in s230 is greater than or equal to a certain value specifies the lowest sound in the range in the subsequent processing. Is extracted as a target band at the time (s250). In the present embodiment, as shown in FIG. 5E, the adjacent error ratio calculated as the rate of change is the smallest (the farthest away from “1” in the negative direction) on the bass side, and the group m , And a group m−1 corresponding to the average error d [m−1] referred to when calculating the adjacent error ratio, and a low frequency band (low frequency error inflection band), It is extracted as a target band (see the shaded part B in the figure).

  Here, on the assumption that the adjacent error ratio is calculated based on the relationship between the group m and the group m-1, the group m and the group m-1 having the minimum adjacent error ratio are configured. Bands are extracted. However, when the adjacent error ratio is calculated based on the relationship between the group m and the group m + 1, the band composed of the group m and the group m + 1 having the minimum adjacent error ratio is set to the minimum side error inflection. It is good also as extracting as a zone | band.

  Thus, after the target band is extracted in s250, the difference Δd between the guide pitches of the groups constituting the target band and the error in the adjacent guide pitches is calculated (s410). In the present embodiment, the difference Δd [k] (= d [k + 1] −d [k]) and the difference Δd [k] (= d [k] −d [k−1]) are the same as in 310 above. Calculated.

  Next, for each of the guide pitches of the group constituting the target band, an error ratio obtained by comparing the difference Δd [k] and the difference Δd [k−1] calculated in s310 with respect to the corresponding guide pitch k | Δd [k] / Δd [k−1] | is calculated (s420).

  Next, among the guide pitches of the groups constituting the target band, the difference Δd [k] calculated in s410 is smaller than 0, and the error ratio | Δd [k−1] calculated in s420 is set. ] / Δd [k] | is extracted, and the guide pitch k is extracted, and this is determined as the lowest sound klo in the user's range (s430).

  As described above, the guide pitch k having a small error d [k] can be said to be a pitch at which the corresponding user can properly utter in a similar transition pattern, whereas the guide pitch having a large error d [k]. High can be said to be the pitch that the user is uttering forcibly, so in the error inflection band on the low tone side, the guide pitch k + 1 adjacent to the high tone side is higher than the lowest guide pitch k that can be properly uttered. The error d [k] at the guide pitch k tends to increase rapidly with respect to the error d [k + 1] at (see FIG. 7A).

  In this case, the difference Δd [k] between the error d [k] for the guide pitch k and the error d [k + 1] for the guide pitch k + 1 adjacent to the high pitch side is naturally smaller than zero. Further, the error ratio | Δd [k−1] / Δd [k] |, which is a ratio with the guide pitch k−1 adjacent to the low tone side, shows a constant small value. Specifically, this error ratio has a significantly large Δd [k−1] as a denominator with the guide pitch k of the highest sound that the user can properly utter in the error inflection band as an inflection point. As a result, the minimum value is indicated.

  In this way, the guide pitch k in which the difference Δd [k] is smaller than 0 and the error ratio | Δd [k−1] / Δd [k] | is minimum can be determined as the highest tone klo (FIG. 7 ( (See shaded portion B in b)).

  Then, the range of the pitch from the highest sound kup determined at s330 to the lowest sound klo determined at s430 is determined as the user's range (FIG. 7 (b) (s260).

Thus, after finishing s260, the process returns to the error totaling process (s180).
(3) Operation, effect In the range determination system 1 according to the present embodiment, based on the highest sound kup and the lowest sound klo in the user's range determined as the pitch excluding the pitch that is forcibly uttered, The user's range can be appropriately determined (FIG. S260).

  In determining the highest sound kup and the lowest sound klo, first, the transition pattern of the pitch (singing pitch) along the time axis associated with the user's singing is set to the pitch (guide sound) when the song is properly sung. By comparing with the high transition pattern, the error d of these transition patterns is calculated (s140 in FIG. 6), and this is reflected in the error distribution of the user (s160 in FIG. 6).

  Then, of the guide pitches 1 to n in the error distribution, the guide that minimizes the error ratio | Δd [k−1] / Δd [k] | that compares each difference Δd with the error d in the adjacent guide pitch. The pitch k is determined as the highest sound kup in the user's range corresponding to the error distribution (s310 to s330 in the figure). Similarly, the guide pitch k that minimizes the error ratio | Δd [k] / Δd [k−1] | that compares each difference with the error in the adjacent guide pitch is determined by the user corresponding to the error component. The lowest sound klo in the range is determined (s410 to s430 in the figure).

  Thus, in the above embodiment, the guide pitch k at which the error ratio | Δd [k−1] / Δd [k] | is minimum can be determined as the highest sound kup, and the error ratio The guide pitch k at which | Δd [k] / Δd [k−1] | is minimized can be determined as the highest tone klo.

