JP2765192B2 - Electronic musical instrument - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention extracts a formant of a sound from the outside, and then reproduces a sound generated by resynthesizing the extracted formant in accordance with a performance pitch. Electronic musical instruments.

[Prior art] Conventionally, in an electronic musical instrument that produces a sampled sound, for example, to produce a vowel / a / of a voice,
The formant of the vowel / a / is stored in advance, and at the time of sound generation, the stored formant is read out and re-synthesized.
At this time, if the first formant having a close frequency component to the higher-order formant having a high frequency component are stored in advance, a sound with good reproducibility can be generated by synthesizing these.

In the conventional electronic musical instrument described above, the vowel / a, i,
In the case where all u, e, o / are to be pronounced, the formants of each vowel are also stored from low to high order formant frequencies, and a predetermined formant is read out at the time of pronunciation, It was pronounced by recombining.

"Problems to be Solved by the Invention" By the way, in the above-mentioned conventional electronic musical instruments, when a sound (vowel or wind instrument sound) having various characteristics (tones or vocal sounds) is to be pronounced, the timbre or timbre changes. Each time, it is necessary to store the formants (plural) of all the sounds to be produced, so that a large memory capacity is required, and there has been a problem that the size and cost of the musical instrument cannot be reduced.

The present invention has been made in view of the above problems,
It is an object of the present invention to provide an electronic musical instrument capable of reducing the memory capacity, the size of the device, and the cost.

[Means for Solving the Problems] In order to solve the above-mentioned problems, the present invention provides a formant extracting means for sampling a sound supplied from the outside and extracting a formant characterizing noise, and a plurality of formants each having a different timbre. Standard value storage means in which sound formants are respectively stored as standard values, and the plurality of sounds based on the extracted formants based on a relationship between the standard formants corresponding to the sampled sounds and the extracted formants of the sampled sounds. And formant estimating means for estimating a formant corresponding to each of the following, and formant synthesizing means for resynthesizing the estimated formant and outputting it as a tone signal.

[Operation] A sound supplied from the outside is sampled by the formant extracting means, and a formant characterizing the sound is extracted. Based on the relationship between the extracted formant and the standard formant stored in the standard value storage means, formant estimation means estimates formants of other sounds having different timbres based on the extracted formant. Then, the formant synthesizing means re-synthesizes the formant estimated by the formant estimating means and outputs it as a tone signal.

"Example" Next, an example of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of one embodiment of the present invention. In this embodiment, an electronic musical instrument that produces sounds having a plurality of timbres (particularly, vowels) will be described. In this figure, 1 is a formant extraction unit,
Fourier transforms the voice supplied from the outside and converts the first formant F1, the second formant F2, and the third formant F3
And the fourth formant F4 are extracted, and the first to third formants are supplied to the movement amount calculation unit 2. The fourth formant is directly supplied to a formant synthesizing unit 4 described later. This is because the fourth formant has high-order frequency components that are affected by the vocal tract and pronunciation morphological differences, which are components representing “individuality,” and does not vary much depending on the type of vowel. The point is that by adding it to other formants in formant synthesis, individual characteristic voices can be reproduced. The movement amount calculation unit 2 determines the movement amount of the standard putter based on the ratio of the first to third formants, and supplies the movement amount to the vowel estimation unit 3. Here, the standard pattern and the amount of movement thereof will be described. In general, the formant components of the vowels / a, i, u, e, o / are shown in FIG. 2 with the first formant being on the horizontal axis and the second formant being on the vertical axis. The shape formed by connecting each point is called a standard pattern. The formant frequency of this standard pattern is as follows.

However, since the actual frequency of each formant varies depending on the speaker and the environment, it is not always located at the standard pattern shown in FIG. But,
In general, even a pattern based on a formant of a voice under different conditions moves in parallel while maintaining the relative positional relationship of the five vowels (/ a, i, u, e, o /) (broken line in FIG. 2). See pattern by.) Therefore, if the formant of any one vowel supplied from outside is known,
The amount of movement (deviation) can be determined from the ratio between the formant of the vowel and the formant of the corresponding vowel in the standard pattern. That is, the above-described movement amount calculation unit 2 calculates the movement amount indicating how much the standard parameter should be moved so that the vowel corresponding to the standard pattern matches the vowel formant supplied from the outside.

