JP2722482B2 - Tone generator - Google Patents

Description

Description: BACKGROUND OF THE INVENTION The present invention relates to a musical tone generator that generates musical tones (including human voices) having fixed formant characteristics by using a waveform memory reading method.

(Prior Art) Conventionally, this type of apparatus divides a range over a plurality of octaves into a plurality of range groups, stores musical tone waveform data for each range group in a waveform memory, and generates a tone when generating a tone. By reading out the musical tone waveform data of the musical range group to which the frequency of the musical tone to be generated belongs at a rate corresponding to the frequency, the formants do not move even if the musical range changes, and the frequency corresponding to the frequency of the musical tone to be generated is adjusted. In addition, a tone with a fixed formant characteristic is obtained for tone frequencies over a plurality of octaves.

(Problems to be Solved by the Invention) In the above-mentioned conventional apparatus, although the fixed formant characteristics of the musical tones between the respective musical range groups are secured, the formants of the musical tones within the respective musical range groups are different from the frequencies of the generated musical tones. It moves according to the corresponding read rate. Accordingly, when the divided range is set wide, a large difference is generated between the lowest frequency and the highest frequency belonging to each range group, so that the movement of the formants within the range group is a serious problem in realizing a fixed formant. Becomes On the other hand, if you set the divided range narrow,
Although this problem is solved, in such a case, there is a problem that the capacity of the waveform memory required for storing the musical tone waveform data increases.

SUMMARY OF THE INVENTION An object of the present invention is to solve the above-mentioned problem, and an object of the present invention is to provide a tone generator capable of obtaining a tone having a good fixed formant characteristic even if the capacity of a waveform memory is saved. is there.

(Means for Solving the Problems) In order to achieve the above-mentioned object, a configurational feature of the present invention is a waveform storing musical tone waveform data obtained by processing waveform data representing an original sound so that its spectral envelope becomes smooth. A memory, reading means for reading out and outputting the musical tone waveform data stored in the waveform memory at a rate corresponding to the frequency of the musical tone to be generated, and inputting the musical tone waveform data output from the reading means, There is provided a filter means for giving a frequency characteristic corresponding to a difference between the spectrum envelope of the waveform data before processing and the spectrum envelope of the musical tone waveform data after processing.

(Operations and Effects of the Invention) In the present invention configured as described above, the waveform memory stores the musical tone waveform data obtained by processing the waveform data representing the original sound so that its spectral envelope becomes smooth. The musical tone waveform data stored in the waveform memory does not have a formant (a peak of a spectrum envelope), and well expresses the characteristics of the original sound except for the formant characteristics. Then, since the tone waveform data is read out at a rate corresponding to the frequency of the tone to be generated by the reading means and supplied to the filter means, the tone waveform data supplied to the filter means is:
It corresponds to the frequency of the musical tone to be generated, has a smooth spectral envelope by eliminating the shift of the formant of the original sound, and well expresses the characteristics of the original sound excluding the formant characteristics. Then, the filter means assigns a frequency characteristic corresponding to the difference between the spectrum envelope of the waveform data before processing and the spectrum envelope of the processed sound waveform data to the read and output tone waveform data. The tone waveform data output from the above-described device has the formant characteristics of the original sound, irrespective of the readout rate.

Therefore, according to the present invention, the musical tone represented by the musical tone waveform data output from the filter means has a desired frequency and faithfully reproduces the formant characteristics of the original sound. Also, it is not necessary to divide the sound range for the fixed formant, and it is possible to obtain a musical sound with good fixed formant characteristics even if the capacity of the waveform memory is reduced while taking advantage of the waveform memory system.

(Embodiment) Hereinafter, a specific embodiment in which a tone generator according to an embodiment of the present invention is applied to an electronic musical instrument will be described. Before that, the basic principle of the tone generator and the device used in the same are described. A method for creating various data will be described.

