JP3966074B2 - Pitch conversion device, pitch conversion method and program - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a pitch conversion device, a pitch conversion method and a program suitable for use in singing synthesis.
[0002]
[Prior art]
Conventionally, as a speech synthesizer, a device that gives fluctuations to the pitch of synthesized speech is known (see, for example, JP-A-9-281994).
[0003]
In this prior art, as a clock signal to be supplied to a D / A converter that converts digital audio waveform data into an analog audio signal, a clock signal having a fluctuation in the clock cycle in accordance with the clock interval fluctuation data read from the storage unit. By using it, fluctuations are given to the pitch of the D / A conversion output (analog audio signal).
[0004]
[Problems to be solved by the invention]
It is difficult for a person who is a professor to sing at a physically constant height (pitch) when generating a sound corresponding to a certain note. Generally speaking, the utterance pitch is slightly different from the note pitch. In addition, pitch variation with time also occurs. In particular, when a general person who does not sing as a profession sings, the tendency of pitch shift and pitch fluctuation as described above is strong, and this is one element for evaluating the skill (or skill) of singing. . In addition, the characteristics of the singer may be seen in the pitch deviation. In addition, if you try to generate a sound with a pitch close to the upper limit or lower limit that a person can utter, the voice generation mechanism is physically burdened, so the pitch you want to generate differs from the pitch you actually uttered (near the upper limit). The pitch tends to decrease with high pitches, and the pitch tends to increase with low pitches near the lower limit.
[0005]
According to the above-described prior art, the direction and amount of pitch fluctuation can be changed by changing the value of the clock interval fluctuation data in the pitch increasing direction or the pitch decreasing direction. It is not possible to change the pitch (reference pitch) before adding, and to change the temporal pattern of pitch fluctuation. In other words, it is impossible to reproduce the vocal pitch of the singer as described above and the pitch variation over time.
[0006]
An object of the present invention is to provide a novel pitch conversion device, a pitch conversion method, and a program that can reproduce a vocal pitch of a singer and a pitch variation with time during singing synthesis.
[0007]
[Means for Solving the Problems]
A first pitch conversion apparatus according to the present invention is a pitch conversion apparatus used in a singing voice synthesizing apparatus including a singing voice synthesizing unit that synthesizes a singing voice signal having a pitch indicated by pitch data, and is a sequential pitch to be synthesized. Input means for sequentially inputting the pitch corresponding to each singing voice, and for converting a plurality of input pitches into a plurality of voice pitches, respectively.An input pitch function that is higher than the input pitch when the input pitch is lower than the predetermined lower limit pitch, and lower than the input pitch when the input pitch is higher than the predetermined upper limit pitch. Storage means for storing a pitch conversion function for converting the pitch to be equal to the input pitch when the pitch is between a predetermined lower limit pitch and a predetermined upper limit pitchAnd the pitch for each pitch input from the input meansPitch conversion functionConversion means for converting the voice pitch into a voice pitch and supplying the data indicating the voice pitch as the pitch data to the singing voice synthesis means.
[0008]
According to the first pitch conversion device, a plurality of input pitches are converted into a plurality of audio pitches, respectively.Pitch conversion functionIs stored in the storage means, thisPitch conversion functionBased on, each pitch related to the input is converted into a voice pitch for singing voice synthesis.Pitch conversion functionIn this case, if a plurality of vocal pitches of the singer are used as a plurality of voice pitches, the vocal pitch and pitch characteristics of the singer can be reproduced in the synthesized singing voice. The pitch can be increased slightly in the vicinity of the lower limit pitch that can be uttered while lowering.
[0009]
In the first pitch conversion device, the input means inputs singer data indicating a singer, and the storage meansPitch conversion functionIs stored for each singer, and the conversion means corresponds to the singer indicated by the singer data.Pitch conversion functionPitch conversion may be performed based on the above. In this way, the utterance pitch and pitch characteristics can be reproduced for each singer.
[0010]
In the first pitch conversion device, a random (random) pitch variation depending on the input pitch may be added to the voice pitch during pitch conversion. If it does in this way, a more natural pitch change can be given to synthetic singing voice. Moreover, you may make it provide the audio | voice pitch with the time-dependent pitch fluctuation | variation contained in a singer's actual audio | voice in the case of pitch conversion. In this way, it is possible to reproduce unstable pitch fluctuations over time of the singer.
