JPH10198381A - Music generator - Google Patents

Abstract

(57) [Summary] To provide a musical sound generation device capable of generating a high-quality musical sound waveform in which the tone color naturally changes in accordance with a reproduction pitch and generation of aliasing noise is suppressed. SOLUTION: Waveform data WD corresponding to various timbres and an interpolation table are stored in a waveform ROM 6, an interpolation table suitable for a reproduction pitch is selected, and the interpolation table is read from the interpolation table in accordance with a decimal part address of an address signal AD. Address signal A based on a plurality of interpolation coefficients IPC
Waveform data W read according to the integer part address of D
Since the D group is polynomial-interpolated, it is possible to generate a musical tone waveform to which a natural tone color change corresponding to the reproduction pitch is given, and to suppress the generation of aliasing noise.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a tone generator suitable for use in, for example, electronic musical instruments.

[0002]

2. Description of the Related Art Heretofore, there has been known a so-called waveform memory reading type tone generating apparatus which stores PCM waveform data in a memory and reads out the PCM waveform data at a rate corresponding to a tone pitch to generate a tone. . In this type of apparatus, a read address is generated by sequentially accumulating rate data proportional to a ratio of a pitch of a musical tone to be generated (hereinafter, referred to as a reproduction pitch) to an original tone pitch, and a memory corresponding to the read address is generated. The PCM waveform data stored in the PCM is read out to generate a musical tone having a desired pitch.

[0003] In the musical tone generating apparatus of the waveform memory reading method, since the rate data has a decimal part for reasons such as frequency accuracy, the read address is inevitably composed of an integer part and a decimal part. On the other hand, since the PCM waveform data stored in the memory is addressed by an integer, the waveform data corresponding to the decimal part address is generated by interpolating the waveform data corresponding to the integer part address.

[0004]

The above-mentioned conventional tone generating apparatus is excellent in that the original sound is faithfully reproduced as it is, but when the readout rate is changed and the reproduction pitch is varied as described above. In this case, the generated waveform becomes uniform. In other words, most natural musical instruments have a waveform spectrum that changes according to the pitch based on the movable formant characteristics of the sounding body of the musical instrument. In the musical sound generation device, the readout waveform is uniformed irrespective of the pitch, so that the further the reproduction pitch is away from the original sound, the more the unnaturalness increases. In addition, since the above-described interpolation calculation performed at the time of pitch conversion is performed similarly for all waveforms, there is a problem that aliasing noise is remarkably displayed depending on the waveform.

Accordingly, the present invention has been made in view of such circumstances, and it is possible to generate a high-quality musical sound waveform in which the tone color naturally changes according to the reproduction pitch and the generation of aliasing noise is suppressed. It is an object of the present invention to provide a musical sound generating device capable of performing the above.

[0006]

In order to achieve the above object, according to the present invention, a waveform storage means for storing waveform data and an interpolation means for storing a plurality of interpolation coefficients for performing waveform interpolation on the waveform data. Coefficient storage means, and address generation means for generating a first address for reading the waveform data and a second address for reading the interpolation coefficient, the first address being changed in accordance with the pitch of a musical tone to be generated; Reading the waveform data from the waveform storage means in accordance with the first address generated by the address generation means, and storing the interpolation coefficient adapted to the pitch of the musical tone to be generated based on the second address. Reading means for selecting and reading from a plurality of interpolation coefficients stored in the means, and reading the waveform data read by the reading means. Is characterized by comprising a pitch conversion means for converting the pitch of the waveform data to the waveform interpolation by the read interpolation coefficients, the pitch of a tone to be generated.

According to the second aspect of the present invention, the interpolation coefficient storage means stores a plurality of interpolation coefficients in a storage medium common to the waveform storage means.

According to the third aspect of the present invention, there is provided a waveform storage means for storing waveform data, an interpolation table storage means for storing, for each waveform type, an interpolation table storing interpolation coefficients for performing waveform interpolation of the waveform data, Address generating means for generating a first address for reading out the waveform data and a second address for reading out the interpolation coefficient, the address changing means corresponding to a pitch of a musical tone to be generated; The waveform data is read from the waveform storage means in accordance with the first address at which the waveform data is generated, and an interpolation table adapted to the waveform type of the read waveform data is stored in the interpolation table storage means for each waveform type. A table is selected from among a plurality of interpolation coefficients constituting a selected interpolation table and should be generated based on the second address. Reading means for selecting and reading an interpolation coefficient adapted to the pitch of a musical tone; and performing waveform interpolation on the waveform data read by the reading means using the interpolation coefficient read by the reading means to determine the pitch of the waveform data. And a pitch converting means for converting the pitch into a tone to be generated.

