JPH0812559B2 - Singing voice generator - Google Patents

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a singing voice sound generator, and more particularly to a singing voice sound generator that synthesizes formant frequencies to generate and output as a singing voice sound.

2. Description of the Related Art There are sound synthesis systems for playing music, one for synthesizing musical instrument sounds and the other for singing voice synthesis. Among them, the singing voice synthesizing system includes, for example, "singing voice synthesizing system for personal computer using formant type speech synthesizing LSI" (Ishihara, Fushikida, Sanru,
Inoue: 1984 IEICE General Conference, 6-198)
Etc. are known. This one decomposes the input voice character string and the musical score character string into parameters with the basic units of consonants and vowels, and creates a table of pitch parameters according to the scale data in the musical score character string. Then, the formant parameters and amplitudes retrieved from the data ROM are combined with the pitch parameters of the scale described above according to the table created by decomposing the consonant and vowel parameters, and the synthetic parameters are generated. Create a time length table according to the note length data. Then, the above-mentioned synthesis parameters are sequentially transferred to the voice synthesis LSI (large-scale integrated circuit) every predetermined frame period according to the time length table, and then,
A synthetic waveform is generated at 10 kHz sampling.

DISCLOSURE OF THE INVENTION Problems to be Solved by the Invention However, the conventional singing voice system described above can only generate a singing voice sound, and it is difficult to make a musical change such as a vibrato and the musical expression is poor. Atsuta

Further, as a device for generating a human voice sound at a designated pitch, there is also a device disclosed in JP-A-55-77799, but this one approximates consonants with white noise,
There was a problem that abrupt changes in formant frequency could not be accurately reproduced, and the human voice generated subsequently was unclear and unnatural.

Therefore, the present invention is particularly applicable to MIDI (Musical Instrument Di
It is an object of the present invention to provide a singing voice sound generator that solves the above problems by connecting to a synthesizer using an interface of the standard.

Means for Solving Problems A singing voice sound generating apparatus according to the present invention includes a table creating means,
Pitch parameter conversion means, synthetic parameter creation means,
It comprises a data conversion means, a data transfer means, a synthesizer and an interface means. The table creating means decomposes the input lyric data into parameters in units of consonants and vowels to create a first table, and the pitch parameter converting means creates pitch parameters in accordance with the scale data in the input score data. For example, it is read from the second table corresponding to the scale prepared in advance. The synthesis parameter creating means combines the formant parameters read from the memory according to the first table with the pitch parameters to generate edited and interpolated synthesis parameters, and according to the tone value data in the score data. Create time length tables and store them temporarily.

This synthesis parameter is converted into data of a specific standard by the data conversion means, data of a plurality of formant frequencies of the converted data is transferred by the data transfer means, and pitch frequency data of pitch frequencies is converted into vowels. Only during the vocalization period, a period is generated and output based on the time length table. The interface means separately supplies data concerning a plurality of formant frequencies and pitch frequency data among the transfer data by the data transfer means as control signals for a plurality of variable frequency oscillators in the synthesizer.

Function A plurality of variable frequency oscillators in the synthesizer generate and output signals of a plurality of formant frequencies that form consonants and vowels, respectively, and also generate a pitch frequency signal that determines the scale based on the score data during the vowel vocalization period. Since it is output, a singing voice sound is produced from the speaker of the synthesizer. Since the synthesizer is provided with a means for giving a musical effect to the sound to be pronounced, it can be used to give a musical effect to the singing voice. Further, according to the present invention, since the parameters are edited and interpolated by the synthetic parameter creating means, it is possible to generate a singing voice sound with a fine nuance. Hereinafter, the present invention will be described in detail together with examples.