[0014]
A first pitch conversion method according to the present invention is for converting a plurality of input pitches into a plurality of audio pitches.An input pitch function that is higher than the input pitch when the input pitch is lower than the predetermined lower limit pitch, and lower than the input pitch when the input pitch is higher than the predetermined upper limit pitch. Storage means for storing a pitch conversion function for converting the pitch to be equal to the input pitch when the pitch is between a predetermined lower limit pitch and a predetermined upper limit pitchAnd a pitch conversion method used in a singing voice synthesizing device that synthesizes a singing voice signal having a pitch indicated by pitch data, and sequentially corresponding to each of the sequential singing voices to be synthesized. A step of inputting a pitch, and for each pitch input in this step, the pitch isPitch conversion functionAnd converting the voice pitch into data representing the voice pitch as the pitch data to the singing voice synthesizing means.
[0015]
According to the first pitch conversion method, pitch conversion can be performed in the same manner as described above with respect to the first pitch conversion device.
[0018]
A first program according to the present invention is a program used in a singing voice synthesizing apparatus including a computer and a singing voice synthesizing unit that synthesizes a singing voice signal having a pitch indicated by pitch data. Input means for sequentially inputting pitches corresponding to sequential singing voices, and for converting a plurality of input pitches into a plurality of voice pitches, respectively.An input pitch function that is higher than the input pitch when the input pitch is lower than the predetermined lower limit pitch, and lower than the input pitch when the input pitch is higher than the predetermined upper limit pitch. Storage means for storing a pitch conversion function for converting the pitch to be equal to the input pitch when the pitch is between a predetermined lower limit pitch and a predetermined upper limit pitchAnd the pitch for each pitch input from the input meansPitch conversion functionIs converted to a voice pitch, and the data indicating the voice pitch is made to function as a conversion means for supplying the data as the pitch data to the singing voice synthesizing means.
[0019]
A second program according to the present invention is a program used in a singing voice synthesizing apparatus provided with a computer, wherein the computer sequentially inputs pitches corresponding to sequential singing voices to be synthesized. To convert multiple input pitches into multiple audio pitches.An input pitch function that is higher than the input pitch when the input pitch is lower than the predetermined lower limit pitch, and lower than the input pitch when the input pitch is higher than the predetermined upper limit pitch. Storage means for storing a pitch conversion function for converting the pitch to be equal to the input pitch when the pitch is between a predetermined lower limit pitch and a predetermined upper limit pitchAnd the pitch for each pitch input from the input meansPitch conversion functionIs converted into a voice pitch on the basis of the voice, and functions as a singing voice synthesizing means for synthesizing a singing voice signal having a voice pitch indicated by the pitch data sent from the converting means. Is.
[0020]
According to the first or second program, pitch conversion can be performed in the same manner as described above with respect to the first pitch conversion device.
[0021]
A third program according to the present invention is a program used in a singing voice synthesizing apparatus including a computer and a singing voice synthesizing unit that synthesizes a singing voice signal having a pitch indicated by pitch data.
Input means for sequentially inputting pitches corresponding to sequential singing voices to be synthesized;
Storage means for storing, for each input pitch among a plurality of input pitches, pitch difference data indicating a temporal variation of the voice pitch with respect to the input pitch;
For each pitch input from the input means, the pitch difference data corresponding to the pitch is read from the storage means, and the temporal variation of the voice pitch indicated by the pitch difference data for reading is added to the pitch according to the input. Then, the pitch conversion is performed, and the data indicating the pitch after the pitch conversion is made to function as a conversion means for supplying the data as the pitch data to the singing voice synthesizing means.
[0022]
A fourth program according to the present invention is a program used in a singing synthesizer provided with a computer,
Input means for sequentially inputting pitches corresponding to sequential singing voices to be synthesized;
Storage means for storing, for each input pitch among a plurality of input pitches, pitch difference data indicating a temporal variation of the voice pitch with respect to the input pitch;
For each pitch input from the input means, the pitch difference data corresponding to the pitch is read from the storage means, and the temporal variation of the voice pitch indicated by the pitch difference data for reading is added to the pitch according to the input. Converting means for performing pitch conversion and sending out pitch data indicating the pitch after the pitch conversion;
Singing voice synthesizing means for synthesizing a singing voice signal having a voice pitch indicated by pitch data sent from the converting means;
To function.