According to the present invention, first and second addresses that change in accordance with the pitch of a musical tone to be generated are generated, and the first and second addresses are generated.
The waveform data is read out from the waveform storage means in accordance with the address of the tone signal, and an interpolation coefficient adapted to the pitch of a musical tone to be generated is selected and read out from the interpolation coefficient storage means based on the second address. Performs waveform interpolation based on the data and the interpolation coefficient, and calculates the pitch of the waveform data.
Converts to the pitch of the musical tone to be generated. In other words, since pitch conversion is performed by waveform interpolation adapted to the pitch of the musical tone to be generated, a natural tone color change corresponding to the playback pitch is realized, and a high-quality musical tone waveform with suppressed generation of aliasing noise is generated. It becomes possible to do.

[0010]

BEST MODE FOR CARRYING OUT THE INVENTION A musical sound generating apparatus according to the present invention can be applied to an electronic musical instrument, a computer music apparatus using a personal computer, and the like. In the following, before describing the examples, after describing the principles of the present invention,
Next, an electronic musical instrument according to an embodiment of the present invention will be described as an example with reference to the drawings.

A. The principle of the present invention will be described below with reference to FIGS. First, FIG. 5 is a diagram for explaining the principle of pitch conversion in the waveform memory readout system which is a premise of the present invention. Considering the frequency conversion on the time axis, it simply reads out the original waveform data stored in the memory at a desired rate.However, in order to evaluate the interpolation filter on the frequency axis, it is interpreted in terms of signal processing. Go.

Now, the waveform data obtained by sampling the original sound at a predetermined frequency is converted to a 3/4 frequency (reproduction pitch).
And reading from the waveform memory. here,
“To increase the frequency of the waveform data by 3/4” can be interpreted as “converting the sampling frequency of the waveform data to 4/3 and outputting at the original sampling frequency”. After interpolating four times, the time axis is thinned out to one third and the time axis is made three fourths. "

This relationship will be described with reference to FIG. FIG.
(A) shows a waveform W1 of waveform data obtained by sampling an original sound at a predetermined frequency and its spectra OWS, OWS.
It is a figure showing 1-OWS4. The spectrum OWS is a spectrum of the waveform W of the original sound, and OWS1 to OWS4 are mirror images of the spectrum OWS generated by sampling, and are repeatedly generated at angular frequencies of 2π / T. IPF
Represents the characteristic of the interpolation filter. When the waveform W1 shown in FIG. 5A is quadrupled by the interpolation filter IPF, as shown in FIG. 5B, the waveform W2 has four times the number of sampling points. It will occur repeatedly at every angular frequency 8π / T.

In the waveform W3 obtained by thinning out the sample points of the quadrupled waveform W2 to 1/3, the angular frequency 2
Spectra OW every π / T '(where T' = 3T / 4)
S occurs. Here, “T ′ = 3T / 4” represents the converted sampling period. Next, when the time axis of the waveform W3 in FIG. 9C is expanded by 4/3 times to restore the converted sampling period, the expansion on the time axis corresponds to the compression on the frequency axis. Therefore, the spectrum distribution shown in FIG. That is, when the pitch of the original sound waveform W1 is reduced to 3/4 times for reproduction, the spectrum of the reproduced waveform is uniformed only by narrowing the band. This is the cause of the unnaturalness of the conventional readout waveform described above.

Therefore, in the present invention, an interpolation filter that does not narrow the band after frequency conversion is used. The pitch conversion when the interpolation filter IPF1 is used will be described with reference to FIG. FIGS. 6 (a) to 6 (d) show, similarly to FIGS. 5 (a) to 5 (d) described above, "after interpolating the waveform data four times, thinning it out to 1/3 and reducing the time axis to 3/4 times. FIG. In this case, the interpolation filter IP
The pitch-converted waveform data W4 according to the characteristics of F1
As is clear from the spectrum of FIG. 6 (d), the high-frequency component HFS appears in addition to the basic spectrum in the half band of the sampling frequency. In other words, the tone is not an unnaturally tone-limited tone, but rather a harmonic component (harmonic component).