  Further, in the above embodiment, the range of the guide pitch k that is referred to when determining the highest sound kup is limited to a range in which the difference Δd [k] is greater than 0 (s330 in FIG. 6).

  In the inflection region described above, it is expected that the error d [k] in the guide pitch k + 1 adjacent to the higher side will increase as the guide pitch K becomes higher than the highest sound that can be properly spoken. Therefore, the difference Δd [k] between the error d [k] and the error d [k + 1] is naturally larger than zero.

  Then, even if the target guide pitch k is limited to a difference Δd [k] larger than 0, the guide pitch k to be extracted as the highest sound kup is not excluded from the extraction target. The processing load can be reduced. Furthermore, since only the appropriate guide pitch k is targeted, it is possible to prevent the determination accuracy of the highest sound kup from being lowered due to extraction of an inappropriate guide pitch k. be able to.

  Further, in the above-described embodiment, the range of the guide pitch k referred to when determining the lowest sound klo is limited to a range in which the difference Δd [k] is smaller than 0 (s430 in FIG. 6).

  In the inflection region described above, it is expected that the error d [k] in the guide pitch k + 1 adjacent to the higher side will decrease as the guide pitch k becomes higher than the lowest sound that can be properly spoken. Therefore, the difference Δd [k] between the error d [k] and the error d [k + 1] is naturally smaller than 0.

  Then, even if the target guide pitch k is limited to a difference Δd [k] smaller than 0, the guide pitch k to be extracted as the lowest sound klo is not excluded from the extraction target. The processing load can be reduced. Furthermore, since only the appropriate guide pitch k is the target, it is possible to prevent the determination accuracy of the lowest sound klo from being lowered due to the extraction of the inappropriate guide pitch k. be able to.

  Further, in the above embodiment, after extracting a band composed of a predetermined number of guide pitches located on the higher pitch side of the guide pitch k in the error distribution as an error inflection band on the higher pitch side, The determination of kup is performed (s240 to s330). The guide pitch k that becomes the highest sound kup is in the inflection region on the high sound side of all the guide sound pitches in the error distribution, so the range of the guide pitch k that is referred to when determining the highest sound kup is the high sound side. There is no problem even if it is limited to the band that forms the inflection region, so that the processing load required for the determination of the highest sound kup can be suppressed.

  Further, in the above embodiment, a band composed of a predetermined number of guide pitches located on the lower pitch side of the guide pitch k in the error distribution is extracted as an error inflection band on the lower pitch side, and then the lowest pitch is calculated. klo is determined (s250 to s430). The guide pitch k that is the lowest pitch klo is in the inflection region on the low pitch side among all the guide pitches in the error distribution, so the range of the guide pitch k that is referred to when determining the minimum pitch klo is There is no problem even if it is limited to the band that forms the inflection region, so that the processing load required for determining the lowest sound klo can be suppressed.

  In the above embodiment, the guide pitch of the error distribution is located on the higher pitch side, and the rate of change in error between the adjacent guide pitches is more than a certain value. A treble error inflection band consisting of each guide pitch is extracted (s240 in FIG. 6).

  In the band on the high pitch side in the error distribution, the error is expected to increase as the guide pitch k becomes higher than the pitch that can be properly spoken. The degree of change in error between adjacent guide pitches also increases.

  Therefore, as described above, the rate of change at the time of extracting a band is a value assumed as the rate of change in the pitch above the guide pitch that can be appropriately spoken by the user (in this embodiment, “the adjacent error ratio is the maximum). By setting “a constant value” as “a constant change rate”, it is possible to extract an appropriate band as the treble error inflection band.

  Moreover, in the said embodiment, it is located in the low pitch side among the guide pitches of an error distribution, and the rate of change of the error between each adjacent guide pitch is not less than a certain level. A bass error inflection band consisting of each guide pitch is extracted (s250 in FIG. 6).

  In the band on the low frequency side in the error distribution, the error is expected to increase as the guide pitch k becomes lower than the pitch that can be properly spoken. The degree of change in error between adjacent guide pitches also increases.

  Therefore, as described above, the rate of change at the time of extracting the band is a value assumed as the rate of change in the pitch below the guide pitch that can be appropriately spoken by the user (in this embodiment, “the adjacent error ratio is the minimum By setting “a constant value” to “a constant change rate”, it is possible to extract an appropriate band as the bass error inflection band.

  In the above embodiment, the errors in the guide pitches for each group consisting of a predetermined number of guide pitches are averaged (s220 in FIG. 6), and the adjacent error ratio between the adjacent groups is maximized. With respect to the above group, it is possible to extract a treble error inflection region made up of each guide pitch in this group (s240 in the figure). Similarly, for one or more groups having a minimum adjacent error ratio, it is possible to extract a bass error inflection region made up of each guide pitch in this group.