FIG. 3 is a block diagram showing a basic circuit of a tone generator, which comprises a waveform memory 11, a formant filter 12, and a formant filter coefficient memory 13.
Consists of The waveform memory 11 stores tone waveform data having a spectral envelope in which the formant characteristics of the original sound (musical sound to be generated) are smoothed, that is, if the original sound has a spectral envelope as shown by a broken line A in FIG. Musical tone waveform data having a spectrum envelope as shown by a solid line B connecting the vertices a, b, and c of FIG. The formant filter 12 is controlled by a filter coefficient output from the formant filter coefficient memory 13, and reads out and outputs from the waveform memory 11 a frequency characteristic for changing a spectrum envelope like a solid line B to a spectrum envelope like a broken line A. As shown in FIG. 7, the frequency characteristic is changed from the spectrum envelope of the original sound (broken line A in FIG. 6) to the spectrum envelope of the musical sound waveform data (solid line B in FIG. 6). ) Is subtracted. In this way, after the tone waveform data having the characteristics of the original sound excluding the formant characteristics are read from the waveform memory 11, the read tone waveform data is given the formant characteristics of the original sound independently of the reading. Therefore, even if the read rate changes, the formants do not move. As a result, a tone having good fixed formant characteristics can be obtained while utilizing the characteristics of tone synthesis by the waveform memory system that faithfully reproduces the waveform of the original sound.

Next, a method of creating the musical tone waveform data stored in the waveform memory 11 and the filter coefficients stored in the formant filter coefficient memory 13 will be described. FIGS. 4 and 5 are block diagrams showing the processing for creating the tone waveform data and the filter coefficients. First, for example, an instrumental sound such as a trumpet or a violin or a human voice is input as an original sound (block 21), and the original sound is subjected to a fast Fourier transform (FFT) process (block 22).
The spectrum envelope of the original sound is detected as shown by the broken line A in the figure (block 23). Next, a spectrum envelope as shown by a solid line B in FIG. 6 connecting the peaks a, b, and c of the peaks of the spectrum envelope with a straight line is derived, and the spectrum envelope of the solid line B is subtracted from the spectrum envelope of the broken line A. Thus, a frequency characteristic as shown in FIG. 7 is obtained, and a filter coefficient for obtaining the frequency characteristic is calculated (block 24). Then, the filter coefficients are stored in the formant filter coefficient memory 13 (FIG. 3).

Also, in parallel with or after the derivation of the frequency characteristic (block 24), in contrast to the derivation, the frequency envelope shown in FIG. 8 is obtained by subtracting the spectrum envelope of the dashed line A from the spectrum envelope of the solid line B. A characteristic is obtained, and an inverse filter coefficient for obtaining the frequency characteristic is calculated (block 25). After calculating the inverse filter coefficient, the coefficient is stored in the inverse filter coefficient memory 28 shown in FIG. 5, and the frequency characteristic of the pre-filter 27 is set as shown in FIG. Then, the original sound is input again (block 26), and the original sound is passed through the pre-filter 27, whereby the solid line B (6
Waveform memory which forms the musical tone waveform data having the spectrum envelope shown in FIG.
11 (FIG. 3) (block 29). Note that these processes may use a dedicated hardware circuit, but it is convenient to use a computer.

Specific Embodiment Next, a specific embodiment of the present invention in which the above-described tone generator is applied to an electronic musical instrument will be described. FIG. 1 is a block diagram showing the entirety of the electronic musical instrument. The electronic musical instrument has a waveform memory 31, a formant filter 32, and a formant filter coefficient memory 33 which are the same as those described above.

As described above, the waveform memory 31 stores a musical tone waveform data including a plurality of sampling data having a spectral envelope in which the formant characteristics of a musical tone to be generated are smoothed. For each tone composed of instrument sounds or human voices, etc., and for each key range for improving sound quality, for example, a half octave,
A plurality of sets of musical tone waveform data are stored for each octave. Further, the tone waveform data for each tone color and for each key range are stored in a plurality of frames, for example, as shown in FIG. 2, the first tone waveform data corresponding to the sampling data in the attack portion of the tone.
Are stored is divided into second through the N-th frame FR ₂ ～FR _N corresponding to a predetermined length of the sampling data in the sustain portion and the decay portion of the frame FR ₁ the same tone.

The formant filter 32 adds the formant frequency characteristic to the tone waveform data from the waveform memory 31 and outputs the same, as in the case described above. The frequency characteristic is controlled by the filter coefficient from the formant filter coefficient memory 33. It has become. The formant filter coefficient memory 33 stores a plurality of sets of filter coefficients corresponding to the tone waveform data for each timbre, key range, and frame.

The waveform memory 31 and the formant filter coefficient memory 33 are controlled by a key switch circuit 34, a key press detection circuit 35, an address generation circuit 36, a timbre selection switch circuit 37, and a selected timbre detection circuit 38. The musical tone waveform data and the filter coefficient are read out according to the key pressed on the keyboard and the selected timbre in the timbre selection operator group.