Next, FIGS. 7A to 7C show a process of "doubling the time axis after interpolating the waveform data twice," that is, reproducing the original sound pitch lowered by one octave. Is illustrated, in which case the interpolation filter I
PF2 is used. Since the transition region of the interpolation filter IPF2 is gentle, many harmonic components appear. As described above, when performing the pitch conversion that lowers the reproduction pitch from the original sound pitch, it is possible to prevent the bandwidth of the reproduction waveform from being narrowed by using an interpolation filter having characteristics according to the frequency of the reproduction pitch. I understand.

Next, an example of pitch conversion for increasing the reproduction pitch from the original sound pitch will be described with reference to FIG.
FIGS. 8A to 8D show a process of “interpolating the waveform data by a factor of two and then decimating the waveform data by a factor of three to increase the time axis by a factor of two.”
In other words, the figure illustrates a process of reproducing the original sound by increasing the pitch of the original sound by 3/2 times. In this case, the interpolation filter IPF3 is used. The interpolation filter IPF3 has a filter characteristic of removing the original waveform spectrum in order to prevent a synonymous phenomenon.

The spectrum of the waveform which has been thinned out to one third after the double interpolation (see FIG. 3 (c)) does not overlap based on the characteristics of the interpolation filter IPF3. Then, when this is multiplied by 2/3 on the time axis, the spectrum is expanded to 3/2 times as shown in FIG.

As described above, in the pitch conversion in the waveform memory reading method, it is necessary to use an interpolation filter having various characteristics corresponding to the frequency of the reproduction pitch in order to eliminate the above-mentioned nickname phenomenon and the unnaturalness of the waveform spectrum. Based on such knowledge, the present invention controls the characteristics of the interpolation filter in consideration of the “relationship between the original sound pitch and the reproduction pitch” and the “spectral characteristics of the original waveform” to provide a natural sound corresponding to the reproduction pitch. This is to generate a high-quality musical sound waveform that realizes tone change and suppresses generation of aliasing noise.

B. 1. Embodiment (1) Overall Configuration Next, with reference to FIG. 1, the overall configuration of an electronic musical instrument having a musical tone generating device that performs pitch conversion based on the above-described principle will be described. In FIG. 1, reference numeral 1 denotes a keyboard, which detects a key press / release operation for each key and generates performance information such as a key-on / key-off signal, a key code, and a key-press speed signal corresponding to the detected key release / key operation. . Reference numeral 2 denotes a panel switch unit provided on the operation panel, and is composed of various operation switches such as a tone color selection switch and a power switch. Reference numeral 3 denotes a CPU for controlling each section of the musical instrument. The CPU 3 mainly supplies performance information supplied from the keyboard 1 via the bus B and various operation switch signals supplied from the panel switch section 2 to a sound source 7 (described later). A control signal Dc for instructing tone control is supplied.

A ROM 4 stores various control programs loaded into the CPU 4 and various data used in these programs. Reference numeral 5 denotes a RAM that is used as a work area of the CPU 3 and temporarily stores various calculation results, register / flag data, and the like. 6 is waveform R
OM, which stores waveform data WD corresponding to various timbres and a plurality of interpolation tables. The interpolation table is made up of an interpolation coefficient IPC group for realizing a spectral characteristic of the original sound and a filter characteristic adapted to the reproduction pitch, and is used for the above-described interpolation calculation at the time of pitch conversion.

The sound source 7 is constituted by a waveform memory reading method, and has a plurality of simultaneous sounding channels so as to sound polyphonically. The sound source 7 outputs an address signal AD for reading out predetermined tone color waveform data and an interpolation coefficient from the waveform ROM 6 based on a control signal Dc corresponding to performance information generated by the CPU 4. Also, the sound source 7
In order to generate a tone corresponding to the tone parameter indicated by the control signal Dc, the waveform ROM 6 is generated in accordance with the address signal AD.
Data WD and interpolation coefficient IPC read from
To generate and output a tone signal OUT. 8 is D /
A tone converter which is an A converter and is output from a tone generator 7
Is converted to an analog musical tone waveform MW and output. 9
Is an amplifier that amplifies the level of the musical tone waveform MW and emits the sound from the speaker 10.

(2) Configuration of the Sound Source 7 Next, the configuration of the sound source 7 will be described with reference to FIG.
An interface circuit 11 exchanges data with the CPU 3. The control signal Dc supplied from the CPU 3 via the interface circuit 11 is supplied to the envelope generator 12, the waveform generator 13, the filter 14, and the multiplier 15. The envelope generator 12
An amplitude envelope from the rise of the musical tone to its decay is generated and supplied to the multiplier 15.