[0023]
According to the third or fourth program, pitch conversion can be performed in the same manner as described above with respect to the second pitch conversion device.
[0024]
DETAILED DESCRIPTION OF THE INVENTION
FIG. 1 shows a singing voice synthesizing apparatus according to an embodiment of the present invention.
[0025]
The song synthesizer in FIG. 1 includes an input unit 10, a pitch converter 12, and a song synthesizer 18, and the pitch converter 12 includes a pitch converter 14 and a database 16.
[0026]
The input unit 10 includes singer data indicating a singer, speech segment data indicating a speech segment (single phoneme [phoneme] or phoneme chain), note data indicating the pitch and length of a note, and synthesized speech sound. The sound intensity data indicating the intensity is input, and the note pitch data indicating the note pitch Pi related to the input and the singer data indicating the singer S related to the input are supplied to the pitch converter 14.
[0027]
The database 16 includes pitch conversion data for converting a plurality of input pitches (note pitches) into a plurality of voice pitches (output pitches) in the form of a pitch conversion function [FT (S, p)] or a pitch conversion table. It is remembered for each singer.
[0028]
FIG. 2 shows an example in which three pitch conversion functions FT (S1, p), FT (S2, p), and FT (S3, p) respectively corresponding to the singers S1, S2, and S3 are stored in the database 16. . Here, p represents the input pitch.
[0029]
In the pitch converter 12 shown in FIG. 2, the pitch converter 14 refers to the pitch conversion function corresponding to the singer S indicated by the singer data from the input unit 10 in the database 16, and notes from the input unit 10. The voice pitch Po corresponding to the note pitch Pi indicated by the pitch data is calculated based on the pitch conversion function related to the reference. Then, pitch data indicating the calculated voice pitch Po is output to the song synthesizer 18.
[0030]
When the database 16 stores the pitch conversion data in the form of a pitch conversion table, the pitch converter 14 refers to the pitch conversion table corresponding to the singer S from the input unit 10 and notes from the input unit 10. The voice pitch Po corresponding to the note pitch Pi indicated by the pitch data is read from the pitch conversion table according to the reference. Then, pitch data indicating the read voice pitch Po is supplied to the song synthesizer 18.
[0031]
The song synthesizer 18 synthesizes a song voice signal based on the singer data, voice segment data, note length data, and sound intensity data from the input unit 10 and the pitch data from the pitch converter 14. . Various song synthesis methods are known, and an appropriate one can be selected to configure the song synthesizer 18.
[0032]
As an example, the synthesizer 18 synthesizes a singing voice signal using voice component data corresponding to a singer indicated by the singer data and a voice element indicated by the voice element data. At this time, the pitch, the sound length, and the sound intensity of the singing voice signal are determined according to the pitch data, the note length data, and the sound intensity data, respectively.
[0033]
FIG. 3 shows an example of the pitch conversion function. In FIG. 3, the input pitch [cent] on the horizontal axis corresponds to the note pitch input to the pitch converter 14, and the output pitch [cent] on the vertical axis corresponds to the voice pitch output from the pitch converter 14. To do.
[0034]
The pitch conversion function FT (S, p) shown in FIG. 3 has an output pitch that is equal to the input pitch between the predetermined lower limit pitch PL and the predetermined upper limit pitch PH, but when the input pitch is higher than the upper limit pitch PH, The output pitch gradually becomes lower than the input pitch as it approaches the upper limit pitch that can be uttered, and when the input pitch is lower than the lower limit pitch PL, the output pitch gradually becomes higher than the input pitch as it approaches the lower limit pitch that humans can utter. It has a shape that When such a shape is expressed mathematically, it is as shown in the following equation (1).