The key switch circuit 34 includes a plurality of key switches respectively corresponding to a plurality of keys on the keyboard, and each key switch opens and closes in response to a key pressed or released on the keyboard. Key press detection circuit 35
Is connected to the key switch circuit 34 to detect the opening / closing of each key switch, thereby detecting the key press / release of each key on the keyboard, and generates a key code KC representing the key pressed in accordance with each detection. It outputs to the address generation circuit 36 and also outputs a key-on signal KON indicating a key press. Also, key code K
C is waveform memory 31 and formant filter coefficient memory 3.
3 is also supplied, and specifies the range to which the depressed key belongs. The address generation circuit 36 outputs an address signal ADR that changes at a rate corresponding to a key pitch frequency pressed on the keyboard based on the supply key code KC to the waveform memory 31 and stores the address signal ADR in the memory 31. controls the reading of the musical tone waveform data are to control the reading of filter coefficients stored frame signal FR representing each frame FR ₁ ~FR _N tone to formant filter coefficients and outputs to the memory 33 the memory 33 . In this case, the address signal ADR is used to designate a storage address of each sampling data of the musical tone waveform data composed of the first through the N-th frame FR ₁ ～FR _N have been stored for each tone color and key range,
The upper bits represent the frame and the lower bits represent the storage address of the sampling data in each frame. It should be noted that the frame signal is supplied to the address generation circuit 36.
In order to switch the FR in accordance with the selected timbre, a selected timbre signal TC described later is supplied from the selected timbre detection circuit.

The tone color selection switch circuit 37 includes a plurality of tone color selection switches respectively corresponding to the plurality of tone color selection operators, and each tone color selection switch is opened and closed according to the operation of the tone color selection operator. The selected timbre detection circuit 38 is connected to the timbre selection switch circuit 37, and detects a selected timbre by detecting opening / closing of each timbre selection switch.The selected timbre detection circuit 38 generates a selected timbre signal TC representing the selected timbre according to the detection. Output to the memory 31 and the formant filter coefficient memory 33,
Is instructed to select the tone.

The output of the formant filter 32 is connected to one input of a multiplier 41, and the other input of the multiplier 41 is connected to an envelope signal generating circuit. The envelope signal generation circuit 42 is a key-on signal from the key press detection circuit 35
It outputs an amplitude envelope signal ENV that controls the amplitude envelope of the musical tone according to KON. The envelope signal generation circuit 42 has a selected tone color signal TC and a key code KC.
Is also supplied, and the amplitude envelope waveform is changed and controlled according to the selected timbre and the key range by the amplitude envelope signal ENV.

The output of the multiplication key 41 is connected to the D / A converter 43,
The converter 43 converts the digital signal from the multiplier 41 into an analog signal and outputs the analog signal to the sound system 44. The sound system 44 includes an amplifier, a speaker, and the like, and emits a musical tone corresponding to the supplied analog signal.

In the electronic musical instrument configured as described above, when any key is pressed on the keyboard, the address generation circuit 36
In cooperation with the key switch circuit 34 and the key press detection circuit 35, an address signal ADR that changes at a rate corresponding to the pitch frequency of the key pressed is supplied to the waveform memory 31, and the first to Nth
Controlling the sequential reading of sampling data constituting the tone waveform data of each frame while sequentially switching the frame FR ₁ ~FR _N. In the case where said key depression selected timbre be of sustained system such violin last long after reading the musical tone waveform data of the N frame FR _N, the second to N-th frame FR ₂ ～FR _N One or more repetitive readings of the musical tone waveform data are controlled. As a result, the waveform memory 31
Are the selected timbre specified by the selected timbre signal TC from the selected timbre detection circuit 38 and the key code from the keypress detection circuit 35
First to Nth frames related to the key range specified by KC
The tone waveform data of FR _{1 to} FR _N is output to the formant filter 32 at a rate corresponding to the pitch frequency of the key that has been pressed. As a result, the musical tone waveform data supplied to the formant filter 32, excluding the formant characteristics, well expresses the characteristic color of the musical tone that changes with the timbre, the key range, and the passage of time.

On the other hand, the formant filter coefficient memory 33 is controlled by the selected timbre signal TC from the selected timbre detection circuit 38, the key code KC from the key press detection circuit 35, and the frame signal FR from the address generation circuit 36. Since the filter coefficient corresponding to the data is read out and output to the formant filter 32 in conjunction with the data reading, the filter 32 has the frequency characteristic (corresponding to the formant characteristic of the tone to be generated) corresponding to the supplied musical tone waveform data. A frequency characteristic corresponding to the difference between the spectrum envelope and the spectrum envelope of the musical tone waveform data read from the waveform memory 31 is provided. As a result, the spectral envelope of the musical sound waveform data output from the formant filter 32 is independent of the pitch frequency of the key pressed on the keyboard, and corresponds to the formant of the musical sound to be generated and is fixed. Become.