The waveform generator 13 is provided with the interface circuit 1
1; a tone parameter designated by the control signal Dc supplied from the CPU 3 via the CPU 1; namely, a start address indicating the start point of the loop playback section, an end address indicating the end point of the section, loop length data, and a tone sound to be generated. While the waveform data WD of the corresponding timbre is loop-reproduced from the waveform ROM 6 in accordance with the frequency data representing the height, etc.
The waveform ROM 6 stores the waveform data WD to be loop-reproduced.
The pitch is converted by an interpolation operation based on the interpolation filter characteristic corresponding to the interpolation coefficient IPC read out from the input and output.

The filter 14 adds to the output of the waveform generator 13 an effect set in accordance with the tone color, for example, an effect such as reverb. The multiplier 15 multiplies the waveform data WD added with this effect by an amplitude envelope to form a tone signal OUT. The output section 16 serially converts the parallel tone signal OUT to be input to the D / A converter 8 (see FIG. 1) at the next stage and outputs the converted signal.
In a normal electronic musical instrument, time division processing is performed for the number of simultaneous sounds. In this case, the output unit 16 accumulates the tone signal OUT for each sound channel sequentially inputted for each sampling period, and then accumulates the parallel / sound signals. In this case, serial conversion is performed.

(3) Configuration of Waveform Generator 13 Next, the configuration of the waveform generator 13 will be described in detail with reference to FIG. The waveform generation unit 13 includes a waveform reading unit 13a including components 21 to 32 and an interpolation calculation unit 13b including components 33 to 38.

In FIG. 1, reference numerals 21 to 23 denote flip-flop circuits (hereinafter abbreviated as FF circuits), which control signals Dc supplied from the CPU 3 via the interface circuit 11 described above. Of these, the tone parameters necessary for reading the waveform are temporarily stored. Here, the FF circuit 21 has a loop length LOOPLGT representing a loop reproduction section, the FF circuit 22 has an end address ENDADRS, and the FF circuit 23 has frequency data FR representing a ratio between a reproduction pitch and an original sound pitch.
RATIO is temporarily stored. Note that these F
The clocks of the F circuits 21 to 23 are
This occurs when writing data.

Reference numeral 24 denotes an adder for accumulating the frequency data FRRATIO output from the FF circuit 23 and an output of a selector 27 described later. Reference numeral 25 denotes an FF circuit for holding the output of the adder 24. In the FF circuit 25, a start address is set from the CPU 3 at the start of sound generation, and thereafter, a waveform read address for increasing the start address at a rate corresponding to the frequency data FRRATIO is generated.

Reference numeral 28 denotes a comparison circuit which supplies "High (1)" to the gate circuit 29 when the waveform read address generated by the FF circuit 25 matches the end address ENDADRS set in the FF circuit 22. .
As a result, the gate circuit 29 supplies the loop length LOOPLGT set in the FF circuit 21 to the subtractor 26. The subtractor 26 subtracts the waveform read address generated by the FF circuit 25 by the loop length LOOPLGT, and returns the waveform read address to the start address. The output of the subtractor 26 is supplied to the input terminal A of the selector 27, and is also supplied to the adder 30 and the selector 32.

The output of the subtractor 26 has a data width of 30 bits, the upper 20 bits forming the integer part address and the lower 10 bits forming the decimal part address. The upper 20 bits of the integer part address are supplied to the adder 30, while the lower 10 bits of the decimal part address are supplied to the input terminal B of the selector 32. When the selector signal ADST is “High (1)”, the selector 27
In the other cases, a waveform read address input to the input terminal A is selected and output.

In the adder 30, the signal A is added to the integer part address.
DBIAS is added and supplied to the input terminal A of the selector 31. The signal ADBIAS applies a bias of “−1”, “+0”, “+1”, and “+2” to the integer part address. As a result, the waveform sample corresponding to the integer part address and its surroundings (“−1”, “+1”, and “+”)
2 "is designated. Interpolation calculation described later is performed using these four waveform samples.

When the selector signal ADSEL is "Low"
At the time of (0) ", the selectors 31 and 32 supply the integer part address to which the above-mentioned four bias values are added to the waveform ROM 6 as the address signal AD. On the other hand, when the selector signal ADSEL is “High (1)”, the selector 31 outputs an interpolation coefficient selection signal having a 6-bit width and the upper 4 bits of the frequency data FRRATIO. In this interpolation coefficient selection signal, the upper 4 bits are “1111” and the following 2
A signal INT whose bits indicate switching of the interpolation table
It consists of NO. The selector 32 selects and outputs the decimal part address when the selector signal ADSEL is “High (1)”.