The vowel estimation unit 3 calculates the first vowel based on the amount of movement described above.
To the third formant and supply it to the formant synthesis unit 4. The estimation of the formant frequency is obtained by multiplying the above-described movement amount by the formant frequency of the standard pattern.

The formant synthesizing unit 4 is a well-known one. For example, the formant synthesizing unit 4 multiplies a window function waveform generated at each predetermined pitch cycle by a frequency form having a formant center frequency as a pitch while initializing a phase at each pitch cycle. Find a tone signal of another vowel.

According to the configuration described above, for example, a vowel as an input vowel
When / a / is input to the formant extraction unit 1, the formant extraction unit 1 extracts the first to fourth formants of the input vowel / a /. The first, second, and third formants are supplied to the movement amount calculation unit 2, and the fourth formant is supplied to the formant synthesis unit 4.

Next, the movement amount calculator 2 calculates the movement amount based on the input vowel formants. Here, a simple example will be described. First, the input vowel is set to / a /, and the first and second formants of the input vowel / a / are each 900 Hz.
Let it be 1500 Hz (see input vowel in FIG. 2). Next, the movement amounts of the first and second formants of the input vowel / i / are calculated from these values. The amount of movement of the first formant is the first formant frequency of the input vowel / a / (= 90
0) divided by the first formant frequency (= 700) of the vowel / a / of the standard pattern (moving amount; 900/70
0). Similarly, the movement amount of the second formant is obtained by changing the second formant frequency (= 1500) of the input vowel / a / to the second formant frequency (= 1400) of the vowel / a / of the standard pattern.
It is obtained by dividing by (moving amount; 1500/1400). Hereinafter, the movement amount is similarly calculated for the third formant of the vowel / a /.

These movement amounts are supplied to the vowel estimation unit 3. The vowel estimation unit 3 estimates the first to third formants of other vowels / i, u, e, o /. According to the above example, the first vowel / i /
Estimation of the formant frequency is based on the above movement amount (= 900/70
0) is multiplied by the first formant frequency (= 200) in the vowel / i / of the standard pattern to obtain a value of about 257 (Hz) as the estimated formant frequency. The estimation of the second formant frequency is performed by multiplying the movement amount (= 1500/1400) by the second formant frequency (= 2100) in the standard pattern.
A value of about 2250 (Hz) is obtained as the estimated formant frequency. Similarly, the third formant frequency is obtained.

In addition, the vowel / u / is similarly moved (= 900/700)
Is multiplied by the first formant frequency of the standard pattern (= 300) to obtain a value of about 385 (Hz) as the estimated formant frequency. Also, the moving amount (= 1500/1400) is multiplied by the second formant frequency (= 1300) of the standard pattern,
A value of about 1392 (Hz) is obtained as the estimated formant frequency.

Hereinafter, the vowel / e, o
The first, second and third formants of / are also estimated (see dashed pattern in FIG. 2).

The first to first of each vowel / i, u, e, o / obtained in this way
The third formant is as shown in Table 2 below.

The formant frequency thus obtained can be supplied to the formant synthesizing unit 4. The formant synthesizing unit 4 converts a window function waveform generated every predetermined pitch cycle into
A periodic waveform whose pitch is the formant center frequency. Multiplication is performed to obtain and output a musical tone signal of each vowel.

Next, a specific application example of the above-described embodiment will be described. FIG. 3 is a block diagram showing the configuration of this embodiment. In this figure, reference numeral 10 denotes a key information generation unit, which is constituted by a keyboard. The key information generating unit 10 is operated by a player, and generates a key code KC and a key-on KON corresponding to the operation by each of the synthesizing units 15a, 1 of the formant synthesizing unit.
Output to 5b, 15c, 15d. Reference numeral 11 denotes a formant control unit, which includes a formant analysis unit 11a, a formant parameter memory 11b, and a standard formant parameter memory 11c.
It is composed of The formant analysis unit 11a
It corresponds to the formant extraction unit in the figure, and takes out each formant by Fourier-transforming the input voice. The first to fourth formants of each vowel are stored in the formant parameter memory 11b. Here, the concept of the format of the formant stored in FIG. 4 is shown. In this figure, there are five vowels of / a, i, u, e, o /, and the formants for these vowels are distinguished by subscripts added to the capital letter F of the alphabet, and the n-th formant of the vowel type m is F _mn It is represented by Although only one formant parameter table may be used, a plurality of formant parameter tables may be prepared to correspond to various voice sounds. In this case, the table of each formant parameter is distinguished by the table designation number K.