[0035]
[Expression 1]
FT (S, p)> p if p <PL
FT (S, p) = p if PL ≦ p ≦ PH
FT (S, p) <p if PH <p
As a specific example, in the region where PH <p, the output pitch is about several tens of cents lower than the input pitch at the maximum, and in the region where p <PL, the output pitch is about several tens of cents higher than the input pitch. Functions can be used.
[0036]
The pitch conversion function as shown in FIG. 3 has a suitable shape for each singer and is stored in the database 16 for each singer as described above with reference to FIG. The pitch converter 14 converts the input pitch Pi into the output pitch Po with reference to the pitch conversion function corresponding to the singer S related to the input. When such pitch conversion is expressed mathematically, the following equation 2 is obtained.
[0037]
[Expression 2]
Po = FT (S, Pi)
FIG. 4 shows an example of pitch conversion using the pitch conversion function of FIG. 3, where (A) shows the pitch change before conversion (change in input pitch), and (B) shows the pitch after conversion. Indicates change (change in output pitch). In FIG. 4A, sequential input pitches correspond to sequential singing voices to be synthesized. According to FIG. 4, the output pitch is higher than the input pitch in the lower frequency range than PL, and the output pitch is lower than the input pitch in the higher frequency range than PH, and the output pitch is set to the input pitch in the middle frequency range higher than PL and lower than PH. You can see that they are equal. In the example of FIG. 4, the input pitch is given discretely. However, this need not be the case and may be given continuously.
[0038]
The pitch conversion function shown in FIG. 3 is approximated to a straight line, but the following number 3 is added to a random (random) pitch variation rand (S, p) depending on the singer and the pitch. A pitch conversion function as shown in the equation may be used.
[0039]
[Equation 3]
FT (S, p) + rand (S, p)
When such a pitch conversion function is used, a random pitch change is added to the sequential output pitch in response to the sequential input pitch as shown in FIG. Can be given a more natural change.
[0040]
In the above-described embodiment, an example in which the time-dependent pitch conversion function FT (S, p) is stored in the database 16 is shown. However, the database 16 stores a time-dependent pitch conversion function, You may make it perform pitch conversion with reference to a pitch conversion function. As an example, the database 16 stores pitch difference data indicating the pitch difference ΔFT (S, p, t) for each singer as pitch conversion data. The pitch difference ΔFT (S, p, t) represents the difference of the voice pitch with respect to the note pitch p as the singer S generates the voice corresponding to the note pitch p as time t progresses.
[0041]
When pitch difference data is stored in the database 16 for each singer in the form of a pitch conversion function ΔFT (S, p, t), the pitch converter 14 corresponds to the singer S related to the input. The input pitch Pi is converted to the output pitch Po with reference to ΔFT (S, p, t). When such pitch conversion is expressed mathematically, it is as shown in the following equation (4).
[0042]
[Expression 4]
Po = Pi + ΔFT (S, Pi, t)
The pitch conversion in this case is performed by adding a pitch difference ΔFT (S, Pi, t) corresponding to the input pitch Pi to the input pitch Pi.
[0043]
Instead of storing the pitch conversion function ΔFT (S, p, t) as described above, the database 16 stores pitch difference data representing a temporal change waveform of the pitch difference ΔFT (S, p, t). You may do it. FIG. 5 shows such pitch difference data as singer S.₁... An example in which 25 pitches of p1 to p25 are stored for each singer in Sn (n is an integer of 2 or more). The pitches p1 to p25 are 1200 to 3600 [cent] in steps of 100 cents (semitones). When pitch difference data is stored in the database 16, the amount of data can be reduced as compared with the case where pitch waveform data described later is stored.
[0044]
In the example of FIG. 5, it is preferable to use data based on an actual song as pitch difference data for each pitch. For example, Singer S₁A sound corresponding to the pitch p1 is actually generated at the same time, a time-dependent change waveform of the pitch difference of the generated sound with respect to the pitch p1 is obtained, and pitch difference data representing this change waveform is used. In this way, it is possible to reproduce the pitch change reflecting the characteristics of the singer and to express a more human fine pitch change.