Embodiment FIG. 1 shows a block system diagram of an embodiment of the device of the present invention. In the figure, the lyric keyboard 1 has a structure in which a plurality of alph-abet keys 17 are arranged on a keyboard body 16 in three stages, and alf-abet keys indicating vowels are arranged at the bottom, as shown in FIG. And multiple keys 17
Is selectively pressed according to the lyrics (voice character string), which is a collection of characters written in Roman letters. For example, the letter "bi" is It is input by pressing the two keys of. As shown in FIG. 2, the lyrics keyboard 1 is further provided with a female voice key 18 and a child voice key 19 at the lowest position, and when the female voice is the lyric sound to be pronounced, the key 18 is used. Can be specified by pressing the key, or the key 19 for a child's voice, and not pressing both the keys 18 and 19 for a male voice. The lyrics data (character voice data and voice-specific data) input by the lyrics keyboard 1 is supplied to a central processing unit (CPU) 3 through an I / O (input / output) interface 2.

Other keyboards include a tone value keyboard 4, a control keyboard 5, and a keyboard in a synthesizer 10 described later. The tone value keyboard 4 has a structure in which a plurality of key groups 21 indicating tone values (time values) are arranged in a matrix as shown in FIG. 3, for example. On the other hand, the control keyboard 5 is a keyboard for estimating the pitch of a sound within a sound range of several octaves, for example, by a number within the range of "1" to "88", and designating the strength and weakness of the sound.
Both output data of the note value keyboard 4 and the control keyboard 5 are supplied to the CPU 3 via the I / O interface 2 as musical score data (musical score character string).

A random access memory (RAM) 6, a read only memory (ROM) 7 and a memory (for example, a magnetic disk) 8 are connected to the CPU 3 via a bidirectional bus. The RAM 6 is a memory circuit for data storage and working of the CPU 3, and has previously stored pitch parameters and the like corresponding to scales described later. Further, the ROM 7 pre-stores a control program of the CPU 3 and a formant parameter table regarding the formant frequency. Further, the memory 8 is a memory circuit for storing input or processed data.
Also, the I / O interface 2 is connected to a MIDI interface 9, which will be described later, via a bidirectional bus.
The DI interface 9 is connected to the synthesizer 10 via a bidirectional bus.

The lyrics data and the musical score data input to the CPU3 are written in the video random access memory (V.RAM) 12 via the video controller 11, converted into the video signal by the video controller 11, and then separated into the three primary color signals. A circuit 13 converts the signals into three primary color signals and supplies them to a CRT (cathode ray tube) 14. The CRT 14 is also supplied with the three primary color signals based on the data generated by the CPU 3. As a result, the operator can see the table displayed on the CRT 14 showing, for example, note values, scales, lyrics, and other necessary items while the keyboard 1,
You can check and correct the input of the desired data by operating 4 and 5.

Further, the CPU 3 is configured as shown in FIG. 4, and operates according to the flow chart shown in FIG. CPU3
Is first initialized (indicated by step 40 in FIG. 5) and then input from the keyboard 1, 4 and 5 via the I / O interface 2 by the lyrics data / musical score data receiving unit 25 in FIG. The lyrics data and the score data are respectively received and fetched (step 41 in FIG. 5). 4th in CPU3
The table creating means 26 shown in the figure decomposes the acquired lyrics data into first parameters in units of consonants and vowels, thereby creating a first table, which is stored in the RAM6.
(Step 42 in FIG. 5). In other words, it is known that most Japanese syllables consist of a combination of consonants and vowels in a broad sense, and it is thought that consonants and vowels are used as a unit to display the phonetic form of a word. Because
The first parameter can indicate individual syllables indicating lyrics.

Further, the pitch parameter converting means 27, in accordance with the scale data in the score data taken in, outputs a pitch parameter indicating a fundamental frequency (pitch frequency) F ₀ of a vowel defining a scale from a table stored in advance in the RAM 6. It is read (that is, converted into pitch parameters, which is indicated by step 43 in FIG. 5). Also, the CPU3 shown in FIG.
The synthesis parameter creating means 28 in the table creating means 26 is a table creating means 26.
By reading the formant parameter from the ROM 7 while referring to the first table stored in the RAM 6 and combining it with the pitch parameter obtained by the pitch parameter conversion means 27 to edit the parameter. Interpolation (such as dividing the sound so that the sound changes smoothly) is performed to create synthesis parameters, and a time length table is created according to the note value data in the score data, and these are created. It is temporarily stored in RAM6 (step 44 in FIG. 5).

Here, the singing voice is composed of individual syllables indicating lyrics and sounds generated together with the syllables. The former is represented by consonants and vowels by the first parameter, and has a formant frequency vs. time characteristic as shown in FIG. It is known to show. That is, it is considered that the voice is identified based on the fundamental frequency (pitch frequency) forming the voice waveform and a plurality of formants. Letting F ₁ , F ₂ and F ₃ be the first, second and third formants (hereinafter also referred to as formant frequencies), one syllable is first pronounced in a consonant part with a gradual change in formant frequency, then The vowel part is pronounced with a substantially constant formant frequency. In the vowel vocalization period, the pitch frequency is generated together with the formant frequencies F _{1 to} F ₃ .

The changes in the formant frequencies F _{1 to} F ₃ in the consonant part differ depending on the consonant. For example, the formant frequencies of the consonants g, d and b are as shown in FIGS. 7 (A), (B) and (C). It is generally known to change to. Here, the formant parameters in the above-mentioned synthetic parameters are parameters that determine the above-mentioned three types of formant frequencies F _{1 to} F ₃ . The pitch parameter defines the pitch frequency.

On the other hand, the above-mentioned sound generated together with the syllable (lyric) is determined by the pitch, size, and length of the sound, and this is input to the CPU 3 as music score data as described above. The formant frequency vs. time characteristic shown in (1) has a characteristic that it is translated in the vertical axis direction, and the pitch parameter (pitch frequency) determines the amount of movement (pitch of sound).
The length of the sound can be obtained by the time length of the vowel part. Further, the loudness of the sound is input by the control keyboard 5, but is included in the above-mentioned synthesis parameter as a volume parameter.

This synthesis parameter is converted by the MIDI data conversion means 29 into data (referred to as MIDI data) conforming to the known MIDI standard (the process of step 45 in FIG. 5 is performed). Here, the MIDI standard converts various operations performed when playing an instrument (pressing a key, turning a volume, etc.) into several bytes of data, and
Or, it is a standard for transmitting and receiving with a machine that controls them. MIDI data is transmitted as 2-3 bytes of serial data, and the first 1 byte is the status, and it is a message that indicates what kind of function the subsequent data bytes have. The data length including the status is indefinite as 1 to 3 bytes, and the status and the data are recognized as the status when the first bit of each byte is "1" and the data when the first bit is "0". The above messages are roughly classified into a channel message for sending individual operation data and a system message for controlling the network.

The above-mentioned converted MIDI data is supplied to the data transfer means 30. Here, according to the processing operation of steps 46 to 56 shown in FIG. 5, the CPU 3 transmits the I / O interface 2 and the MIDI interface 9 respectively. And transferred to the synthesizer 10. That is, in FIG. 5, the data transfer means 30 sets the variable J to "1" (step 46), the variable I to "0" (step 47), and then the synthesizer 10
Based on the data coming in via the MIDI interface 9 and the I / O interface 2, the synthesizer 10 determines whether or not it is receivable, and waits until it becomes receivable (step 48). When the synthesizer 10 is ready to receive (enable), the first data of the consonant formant frequencies F _{1 to} F ₃ and the volume of the MIDI data are I / O.
After transferring to the synthesizer 10 through the O interface 2 and MIDI interface 9, respectively, the value of the variable I is set to "1".
Only add (steps 49, 50). The processing operations of steps 49 and 50 are repeated until the value of the variable I after addition becomes N or more (step 51).

Here, the value of N is the vocalization time zone of the consonant part,
The number of frequency changes to be performed for each of the _first to _third formant frequencies F1 to F3 is shown at regular intervals (for example, 0.33 ms). As a result, each of the formant frequencies which cause the frequency changes of the formant frequencies F _{1 to} F _{3 in} the consonant vocalization time zone as shown in FIGS. The 3-byte first data is transferred, and the formant frequency data is rewritten about every 3 ms (≈3 × 3 × 320 μs).

When the value of the variable I is equal to or greater than N, then the MIDI vowel of data formant frequencies F ₁ to F _3, the pitch frequency F ₀ is transferred second data relating to pitch frequency F ₀ and volume and the like to the synthesizer 10 Are set (step
52). Then, the time count for checking the time length of the vowel part is performed so that the time length is in accordance with the tone value based on the time length table, and when the predetermined time length is reached, the second data is transmitted. Is turned off (step 5
3). The data transfer means 30 then clears the first data (step 54) and then increases the value of the variable J by "1" (step 55). The above steps 47 to 55 are repeated until the value of the variable J after the increase exceeds the total number m of notes to be generated, and when it exceeds m, the operation ends (step 5).
6).

It is supplied to the MIDI interface 9 through the MIDI data I / O interface 2 thus taken out, and further supplied to the synthesizer 10 from this. FIG. 8 shows a circuit diagram of an example of the MIDI interface 9. I / O
The MIDI data (first and second data) from the interface 2 is supplied to the data input terminals D _{0 to} D ₇ of the signal processing device 60, where parallel conversion is performed and the inverters 61, 62 and resistors are connected.
It is transferred to the synthesizer 10 via 63, the connector 64 and the like. The data from the synthesizer 10 is supplied to the signal processing device 60 via the connector 65 and the photo coupler 66, respectively,
It is output in parallel from the data input / output terminals D _{0 to} D ₇ . The signal processing device 60 is supplied with a clock of, for example, 500 kHz from the clock module IC 67.

FIG. 9 shows a block system diagram of an embodiment of the synthesizer 10. MIDI data from the MIDI interface 9 is input / output interface (not shown) in the synthesizer 10.
Are sequentially input to the input terminals 70 ₁ to 70 ₆ via. Input terminal
Input data for 70 ₁ to 70 ₄ is voltage controlled oscillator (VCO) 71 to 74
Are supplied as control signals. Where VCO 71, 72 and 73
Respectively generate and output the signals of the first, second and third formant frequencies, and the VCO 74 oscillates and outputs the signal of the pitch frequency F ₀ . The data incoming to the input terminal 70 ₅ is supplied to the low frequency oscillator (LFO) 75, to control its oscillating operation. The output signal of this LFO75 is supplied to the VCOs 71 to 74, and its output oscillation frequency can be variably controlled to obtain a vibrato effect, and the output signal of the LFO75 is a voltage control filter (VCF) 78.
And a voltage-controlled amplifier (VCA) 79, respectively, to emphasize only a specific overtone or obtain a tremolo effect.

The output signals of the VCOs 71 to 74 are supplied to the ring modulator 76, respectively, where they are converted into signals of the sum or difference frequency of each frequency and then supplied to the mixer 77, where the VCOs 71 to 74 are supplied.
Is mixed with each output signal of. The ring modulator 76 is used to obtain a specific tone, and its operation is
Controlled by data from 70 ₆ . The mixed output signal of the mixer 77 is passed through the VCF 78, the VCA 79 and the amplifier 80, respectively, and converted into a signal having a waveform as shown in FIG. In FIG. 10, T indicates the pitch cycle.