Then, the tone waveform data output from the formant filter 31 is supplied to a multiplier 41, and the multiplier 41 cooperates with the envelope signal generating circuit 42 to generate an amplitude envelope corresponding to the selected timbre and the key range. Give.
Next, the tone waveform data to which the amplitude envelope has been added in this manner is converted into an analog signal by the D / A converter 43 and supplied to the sound system 44.
From 44, a musical tone with an amplitude envelope is generated. Further, when the key being depressed on the keyboard is released,
With the disappearance of the key-on signal KON, the envelope signal generating circuit 42 outputs the amplitude envelope signal ENV which is gradually attenuated, so that the musical tone being generated is gradually attenuated and the sound generation is stopped.

As can be understood from the above description, according to the above embodiment, a musical tone to be generated in the waveform memory 31 such as a trumpet, a violin or other natural musical instrument or a musical tone having a spectral envelope in which a formant of a musical tone collected from a human voice is smoothed By storing the waveform data, the musical tone waveform data which well represents the characteristic of the musical tone excluding the formant characteristic is generated, and the musical tone is generated irrespective of the pitch frequency by the filter means including the formant filter 32 and the formant filter coefficient memory 33. Since a unique formant is added to the musical sound waveform data, it is possible to produce a musical tone having a good fixed formant characteristic even if the capacity of the waveform memory 31 is reduced while taking advantage of the waveform memory system.

Modifications In the above embodiment, the waveform memory 31 and the formant filter coefficient memory 33 store the musical tone waveform data and the filter coefficient respectively for each of a plurality of key ranges and frames corresponding to timbres. If a slight decrease in tone quality is permitted, the tone waveform data and filter coefficients for a common frame or a single frame may be stored in the memories 31 and 33 in correspondence with the tone color. Also, the waveform data for one wave may be used as the musical sound waveform data, not for a plurality of waves. Further, conversely, if the capacity of the waveform memory 31 is allowed to increase for the purpose of improving the sound quality, the waveform memory 31 stores all continuous tone waveform data from the start to the end of the tone generation. You may.

Further, in the above-described embodiment, the sampling data of the musical sound waveform is stored as it is as the musical sound waveform data in the waveform memory 31, but the sampling data may be stored in a compressed form. By doing so, the capacity of the waveform memory 31 can be saved, and this is particularly effective when the entire musical tone waveform data described in the above-described modified example is stored in the waveform memory 31.

Further, in the above embodiment, the reading of the filter coefficients stored in the formant filter coefficient memory 33 is switched by the frame signal from the address generating circuit 36. The switching of the filter coefficient may be controlled in accordance with the passage of time by the means. This makes it possible to easily change the fixed formant characteristic with time even when storing one waveform of musical tone waveform data or all musical tone waveform data in the waveform memory 31 as described above.

[Brief description of the drawings]

FIG. 1 is an overall block diagram of an electronic musical instrument to which a tone generator according to one embodiment of the present invention is applied, FIG. 2 is a waveform diagram showing an example of musical tone waveform data stored in a waveform memory of FIG.
FIG. 3 is a block diagram for explaining the basic principle of the present invention, FIGS. 4 and 5 are block diagrams for explaining a method of forming data used in FIGS. 1 and 3, and FIG. FIG. 7 is a spectrum envelope waveform diagram for explaining the operation of FIGS. 3 to 5, and FIGS. 7 and 8 are FIGS.
FIG. 3 is a frequency characteristic diagram for explaining the operation of FIG. Description of reference numerals 11,31 …… Waveform memory, 12,32 …… Formant filter, 13,33 …… Formant filter coefficient memory, 34…
... Key switch circuit, 35 ... Key press detection circuit, 36 ... Address generation circuit.

Claims (1)

(57) [Claims] 1. A waveform memory for storing tone waveform data obtained by processing waveform data representing an original sound so that its spectral envelope becomes smooth, and a frequency of a tone to be generated from the tone waveform data stored in the waveform memory. Reading means for reading and outputting at a rate corresponding to the following; and inputting the musical tone waveform data output from the reading means, and inputting the musical tone waveform data into the spectral envelope of the waveform data before the processing and the spectrum of the musical tone waveform data after the processing. And a filter means for giving a frequency characteristic corresponding to a difference from the envelope.