[0045]
The pitch converter 14 refers to the pitch difference data corresponding to the input pitch Pi among the pitch difference data corresponding to the singer S related to the input, and converts the input pitch Pi into the output pitch Po according to the above-described equation (4). . FIG. 6 shows an example of pitch conversion using the pitch difference data of FIG. 5, and (A) shows the pitch change (change of input pitch) before conversion as in FIG. B) shows the pitch change (change in output pitch) after conversion. According to FIG. 6, in the vicinity of the upper limit pitch or lower limit pitch at which a person can speak, the output becomes lower or higher than the input pitch, respectively, as described above with reference to FIG. It can be seen that the amount of fluctuation (fluctuation) of the pitch increases in the vicinity. Therefore, it is possible to reproduce human voice pitch and pitch fluctuation.
[0046]
Since the pitch is not a discrete value but a continuous value, in order to be able to handle all pitches in the example of FIG. 5, pitch difference data for all the pitches is stored, and the amount of data to be stored is enormous. Become. Also, the amount of data to be stored for the pitch difference data in which the change in pitch difference ΔFT (S, p, t) continues for a long time becomes enormous. In order to avoid such an increase in the amount of stored data, the following measures (A) or (B) can be appropriately adopted.
[0047]
(A) When pitch difference data representing the pitch difference ΔFT (S, p, t) is stored for each of a plurality of discrete pitches, if pitch difference data whose pitch exactly matches the input pitch cannot be detected, input is performed. Pitch conversion is performed with reference to the pitch difference data having the closest pitch. Alternatively, pitch conversion may be performed by obtaining new pitch difference data by interpolation from two pitch difference data having a pitch close to the input pitch.
[0048]
(B) As the pitch difference data, data having a change duration of the pitch difference within a predetermined value is stored, and the time length of the input pitch exceeds the change duration of the pitch difference ΔFT (S, p, t). In this case, the pitch difference data is read again after returning to an appropriate position such as time 0 in the change waveform of the pitch difference.
[0049]
In the example of FIG. 5, the database 16 stores pitch difference data representing a change waveform of the pitch difference corresponding to each of a plurality of pitches for each singer, but corresponds to a plurality of pitches for each singer. You may make it memorize | store the pitch waveform data showing a pitch change waveform as pitch conversion data. In this case, pitch conversion is performed by reading the pitch waveform data corresponding to the input pitch Pi from among the plurality of pitch waveform data corresponding to the singer S related to the input, and setting it as the output pitch Po. When the pitch waveform data is created based on an actual song, it is possible to reproduce the vocal pitch of the singer and the pitch variation over time.
[0050]
The pitch conversion process as described above may be executed as a software process in a small computer such as a personal computer. That is, the CPU (central processing unit) may execute pitch conversion processing according to a program stored in storage means such as ROM or RAM.
[0051]
FIG. 7 shows a singing voice synthesizing apparatus according to another embodiment of the present invention. This apparatus performs singing synthesis using, for example, an SMS (Spectral Modeling Synthesis) technique disclosed in Japanese Patent No. 2906970. .
[0052]
In step S1, a singing voice signal is input, and in step S2, an SMS analysis process and a segment extraction process are performed on the input singing voice signal.
[0053]
In the SMS analysis process, the input speech signal is divided into a series of time frames, and one set of intensity (magnitude) spectrum data is generated by FFT (Fast Fourier Transform) for each frame, and one set of intensity for each frame. A line spectrum corresponding to a plurality of peaks is extracted from the spectrum data. Data representing the amplitude value and frequency of these line spectra is referred to as deterministic component data. Next, a residual spectrum is obtained by subtracting the spectrum of the harmonic component from the spectrum of the input speech waveform. This residual spectrum is referred to as an inharmonic component.
[0054]
In the segment extraction process, the harmonic component data and the anharmonic component data obtained by the SMS analysis process are classified according to the speech segment. A speech segment is a component of lyrics, for example, a single phoneme (or phoneme) such as [a], [i] or “a”. i "," a " It consists of a phoneme chain (a chain of multiple phonemes) like “p”.
[0055]
The database 20 stores harmonic component data and anharmonic component data for each speech unit. Further, the database 20 stores pitch conversion data (including the case of pitch difference data or pitch waveform data) as described above with respect to the database 16.