Therefore, if the selector signal ADSEL is "High (1)", a corresponding interpolation table is selected from among the interpolation tables of various filter characteristics stored in the waveform ROM 6 according to the interpolation coefficient selection signal, and the interpolation table is selected. The interpolation coefficient IPC corresponding to the decimal part address is read from the table. Thus, the components 21 to 3
The waveform readout unit composed of 2 generates a waveform readout address of the loop reproduction section according to the instruction of the CPU 3, and reads out four waveform samples based on the integer part address and the bias-added integer part address.
The interpolation coefficient IPC corresponding to the decimal part address is read from the interpolation table selected by the interpolation coefficient selection signal.

The configuration 33 of the interpolation operation unit 13b is an FF circuit for temporarily storing waveform data WD read from the waveform ROM 6 in accordance with the address signal AD. Reference numeral 34 denotes an FF circuit which holds four interpolation coefficients IPC read from the interpolation table corresponding to the above-mentioned decimal part address. A coefficient multiplier 35 multiplies the waveform data WD by an interpolation coefficient IPC and outputs the result. An FF circuit 36 holds the multiplied output. 37 is an adder 3
The accumulator 7a includes an FF circuit 37b and a gate circuit 37c, and accumulates the output of the FF circuit 36 to generate one waveform sample.

This accumulation is performed in synchronization with the gate signal IPST supplied from the CPU 3 side.
Execute Y = IPC (−1) × WD (−1) + IPC (0) × WD (0) + IPC (+1) × WD (+1) + IPC (+2) × WD (+2) (1) where IPC ( -1) to IPC (+2) represent interpolation coefficients IPC respectively corresponding to the four waveform data WD (-1) to WD (+2) described above. The accumulated value Y obtained by the third-order polynomial interpolation operation is taken into the FF circuit 38 and output to the next stage as a tone signal OUT.

(4) Operation of Waveform Generator 13 The operation of the waveform generator 13 having the above configuration will be described with reference to the time chart of FIG. Here, for simplification of the description, the description will be made as a single sound generation process. First, in FIG. 4, CLK is a basic clock and CK (FF2
5) is an address increment clock obtained by dividing the basic clock CLK by eight. In this description of the operation, since the tone processing is performed, the address increment clock coincides with the sampling cycle.

In the state where predetermined values are set in the FF circuits 21 and 22 and zero data is set in the FF circuit 23, when a note-on is instructed from the CPU 3, the select signal ADST is shown. And the CPU 3 stores the start address A in the FF circuit 25.
Set DD. Subsequently, when the frequency data FRRATIO corresponding to the key code is set in the FF circuit 23, the waveform read address stored in the FF circuit 25 by the adder 24 is sequentially changed from the start address ADD to the rate corresponding to the frequency data FRRATIO. Will be added.

The output of the FF circuit 25 is the end address END set in the FF circuit 22 by the comparison circuit 28.
When the output of the FF circuit 25 is compared with the ADRS and becomes equal to or greater than the end address ENDADRS, the comparison circuit 28 generates "High (1)". Then, the gate circuit 29 inputs the loop length LOOPLGT set in the FF circuit 21 to the subtractor 26.
Thus, the waveform read address stored in the FF circuit 25 is the start address AD which is the loop start position.
Returned to D.

In the process of generating the waveform read address in the loop reproduction section in this manner, the integer part address of the waveform read address stored in the FF circuit 25 is input to the adder 30, and "-1", "0", " +1 "and" +2 "are added and input to the input terminals A of the selectors 31 and 32, while the decimal part address of the waveform read address is input to the input terminal B of the selector 32. The input terminal B of the selector 31 receives the interpolation coefficient selection signal and the upper 4 bits of the frequency data FRRATIO.

As shown in the time chart of FIG. 4, the selectors 31 and 32 alternately output an integer part address, a decimal part address, an interpolation coefficient selection signal associated therewith, and frequency data FRRATI in accordance with the signal ADSEL.
O is supplied to the waveform ROM 6 as an address signal AD. As a result, first, the waveform data WD (-1) corresponding to the integer part address "-1" is read from the waveform ROM 6, and then the interpolation table in the waveform ROM 6 is designated according to the interpolation coefficient selection signal, and the decimal number is designated therefrom. The interpolation coefficient IPC (-1) corresponding to the section address is read. Thereafter, similarly, the waveform data WD and the interpolation coefficient IPC are alternately read.

The waveform data WD (-1) to WD (+2) thus read from the waveform ROM 6 and the interpolation coefficient I
The PC (-1) to IPC (+2) are subjected to a third-order polynomial interpolation operation which is multiplied by each other in the interpolation operation unit 13b and then accumulated, and the accumulation result Y is converted to the musical tone signal OU.
Output as T.

As described above, according to this embodiment,
Waveform data WD and interpolation tables corresponding to various timbres are stored in the waveform ROM 6, an interpolation table suitable for the reproduction pitch is selected, and a plurality of interpolation coefficients read out from the interpolation table in accordance with the decimal part address of the address signal AD. Since the waveform data WD group read in accordance with the integer part address of the address signal AD is polynomial-interpolated based on the IPC, a tone waveform to which a natural tone change corresponding to the reproduction pitch is given is generated, and generation of aliasing noise is suppressed. It is possible to do.

In this embodiment, as described above, the interpolation table is selected according to the ratio between the original sound pitch and the reproduction pitch. In addition, the interpolation table is selected according to the designated tone color. It may be specified. That is, while specifying the type of the waveform data WD to be read out by operating the timbre switch, an interpolation table group adapted to the spectral characteristic of the waveform data WD is selected, and an interpolation table adapted to the reproduction pitch is selected from the interpolation table group. In this embodiment, a table is specified, and the interpolation coefficient IPC is read from the table according to the decimal part address. It is also possible to adopt a mode in which interpolation coefficients having different interpolation characteristics are interpolated with each other to derive an interpolation coefficient having a new interpolation characteristic.

[0044]

According to the present invention, since the pitch conversion by the waveform interpolation adapted to the pitch of the musical tone to be generated or the waveform interpolation adapted to the spectral characteristics of the original sound is performed, a natural tone corresponding to the reproduction pitch is performed. It is possible to generate a high-quality musical sound waveform that realizes a change and suppresses generation of aliasing noise.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an overall configuration of an embodiment according to the present invention.

FIG. 2 is a block diagram showing a configuration of a sound source 7;

FIG. 3 is a block diagram illustrating a configuration of a waveform generation unit 13.

FIG. 4 is a time chart for explaining the operation of the waveform generation unit 13;

FIG. 5 is a diagram for explaining the principle of the present invention.

FIG. 6 is a diagram for explaining the principle of the present invention.

FIG. 7 is a diagram for explaining the principle of the present invention.

FIG. 8 is a diagram for explaining the principle of the present invention.

[Description of Signs] 6 Waveform ROM (waveform storage means, interpolation coefficient storage means) 13 Waveform generation unit 13a Waveform readout unit (address generation means, readout means) 13b Interpolation calculation unit (pitch conversion means)

Claims (3)

[Claims] 1. A waveform storage means for storing waveform data, an interpolation coefficient storage means for storing a plurality of interpolation coefficients for waveform-interpolating the waveform data, and an address changing corresponding to a pitch of a musical tone to be generated. Address generating means for generating a first address from which the waveform data is read and a second address for reading the interpolation coefficient; waveform data from the waveform storage means in accordance with the first address generated by the address generating means And read
Reading means for selecting and reading out from among a plurality of interpolation coefficients stored in the interpolation coefficient storage means, an interpolation coefficient adapted to a pitch of a musical tone to be generated based on the second address; Pitch converting means for performing waveform interpolation on the read waveform data with the interpolation coefficient read by the reading means and converting the pitch of the waveform data into the pitch of a musical tone to be generated. Musical tone generator. 2. The tone generating apparatus according to claim 1, wherein said interpolation coefficient storage means stores a plurality of interpolation coefficients in a storage medium common to said waveform storage means. 3. A waveform storage means for storing waveform data; an interpolation table storage means for storing, for each waveform type, an interpolation table storing interpolation coefficients for performing waveform interpolation on the waveform data; Address generating means for generating a first address for reading the waveform data and a second address for reading the interpolation coefficient, and a first address generated by the address generating means. While reading the waveform data from the waveform storage means in accordance with the above, while selecting an interpolation table corresponding to the waveform type of the read waveform data from the interpolation table for each waveform type stored in the interpolation table storage means, A complement adapted to the pitch of a musical tone to be generated based on the second address from among a plurality of interpolation coefficients constituting the selected interpolation table. Reading means for selecting and reading an interval coefficient, and performing waveform interpolation on the waveform data read by the reading means using the interpolation coefficient read by the reading means to obtain a pitch of the waveform data and a tone of a musical tone to be generated. A tone conversion device for converting pitch into high pitch.