Next, the standard formant parameter memory 11c is a table similar to the above-described formant parameter memory 11b, and stores the formants of each vowel for the above-described standard pattern. Each formant is distinguished by a subscript attached to the alphabet Fs. A plurality of standard formant parameter tables may be prepared similarly to the above-described formant parameter table, and may be distinguished by the table output number K.

Reference numeral 12 denotes a musical tone setting unit which supplies various variables to each unit in order to control the entire speech synthesis system. Music setting section 12
Outputs the input vowel type designation variable M and the formant parameter table designation number K to the formant control unit 11. Also, a formant sound control unit described later
The table designation number K, the pronunciation vowel designation variable M ', and the mode MODE are output to 13. The pronunciation vowel designation variable M 'is a variable for designating a vowel to be pronounced. Mode MODE
Is a variable for designating which formant data of the formant parameter memory 11b or the standard parameter memory 11c to use at the time of sound generation, and takes a state of "0" or "1".

The formant sound control unit 13 described above is provided with the mode MO
If DE is "1", formant parameter memory
The formant of the vowel designated by the pronunciation vowel design variable M '(M' = 1, 2, 3, 4, or 5) in the K-th formant table of 11b is read, and the first formant is sent to the synthesis unit 15a. The second formant is sent to the synthesis unit 15b,
The third formant is output to the synthesizing unit 15c, and the fourth formant is output to the synthesizing unit 15d. When the mode MODE is "0", the vowel formant designated by the vowel designation variable M 'of the K-th formant table in the standard parameter memory 11c is read out, and each formant is read out in the same manner as described above. Is output to each of the combining units 15a to 15d.

The synthesizing units 15a to 15d correspond to the formant synthesizing units in FIG. 1. In each unit, a phase of a window function waveform generated at each pitch cycle specified by the key code KC is initialized at each pitch cycle. While, based on each formant, multiply the periodic waveform with the pitch of the formant center frequency to obtain the tone signal of each vowel,
Output to multipliers 16a, 16b, 16c and 16d. Multipliers 16a-1
6d adjusts the level of the tone signal according to the MIX signal, and then outputs it to the mixing unit 17. Mixing part 17
Mixes the level-adjusted first to fourth formants to form a tone signal and outputs it to the next-stage circuit (such as a sound system; not shown).

Next, the operation of the electronic musical instrument having the above-described configuration will be described with reference to the flowcharts shown in FIGS.

First, the operation of the formant control unit 11 will be described. When a predetermined switch operation is performed and a voice input is instructed by the performer, a flowchart shown in FIG. 6 is executed. First, in step SA1, the input voice is sampled. In this case, the input vowel is set in advance by the player, and is output from the tone setting unit 12 as an input vowel designation variable M (M = 1, 2, 3, 4, or 5). Next, in step SA2, the first formant _FM1 , the second formant _FM2 , the third formant _FM3, and the fourth formant _FM4 are extracted from the input voice by the formant analyzing section 11a. Here, for example, assuming that the input vowel designation variable M is designated as “1”,
/ a / first formant F ₁₁ , second formant F ₁₂ , third
Formant F ₁₃ and fourth formant F ₁₄ is extracted. Then, the process proceeds to Step SA3, where variables m = 1 and n = 1 are set as initial settings.