[0056]
When singing a song, lyrics data and melody data are input in step S3. In step S4, the phoneme sequence represented by the lyrics data is subjected to a phoneme sequence / speech unit conversion process to divide the phoneme sequence into speech units, and for each speech unit, harmonic component data and anharmonic component corresponding thereto. Are read out from the database 20 as speech segment data.
[0057]
In step S5, speech unit connection processing is performed on the speech unit data (harmonic component data and anharmonic component data) read from the database 20 to connect the speech unit data in the order of pronunciation.
[0058]
In step S6, a pitch conversion process is performed. That is, the note pitch indicated by the melody data input in step S3 is converted into a voice pitch based on the pitch conversion data (including the case of pitch difference data or pitch waveform data) in the database 20 in the same manner as described above. Pitch data indicating the voice pitch is generated.
[0059]
In step 7, new harmonic component data suitable for the voice pitch is generated on the basis of the harmonic component data for each voice unit and the voice pitch indicated by the pitch data generated in step S6. At this time, in the new harmonic component data, if the spectrum intensity is adjusted so that the shape of the spectral envelope represented by the harmonic component data subjected to the processing in step S5 is inherited, the timbre of the voice signal input in step S1 is reproduced. can do.
[0060]
In step S8, the harmonic component data generated in step S7 and the anharmonic component data subjected to step S5 are added for each speech unit. In step S9, the data subjected to the addition process in step S8 is converted into a time domain singing voice signal by inverse FFT or the like for each voice unit. The singing voice signal obtained as a result of the processing in step 9 is a digital signal, and is converted into an analog singing voice signal by the D / A converter 22.
[0061]
As an example, in order to synthesize a singing voice “saita”, the speech unit “#s”, “s” a "," a "," a " i "," i "," i t "," t Speech segment data corresponding to “a”, “a”, and “a #” (# represents silence) are read out and connected in step S5. In step S7, harmonic component data having a pitch related to the conversion in step S6 is generated for each speech unit, the addition process in step S8, the conversion process in step S9, and the D / A conversion process in the converter 22 After that, a singing voice signal of “Cita” is obtained.
[0062]
The processing in steps S1 to S9 described above may be executed as software processing in a small computer such as a personal computer, or may be executed using hardware such as an electronic circuit.
[0063]
【The invention's effect】
As described above, according to the present invention, the pitch conversion data stored in the storage means is used to convert the input note pitch into the voice pitch for singing synthesis. An effect of reproducing the fluctuation can be obtained. In addition, it is possible to faithfully reproduce unstable pitch fluctuations over time in actual human utterances, express differences in utterance pitch accuracy by range, and express differences in pitch changes by singers. It becomes possible.
[0064]
In addition, since the pitch conversion is performed by adding the variation of the voice pitch indicated by the pitch difference data to the input pitch, there is an advantage that the amount of data to be stored can be reduced.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a singing voice synthesizing apparatus according to an embodiment of the present invention.
FIG. 2 is a block diagram for explaining the operation of the pitch converter.
FIG. 3 is a graph showing an example of a pitch conversion function.
4A and 4B are diagrams illustrating an example of pitch conversion using the pitch conversion function of FIG. 3, in which FIG. 4A is a graph showing a pitch change before conversion, and FIG. 4B is a graph showing a pitch change after conversion; It is.
FIG. 5 is a graph showing the storage status of pitch difference data in a database.
6A and 6B are diagrams showing an example of pitch conversion using the pitch difference data of FIG. 5, in which FIG. 6A is a graph showing pitch change before conversion, and FIG. 6B is a graph showing pitch change after conversion. It is.
FIG. 7 is a block diagram showing a singing voice synthesizing apparatus according to another embodiment of the present invention.
[Explanation of symbols]
10: input unit, 12: pitch converter, 14: pitch converter, 16, 20: database, 18: singing synthesizer, 22: D / A converter, S1: singing voice signal input processing, S2: SMS analysis and Section extraction processing, S3: Lyric data and melody data input processing, S4: Phoneme string-speech segment conversion processing, S5: Speech segment connection processing, S6: Pitch conversion processing, S7: Harmonic component generation processing, S8: Addition processing , S9: Processing to convert to a time domain singing voice signal.