Since band consonants utterance time predetermined data only to the terminals 70 ₁ to 70 ₃ is applied, first from VCO71,72 and 73, the second and third formant frequencies F _1, F ₂ and F ₃ Are simultaneously output, and their frequency values are divided into N stages and sequentially changed, but no signal is generated and output from the VCO 74. On the other hand, the vowel vocalization time period is the second
Based on a part of the data of VCO74, the pitch frequency F ₀ that determines the scale is generated and output for the first time, and VCO71-73
From then on, the signals of the _first to third formant frequencies F _{1 to} F ₃ are respectively generated with constant values. As a result, the lyrics input by the lyrics keyboard 1 from the speaker 81 are pronounced as a human voice according to the musical score input by the tone value keyboard 4 and the control keyboard 5. Further, the human voice sound, that is, the singing voice sound can be provided with musical changes such as vibrato and tremolo. As a result, there is a nuance and a more preferable singing voice. Note that the synthesizer 10 shown in FIG. 9 can also obtain only instrument sounds by using the keyboard 82 as in the conventional case. Further, the pitch parameter conversion means is not limited to reading from the table, and its value can be calculated. Furthermore, the number of formants is not limited to three, and the present invention can be applied to a plurality of formants other than three.

EFFECTS OF THE INVENTION As described above, according to the present invention, a singing voice can be generated using a synthesizer, and a formant frequency or the like is synthesized. Therefore, compared with a conventional device that uses white noise as a consonant. In addition, it is possible to generate a natural and clear singing voice sound, and since the synthesizer is used, it is possible to easily give a musical change to the singing voice sound. Further, since the parameters are edited and interpolated according to the present invention, continuous sounds can be generated, and singing voice sounds in which fine nuances are expressed can be generated.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of the device of the present invention,
2 and 3 are diagrams showing a schematic configuration of each keyboard in the block system shown in FIG. 1, FIG. 4 is a block system diagram showing an embodiment of the main part of the device of the present invention, and FIG. FIG. 6 is a flow chart for explaining the operation of the main part of the device of the invention, FIG. 6 is a diagram showing the relationship between the consonant part, the vowel part, and the formant frequency of the synthetic voice,
FIG. 7 is a diagram showing the relationship between the formant frequency of each consonant and time, FIG. 8 is a circuit diagram of an example of the MIDI interface in the block system shown in FIG. 1, and FIG. 9 is the block system shown in FIG. FIG. 10 is a block diagram showing an embodiment of the synthesizer of FIG. 10, and FIG. 10 is a diagram showing an example of an input synthesized voice signal waveform of a speaker in the block system shown in FIG. 1 ... Lyrics keyboard, 2 ... I / O interface,
3 ... Central processing unit (CPU), 4 ... Sound value keyboard,
5 ... Control keyboard, 6 ... Random access memory (RAM), 7 ... Read-only memory (ROM), 9 ... MIDI interface, 10 ... Synthesizer, 11 ... Video controller, 12 ... … Video random access memory (V.RAM), 25 …… Lyrics data / musical score data receiving section, 26 …… Table creating means, 27…
... Pitch parameter conversion means, 28 ... synthesis parameter creation means, 29 ... MIDI data conversion means, 30 ... data transfer means, 70 ₁ to 70 ₆ ... input terminals, 71 to 74 ... voltage controlled oscillator (VCO) , 75 …… Low frequency oscillator (LFO), 77 …… Mixer.

Claims (1)

[Claims] 1. A table creating means for decomposing input lyrics data into parameters in units of consonants and vowels to create a first table, and a pitch for converting into pitch parameters according to scale data in the input score data. Parameter conversion means and formant parameters read from the memory in accordance with the created first table are combined with the pitch parameters to generate edited and interpolated synthetic parameters, and note value data in the score data. A time length table is created in accordance with the above, synthetic parameter creating means for temporarily storing them, data converting means for converting the synthetic parameter into data of a specific standard, and data relating to a plurality of formant frequencies among the converted data. The pitch frequency data relating to the pitch parameter is transferred and vowel vocalization period is transmitted. A data transfer means for generating and outputting a period based on the time length table only, a synthesizer for outputting mixed output signals of a plurality of variable frequency oscillators, and a plurality of formant frequencies among the transfer data from the data transfer means. A singing voice generating device comprising an interface means for separately supplying data and the pitch frequency data to the plurality of variable frequency oscillators in the synthesizer as control signals.