Next, at step SA4, the first formant F ₁₁ a vowel / a / calculated from the equation _{F 11 / Fs 11 × Fs 11} , is stored in the corresponding portion of the K-th formant table. Next, the process proceeds to Step SA5, where it is determined whether or not the variable m has become “5”. In this case, since the variable m is still “1”, the step
The determination result at SA5 is “NO”, and the process proceeds to Step SA6. In step SA6, after the variable m is incremented, the process returns to step SA4. At step SA4, the vowel / i / the first formant F ₂₁ determined from equation _{F 11 / Fs 11 × Fs 21} , also stores the predetermined point K-th formant table. Hereinafter, similarly, the determination in step SA5 is repeated step SA4, SA5 and SA6 until "YES", vowel / u, e, o / the first formant F _31, F ₄₁ and F
₅₁ is obtained and stored in a predetermined location of the K-th formant table. Then, when the variable m becomes “5”, the judgment result in step SA5 becomes “YES”, and
Proceed to SA7. In step SA7, it is determined whether or not the variable n has become "3". In this case, the variable n is set in step SA3
Is "1" as it is in the initial setting. Therefore, the determination result in step SA7 is “NO”, and the
Proceed to A8. In step SA8, the variable n is incremented and the variable m is set to "1".
Return to Then, again, step SA4, SA5, SA6 repeated, the second formant _{F 12, F 22, F 32} , F 42, F 52 of each vowel seeking to store the predetermined portion of the K-th formant table. Then, the variable m becomes “5” and the step
When the determination result at SA5 becomes “YES” again, the step
Proceed to SA7. Since the variable n is “2”, the result of the determination in step SA7 is “NO”, and the flow proceeds to step SA8.
In this step SA8, the variable n is incremented to “3” and the variable m is set to “1”, and then the process returns to step SA4. Then, again, step SA4, SA5, SA6 repeated, third formant F ₁₃ of each _{vowel, F 23, F 33, F} 43,
Seeking F _53, and stores the predetermined point formant table. Then, the variable m becomes “5” and the step SA
When the determination result at step 5 becomes “YES” again, step SA
Go to 7. Since the variable n is “3”, the result of the determination in step SA7 is “YES”, and the flow proceeds to step SA9. In step SA9, as it is stores the fourth formant F ₁₄ extracted in the step SA2 each vowel / i, u, e, as o / the fourth formant, the routine ends.

In this way, the formants of the other vowels are all obtained based on the input vowel by the processing of the formant control unit 11,
Stored in the formant parameter memory 11b. Therefore, since only the standard pattern needs to be stored in advance, the memory capacity can be reduced. In the embodiment, the input vowel designation variable M is 1
May be set to any one of values from 5 to 5, for example, “2”.
Then, formants of other vowels are obtained based on the vowel / i /.

When the voice input is completed, the pronunciation is performed next. That is, when the player performs using the keyboard, the key information generator 10
Supplies a key code KC and a key-on KON corresponding to the performance to each of the synthesizing units 15a, 15b, 15c and 15d. Further, the musical tone setting unit 12 supplies the table designation number K, the pronunciation vowel designation variable M ′ and the mode MODE (0 or 1) to be used to the formant emission control unit 13. Then, the flowchart shown in FIG. 7 is executed. First, in step SB1, it is determined whether or not the variables K and M 'and the mode MODE have been set. If these variables are not set, the result of the determination in step SB1 is "NO",
Step SB1 is repeatedly executed until the above variables are set. In this case, since the variables have already been supplied from the tone setting unit 12, the result of the determination in step SB1 is "YES", and the routine proceeds to step SB2. In step SB2, it is determined whether or not the mode MODE is "0". In this case, assuming that the mode MODE has been set to "0", the result of the determination at step SB2 is "YES", and the routine proceeds to step SB3. In step SB3, the vowel designated by the pronunciation vowel designation variable M 'in the K-th table of the standard formant parameter memory 11c (in this example, for example,
M ′ = 4; / e /), the first formant Fs ₄₁ , the second formant Fs ₄₂ , the third formant Fs ₄₃ and the fourth formant Fs ₄₄ were supplied to the respective synthesis units 15a, 15b, 15c, 15d. Thereafter, the process returns to step SB1, and waits for the next variable setting operation.

Further, in the combining unit 15a, 15b, 15c and 15d, the sound in accordance with the first formant Fs _41, second formant Fs _42, third formant Fs _43, parameters supplied from the fourth formant Fs ₄₄ and tone setting unit 12 The signals are synthesized, and each audio signal is supplied to multipliers 16a, 16b, 16c and 16d, respectively. After the level of the audio signal is adjusted in each multiplier, the audio signal is mixed by the mixing unit 17, and is then emitted by a subsequent circuit as a vowel / e / having a pitch corresponding to the key code KC.