Claims (7)

A pitch conversion device used in a singing voice synthesizing device provided with a singing voice synthesizing means for synthesizing a singing voice signal having a pitch indicated by pitch data,
Input means for sequentially inputting pitches corresponding to sequential singing voices to be synthesized;
A pitch conversion function for converting a plurality of input pitches to a plurality of audio pitches, and when the input pitch is lower than a predetermined lower limit pitch, the input pitch is higher than the input pitch so that the input pitch is higher than the predetermined upper limit pitch. A memory for storing a pitch conversion function for converting the input pitch to be equal to the input pitch when the input pitch is between the predetermined lower limit pitch and the predetermined upper limit pitch so that the input pitch is lower than the input pitch when the input pitch is higher than Means ,
Conversion means for converting the pitch into a voice pitch based on the pitch conversion function for each pitch inputted from the input means, and supplying data indicating the voice pitch to the singing voice synthesis means as the pitch data Pitch converter. The input means inputs singer data indicating a singer, the storage means stores the pitch conversion function for each singer, and the conversion means is a pitch corresponding to a singer indicated by the singer data. The pitch conversion apparatus according to claim 1, wherein the pitch conversion is performed based on the conversion function .   The pitch conversion device according to claim 1 or 2, wherein the conversion means adds a random pitch variation depending on an input pitch to the voice pitch during pitch conversion.   The pitch conversion device according to claim 1, wherein the conversion unit adds a time-dependent pitch fluctuation to the voice pitch during pitch conversion. A pitch conversion function for converting a plurality of input pitches to a plurality of audio pitches, and when the input pitch is lower than a predetermined lower limit pitch, the input pitch is higher than the input pitch so that the input pitch is higher than the predetermined upper limit pitch. A memory for storing a pitch conversion function for converting the input pitch to be equal to the input pitch when the input pitch is between the predetermined lower limit pitch and the predetermined upper limit pitch so that the input pitch is lower than the input pitch when the input pitch is higher than means, a pitch conversion method used in the singing voice synthesizing apparatus and a singing voice synthesis means for synthesizing a singing voice signal having a pitch indicated by pitch data,
Inputting a pitch sequentially corresponding to each of the sequential singing voices to be synthesized;
Converting each pitch into a voice pitch based on the pitch conversion function and supplying data indicating the voice pitch to the singing voice synthesizing unit as the pitch data for each pitch input in this step. . A program used in a song synthesizer comprising a computer and a song synthesis means for synthesizing a song voice signal having a pitch indicated by pitch data, the computer comprising:
Input means for sequentially inputting pitches corresponding to sequential singing voices to be synthesized;
A pitch conversion function for converting a plurality of input pitches to a plurality of audio pitches, and when the input pitch is lower than a predetermined lower limit pitch, the input pitch is higher than the input pitch so that the input pitch is higher than the predetermined upper limit pitch. A memory for storing a pitch conversion function for converting the input pitch to be equal to the input pitch when the input pitch is between the predetermined lower limit pitch and the predetermined upper limit pitch so that the input pitch is lower than the input pitch when the input pitch is higher than Means ,
A program for converting the pitch into a voice pitch based on the pitch conversion function for each pitch input from the input means and functioning as a conversion means for supplying data indicating the voice pitch as the pitch data to the singing voice synthesizing means . A program used in a singing synthesizer provided with a computer, wherein the computer is
Input means for sequentially inputting pitches corresponding to sequential singing voices to be synthesized;
A pitch conversion function for converting a plurality of input pitches to a plurality of audio pitches, and when the input pitch is lower than a predetermined lower limit pitch, the input pitch is higher than the input pitch so that the input pitch is higher than the predetermined upper limit pitch. so as to be lower than the input pitch is higher than, if the input pitch is between a predetermined lower limit pitch and a predetermined upper limit pitch stores pitch conversion function that converts the so that Do equal to the input pitch Storage means ;
Conversion means for converting the pitch into a voice pitch based on the pitch conversion function for each pitch inputted from the input means, and sending pitch data indicating the voice pitch;
A program that functions as a singing voice synthesizing means for synthesizing a singing voice signal having a voice pitch indicated by pitch data sent from the converting means.

Images