Further, in the above embodiment, the pitch error is calculated after the position on the time axis of the singing pitch is corrected so that the transition pattern of the singing pitch and the transition pattern of the guide pitch are most approximated. (S140 in FIG. 2). When the transition pattern is approximated in this way, it means that the timing at which the pitch changes is also approximated. Thus, the timing deviation at which the pitch change is started can be compensated. .
(4) Modifications The embodiment of the present invention has been described above. However, the present invention is not limited to the above embodiment, and can take various forms as long as they belong to the technical scope of the present invention. Needless to say.

  For example, in the above-described embodiment, the range of the guide pitch k referred to when determining the lowest sound kup is limited to a range where the difference Δd [k] is larger than 0 (s330 in FIG. 6). However, the guide pitch k referred to here may be a guide pitch on the high pitch side in the error distribution, and may be determined regardless of the difference Δd [k]. As a specific example, it is conceivable to set a certain ratio (for example, several tens of percent) on the high pitch side in the entire guide pitch range.

Further, in the above embodiment, the range of the guide pitch k referred to when determining the lowest sound klo is limited to a range where the difference Δd [k] is smaller than 0 (s430 in FIG. 6). However, the guide pitch k referred to here may be a guide pitch on the low tone side in the error distribution, and may be determined regardless of the difference Δd [k]. As a specific example, it is conceivable to set a certain ratio (for example, several tens of percent) on the low sound side in the entire guide pitch.
(5) Correspondence with the Present Invention In the embodiment described above, s140 in FIG. 2 is the error calculating means in the present invention, s160 in FIG. 6 is the distribution updating means in the present invention, and s330 in FIG. 6 is the present invention. S240 is the high frequency error inflection band extracting means in the present invention, s430 is the lowest sound determining means in the present invention, and s250 is the low frequency error in the present invention. The inflection band extraction means, and s260 in FIG.

  DESCRIPTION OF SYMBOLS 1 ... Sound range determination system, 2 ... Network, 11 ... Control part, 13 ... Memory | storage part, 15 ... Communication part, 17 ... User interface part, 19 ... Media drive, 21 ... Display part, 23 ... Operation part, 25 ... Microphone, 27 ... Speaker, 29 ... Audio input / output unit.

Claims (14)

Singing data indicating the transition of the pitch along the time axis associated with the singing of the music by the user, and guide data indicating the transition of the pitch along the time axis when the singing music related to the singing data is appropriately sung , The transition patterns of the pitches (hereinafter referred to as “singing pitches”) in the respective unit sections (hereinafter referred to as “singing intervals”) on the time axis indicated by the song data are indicated by the guide data. Both transition patterns are compared with the transition pattern of the pitch (hereinafter referred to as “guide pitch”) in the guide section corresponding to the corresponding singing section in each unit section (hereinafter referred to as “guide section”) on the time axis. Error calculating means for calculating an error Δd [k] of a singing pitch with respect to a guide pitch k (= 1 to n) to be uttered in the guide section And,
Among the error distributions prepared for each of a plurality of users and distributing the error of the singing pitch with respect to the guide pitch of the user for each of the guide pitches, this is an opportunity for calculation by the error calculation means. Each of the calculated errors is additionally distributed in the error distribution corresponding to the user who performed the song as an error d [k] of the singing pitch with respect to the guide pitch k in the guide section referred to in the calculation. A distribution updating means for updating the error distribution;
Among the guide pitches k in the error distribution updated by the distribution updating means, the difference Δd [k−1] (= d [) from the error d [k−1] in the guide pitch k−1 adjacent to the lower side. k] −d [k−1]) is compared with the difference Δd [k] (= d [k + 1] −d [k]) from the error d [k + 1] at the guide pitch k + 1 adjacent to the higher side. A guide pitch k that minimizes | Δd [k−1] / Δd [k] | is extracted, and the extracted guide pitch k is determined as the highest tone kup in the user's range corresponding to the error component. A pitch determination system characterized by comprising the highest pitch determination means. The highest sound determination means has a difference Δd [k] greater than 0 in the guide pitch k in the error distribution updated by the distribution update means, and the error ratio | Δd [k−1] / Δd. The pitch determination system according to claim 1, wherein a guide pitch that minimizes [k] | is extracted. Among the plurality of guide pitches in the error distribution updated by the distribution updating means, a high frequency error variable is extracted that extracts a band consisting of a predetermined number of guide pitches located on the higher pitch side as a high frequency error inflection band. Music band extraction means,
The highest sound determination means extracts a guide pitch that becomes the highest sound kup from guide pitches in the high frequency error inflection band extracted by the high frequency error inflection band extraction means. The pitch determination system according to claim 1 or 2. The high frequency error inflection band extracting means is located on the higher pitch side among a plurality of guide pitches in the error distribution updated by the distribution updating means, and is adjacent to the adjacent guide pitches. The pitch determination system according to claim 3, wherein a treble error inflection band made up of guide pitches each having a rate of change in error between or above a certain value is extracted. The high frequency error inflection band extracting means converts a guide pitch located at least on the higher pitch side among a plurality of guide pitches in the error distribution updated by the distribution updating means to a predetermined number of guide sounds. In addition to classifying each group into high-frequency groups and averaging the guide pitch errors classified into the same group for each group, the average error in each group adjacent to the high-pitched sound side is the average error in that group. The pitch determination system according to claim 4, wherein, for one or more groups including a group having a maximum adjacent error ratio excluding, a band composed of each of the guide pitches classified into these groups is extracted. . Singing data indicating the transition of the pitch along the time axis associated with the singing of the music by the user, and guide data indicating the transition of the pitch along the time axis when the singing music related to the singing data is appropriately sung , The transition patterns of the pitches (hereinafter referred to as “singing pitches”) in the respective unit sections (hereinafter referred to as “singing intervals”) on the time axis indicated by the song data are indicated by the guide data. Both transition patterns are compared with the transition pattern of the pitch (hereinafter referred to as “guide pitch”) in the guide section corresponding to the corresponding singing section in each unit section (hereinafter referred to as “guide section”) on the time axis. Error calculating means for calculating an error Δd [k] of a singing pitch with respect to a guide pitch k (= 1 to n) to be uttered in the guide section And,
Among the error distributions prepared for each of a plurality of users and distributing the error of the singing pitch with respect to the guide pitch of the user for each of the guide pitches, this is an opportunity for calculation by the error calculation means. Each of the calculated errors is additionally distributed in the error distribution corresponding to the user who performed the song as an error d [k] of the singing pitch with respect to the guide pitch k in the guide section referred to in the calculation. A distribution updating means for updating the error distribution;
Among the guide pitches k in the error distribution updated by the distribution updating means, the difference Δd [k] (= d [k + 1] −d [k] from the error d [k + 1] in the guide pitch k + 1 adjacent to the higher side. ]) In comparison with the difference Δd [k−1] (= d [k] −d [k−1]) from the error d [k−1] in the guide pitch k−1 adjacent to the lower side. A guide pitch k that minimizes | Δd [k] / Δd [k−1] | is extracted, and the extracted guide pitch k is determined as the lowest tone klo in the user's range corresponding to the error component. A pitch determination system comprising: a minimum pitch determination unit; The lowest sound determination means has the difference Δd [k] smaller than 0 in the guide pitch k in the error distribution updated by the distribution update means, and an error ratio | Δd [k] / Δd [k− 1] The guide pitch k that minimizes | is extracted. The pitch determination system according to claim 6, wherein: Of the plurality of guide pitches in the error distribution updated by the distribution updating means, a low-frequency error variable that extracts a band consisting of a predetermined number of guide pitches located on the low pitch side as low-frequency error inflection bands. Music band extraction means,
The lowest sound determination means extracts a guide pitch that becomes the lowest sound klo from guide pitches in the low-frequency error inflection band extracted by the low-frequency error inflection band extraction means. The pitch determination system according to claim 6 or 7. The low frequency error inflection band extracting means is located on the low pitch side among a plurality of guide pitches in the error distribution updated by the distribution updating means, and is adjacent to the adjacent guide pitches. The pitch determination system according to claim 8, wherein a bass band composed of guide pitches each having a change rate of an error between the guide pitches is extracted from a certain level. The low-frequency error inflection band extracting unit converts a guide pitch located at least on the low pitch side among a plurality of guide pitches in the error distribution updated by the distribution updating unit to a predetermined number of guide sounds. In addition to classifying each group into high-frequency groups and averaging the guide pitch errors classified into the same group for each group, the average error in each group adjacent to the high-pitched sound side is the average error in that group. The pitch determination system according to claim 9, wherein, for one or more groups including a group having a minimum adjacent error ratio excluding, a band composed of each of the guide pitches classified into these groups is extracted. . A sound range determination system provided with all the means according to any one of claims 1 to 5 and all the means according to any one of claims 6 to 10,
A sound range in which a pitch range from the highest sound determined by the highest sound determination means to the lowest sound determined by the lowest sound determination means is a user's sound range corresponding to the error distribution referred to in these determinations A sound range determination system comprising a determination means.   A program for causing a computer to execute various processing procedures for causing the computer to function as all the means according to claim 1.   The program for making a computer perform the various process procedures for functioning as all the means in any one of Claim 6 to 10.   The program for making a computer perform the various process procedures for functioning as all the means in any one of Claim 11.