On the other hand, if the mode MODE is set to “1”, the result of the determination in step SB2 is “NO”, and the flow proceeds to step SB4. At step SB4, the first vowel designated by the pronunciation vowel designation variable M 'of the K-th table of the formant parameter memory 11b (in this case, for example, M' = 4; / e / as described above). Formant F ₄₁ ,
Second formant F _42, third formant F ₄₃ and the fourth
Formant F ₄₄ each synthesis units 15a, 15b, 15c, 15d
After returning to step SB1, the process returns to step SB1 to wait for the next variable setting operation.

Further, in the combining unit 15a, 15b, 15c and 15d, the first formant F _41, second formant F _42, third formant F _43, the fourth formant F ₄₄ and the tone setting section 12
The audio signals are synthesized according to the parameters supplied from, and the respective audio signals are supplied to multipliers 16a, 16b, 16c and 16d, respectively. After the level of the audio signal is adjusted in each multiplier, the audio signal is mixed by the mixing unit 17,
Vowel at pitch corresponding to key code KC / e
Pronounced as /.

In the tone setting section 12, the pronunciation vowel designation variable M '
Is changed, the next time the keyboard is operated, the changed vowel is pronounced. Thus, in this embodiment, since the fourth formant is added as it is, there is an advantage that a vowel having extremely individual characteristics can be pronounced.

Note that the formant extraction method in the above-described embodiment may be a fast Fourier transform, a composite sine wave method, a linear prediction method (for example, the LSP method), or the like.

In the embodiment, the vowels are five types and the formants are the first to fourth formants. However, the number is not limited to this.

In addition, although the standard pattern and the input voice are Japanese vowels, the present invention is not limited to this, and other languages or instrument sounds may be used.

Further, in the embodiment, the fourth formant is configured to use the input form as it is, but may use the formant obtained by calculation in the same manner as the low-order formant.

In the embodiment, the fourth formant is used as a higher-order formant to be added to a lower-order formant in formant synthesis, but a higher-order formant may be added.

Further, in the embodiment, after inputting a voice, a parameter table is created in the memory and used. However, the input sound is analyzed and calculated in real time to obtain a formant, and the formant is obtained. May be supplied to the synthesizing unit. Therefore, in this case, it is not necessary to store the estimated formants in the memory, and only the standard pattern need be stored, so that the memory capacity can be further reduced.

Further, in the embodiment, the type of the input vowel is set by the variable M, but the input vowel and the standard pattern may be compared (voice recognition) to recognize which vowel is input. This has the advantage that the setting of the input vowel can be omitted.

[Effects of the Invention] As described above, according to the present invention, only the standard formants are stored, and the formants of other sounds having characteristics of the sampled sound and having different timbres are referred to as the standard formants. Estimation is performed based on the ratio of an externally supplied sound to the estimated formant, and the estimated formant is re-synthesized to generate a sound, thereby reducing memory capacity, miniaturizing the device, and reducing cost. The advantage of reduction can be obtained.

[Brief description of the drawings]

FIG. 1 is a block diagram showing the configuration of one embodiment of the present invention,
FIG. 2 is a diagram showing a relationship between vowels and formants for explaining a standard pattern and a vowel estimation method used in the embodiment, FIG. 3 is a block diagram showing a configuration of an application example of the present invention,
FIG. 4 is a conceptual diagram of a format of a formant frequency stored in a formant parameter memory used in the application example, and FIG. 5 is a conceptual diagram of a format of a standard pattern stored in a standard formant parameter memory used in the application example. 6 and 7 are flowcharts for explaining the operation of the application example. 1. Formant extraction unit (formant extraction means)
2... A moving amount calculating section, 3... A vowel estimating section (vowel estimating means), 4... A formant synthesizing section (formant synthesizing means), 12 c... A standard formant parameter memory (standard value storing means).

Claims (1)

(57) [Claims] (1) sampling an externally supplied sound,
Formant extraction means for extracting a formant characterizing the sound; standard value storage means for storing formants of a plurality of predetermined tones each having a different timbre as a standard value; the standard formant corresponding to the sampling sound and the sampling A formant estimating means for estimating a formant corresponding to each of the plurality of sounds based on the extracted formant based on a relationship with a sound extraction formant; and a formant synthesis for resynthesizing the estimated formant and outputting it as a tone signal. An electronic musical instrument comprising: