JPH09325777A - Device and method for musical sound signal generation - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide the device and method for musical sound signal generation which can reduce the amount of waveform data and reproduce a sound including a percussion sound faithfully. SOLUTION: This device is equipped with a 1st storage means 10 which stores waveform data obtained by sampling a sound generated by removing percussion sound components from a source sound, i.e., a string sound at a specified sampling frequency and performing a loop process, a 2nd storage means 11 which stores waveform data obtained by sampling the percussion sound at a sampling frequency lower than the specified sampling frequency, a 1st read means 13 which reads the waveform data out of the 1st storage means 10 on a loop basis, a 2nd read means 14 which reads the waveform data out of the 2nd storage means, and a mixing means 18 which mixes the waveform data from the 1st and 2nd read means 13 and 14, and generates a musical sound signal according to the waveform data obtained by the mixing means.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a musical tone signal generating apparatus and a musical tone signal generating method suitable for, for example, an electronic musical instrument or a sound source board, and particularly when a piano tone color is generated by a pulse code modulation (PCM) method, a memory is stored. A technique for reducing the amount of waveform data to be processed.

[0002]

2. Description of the Related Art Conventionally, there is known a tone signal generator which employs a PCM system. In this tone signal generator, for example, waveform data obtained by pulse code modulating a signal of a natural musical instrument sound is stored in advance in a waveform memory. Then, in the tone signal generation operation, when the tone generation command is received, the waveform data is sequentially read from the waveform memory, and the tone signal is generated based on the read waveform data. In such a PCM type musical tone signal generator, various techniques are adopted in order to faithfully simulate a natural musical instrument sound. In the following, a conventional technique will be described by taking an acoustic piano (hereinafter simply referred to as "piano") as an example of one of the natural musical instruments.

Generally, when a keyboard of a piano is tapped, a striking sound is generated in addition to a sound based on the vibration of the strings. This striking sound is generated by a shock applied to the shelf board by a keystroke, a shock when a hammer strikes a string in response to the keystroke, and the like. This striking sound is particularly important for reproducing the piano sound, because it is audibly good to hear in the middle to high range of the piano. The hitting sound continues for a relatively long time (for example, 1 to 2 seconds) due to the reverberation of the housing. In order to simulate the characteristic of such a piano sound, a technique of generating a musical sound signal including a striking sound has been developed in a musical sound signal generator of the PCM system.

For example, in the first technique, a piano sound generated by keystroke (hereinafter referred to as "original sound") is recorded. Then, the recorded sound signal is subjected to pulse code modulation to create waveform data and stored in the waveform memory. Then, the waveform data is sequentially read from the waveform memory according to the tone generation command, and a musical tone signal is generated based on the read waveform data. In this technique, since the waveform data including the striking sound component is created, a sound close to a natural musical instrument sound can be obtained.

In the second technique, a signal of a sound obtained by removing a striking sound component from an original sound, that is, a sound based on only the vibration of a string (hereinafter referred to as "string sound") is pulse-code modulated to generate string sound waveform data. Is created. For example, a comb filter is used to remove the impact sound component. This string sound waveform data is stored in the first waveform memory. On the other hand, for example, a percussion sound generated by striking a piano shelf is recorded, and the percussion sound waveform data is created by pulse-code modulating the percussion sound. This impact sound waveform data is stored in the second waveform memory. Then, the string sound waveform data is read from the first waveform memory and the impact sound waveform data is read from the second waveform memory according to the sounding command, and mixed. Then, a musical tone signal is generated based on the mixed waveform data.

On the other hand, in many cases, the PCM type musical tone signal generator employs a waveform data storing / reading method called a loop method. In this loop method,
As shown in FIG. 6, only the waveform data of the rising part of the musical sound (hereinafter referred to as “attack part”) and the waveform data of a part of the subsequent steady part (hereinafter referred to as “loop part”) are stored in the waveform memory. To be done. Then, in response to the sounding command, first, the waveform data of the attack portion is read only once, and subsequently the waveform data of the loop portion is repeatedly read. According to this method, since it is not necessary to store the waveform data from the rising of the musical sound to the mute, there is an advantage that the amount of the waveform data to be stored can be reduced.

[0007]

However, there are the following problems in the case of generating the tone signal including the hitting sound by the loop method. As is well known, musical tones are composed of periodic waveforms, but striking sounds are not composed of periodic waveforms. If part of the waveform data of the hitting sound is repeatedly read and reproduced, the sound becomes a periodic waveform and becomes a sound different from the hitting sound. Therefore, when the sound including the hitting sound for a certain length is reproduced by the loop method, the waveform data before the hitting sound is muted (before the position A in FIG. 6) is present in all or part of the loop portion. If included, the striking sound changes to a heterogeneous sound as soon as the playback of the loop portion is started, resulting in a very unnatural sound. Therefore, the loop portion needs to be provided after the impact sound has been silenced (after position A in FIG. 6). This means that the amount of waveform data in the attack portion becomes large. Therefore, if the sound including the hitting sound is faithfully reproduced by the loop method, there is a limit in reducing the amount of waveform data.

Therefore, an object of the present invention is to provide a musical tone signal generating apparatus and a musical tone signal generating method capable of reducing the amount of waveform data and faithfully reproducing a sound including a beating sound. is there.

[0009]

In order to achieve the above-mentioned object, a musical tone signal generator of the present invention is obtained by sampling a sound in which a striking sound component is removed from an original sound, that is, a string sound at a predetermined sampling frequency. Waveform data obtained by performing loop processing, and a waveform obtained by sampling the impact sound removed from the original sound at a sampling frequency lower than the predetermined sampling frequency. Second storage means for storing data, first reading means for reading the waveform data from the first storage means in a loop system, and second reading means for reading the waveform data from the second storage means, The first reading means and the second
Mixing means for mixing each waveform data from the reading means, and is configured to generate a tone signal based on the waveform data mixed by the mixing means.

The waveform data stored in the first storage means is
For example, it can be created as follows. First,
The impact sound is extracted from the original sound. A low pass filter is used for this extraction. The part of the percussion sound component having a large amplitude exists on the frequency side lower than the fundamental tone in the middle to high range. Therefore, if the cutoff frequency of the filter is set to a frequency lower than the fundamental tone and the original sound is filtered, only the impact sound component is extracted. Next, the impact sound component is subtracted from the original sound. By this subtraction, the waveform data of the sound in which the striking sound component is removed from the original sound, that is, the string sound is obtained.

The above-described processing for obtaining the waveform data can be realized by sampling the original sound at a predetermined sampling frequency to obtain sampling data and digitally processing this on a computer, for example. As the sampling frequency, for example, 44.1 KHz can be used.

The waveform data of the string sound is stored in the first storage means after being subjected to loop processing. Here, the loop processing is processing the waveform data so as to conform to the loop method. More specifically, the loop process is a process of determining the attack part and the loop part based on the waveform data of the string sound. At this time, since the waveform data of the hitting sound is not included in the waveform data, the loop portion can be selected from any position immediately after the steady waveform. Therefore,
Since the amount of steady waveform data can be reduced,
The amount of waveform data to be stored can be reduced as a whole.

The waveform data stored in the second storage means is
For example, the waveform data of the impact sound component extracted as described above can be obtained by sampling at a frequency lower than the sampling frequency. This waveform data
For example, the waveform data obtained by sampling at 44.1 KHz can be obtained by digitally resampling on a computer, for example. The frequency component included in the impact sound is generally biased to a low frequency region. Therefore, even if the sampling data obtained by sampling at 44.1 KHz, for example, is resampled at about 10 KHz, the impact sound does not change in hearing.
The waveform data resampled in this way is stored in the second storage means as it is (without performing loop processing).

The striking sound continues for a relatively long time, but is much shorter than the sounding time of the original sound. Therefore, even if the whole waveform data of the hitting sound is stored in the second storage means, the amount of the waveform data of the hitting sound does not become enormous. The hitting sound waveform data may be stored in the second storage means by applying a damping envelope to the extent that the natural damping feeling is not impaired. With this configuration, it is possible to reduce the amount of hitting waveform data.

The effect of reducing the waveform data by the tone signal generator of the present invention will be specifically described in comparison with the conventional effect. It is now assumed that the first storage means stores 0.5 seconds worth of string sound waveform data and the second storage means stores 1 second worth of striking sound waveform data. The data length of each waveform data is 16 bits. In this case, the amount of waveform data stored in the first storage means is obtained by the following equation.

[0016]

[Formula 1] Waveform data amount = 44.1 KHz × 0.5 seconds × 1
6 bits = 352.8 K bits The amount of waveform data stored in the second storage means is obtained by the following formula.

## EQU00002 ## Waveform data amount = 10 KHz.times.1 second.times.16 bits = 160 Kbits Therefore, the amount of waveform data required in the tone signal generator of the present invention is 512.8 Kbits.

On the other hand, in the conventional tone signal generator,
When the waveform data for 1 second is stored in the storage means, the amount of the waveform data is obtained by the following formula.

## EQU00003 ## Waveform data amount = 44.1 KHz.times.1 second.times.16 bits = 705.6 Kbits Therefore, according to the musical tone signal generating apparatus of the present invention, 192.8 Kbits worth of waveform data is provided as compared with the conventional musical tone signal generating apparatus. The amount of can be reduced.

Further, in order to achieve the above-mentioned object, the musical tone signal generating method of the present invention loops to a waveform data obtained by sampling a sound obtained by removing a striking sound from an original sound, that is, a string sound at a predetermined sampling frequency. The processed waveform data is stored, and the waveform data obtained by sampling the impact sound at a sampling frequency lower than the predetermined sampling frequency is stored, and the waveform data of the sound in which the impact sound is removed from the original sound is stored. It is characterized in that it is read by a loop method, the waveform data of the striking sound is read, each of the read waveform data is mixed, and a tone signal is generated based on the mixed waveform data. With this tone signal generating method, the same effect as that of the tone signal generating device described above can be obtained.

[0019]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of a musical tone signal generating apparatus and a musical tone signal generating method of the present invention will be described in detail below with reference to the drawings. In the following embodiments, the case of generating a musical tone signal of a piano sound will be described, but the present invention is not limited to a piano sound, and for example, a musical tone signal of a sound including a harpsichord sound or other striking sound is generated. It is also applicable when doing.

First, the waveform data used in the present invention will be described. In the present embodiment, it is assumed that one tone color waveform data is composed of string sound waveform data and impact sound waveform data. The string sound waveform data and the impact sound waveform data are created by the following procedure.

First, the piano sound (original sound) generated by the keystroke with the strongest touch is recorded. FIG. 3 shows a spectrum of a recorded piano sound. Then, the recorded original sound is sampled at a sampling frequency of, for example, 44.1 KHz to obtain sampling data. By subjecting this sampling data to a predetermined process, string sound waveform data and impact sound waveform data are created. The creation of these waveform data can be performed by computer processing, for example.

First, the procedure for creating the string sound waveform data will be described. That is, the sampling data is filtered by a low-pass filter to obtain sampling data of a hitting sound component. The cutoff frequency of the filter may be, for example, a frequency slightly lower than the fundamental tone. FIG. 4 shows the spectrum of the impact sound obtained by this filtering process. Next, the sampling data of the impact sound is subtracted from the sampling data of the original sound. By this subtraction,
Sampling data of only the sound in which the striking sound component is removed from the original sound, that is, the string sound is obtained. The spectrum of this string sound is shown in FIG. Next, a loop process is performed on the sampling data of this string sound. At this time, a predetermined range can be set as the loop portion immediately after the steady waveform is obtained. After this loop processing is performed, it is stored in the string sound waveform memory 10.

Next, the creation of impact sound waveform data will be described. The sampling data of the hitting sound obtained above is resampled at a sampling frequency of 10 KHz, for example. The sampling data obtained by this resampling is stored in the impact sound waveform memory 11 as impact sound waveform data.

FIG. 1 is a block diagram showing the configuration of an embodiment of the musical tone signal generator of the present invention. The musical tone signal generator includes a string sound wave memory 10 and a percussion sound wave memory 1.
1, weight coefficient generators 12a and 12b, waveform reading circuit 1
3 and 14, multipliers 15 and 16, a filter 17, an adder 18, a multiplier 19, a parameter generator 20, and an envelope generator 21.

Tone data, tone range data, and touch data are supplied to the musical tone signal generator from the outside (for example, an operation panel of an electronic musical instrument, a keyboard device, a MIDI interface, a sequencer, etc.).

The tone color data is supplied from, for example, a tone color selection switch provided on an operation panel of an electronic musical instrument to which the present tone signal generator is applied. The range data and the touch data are supplied from the keyboard device of the electronic musical instrument to which the musical tone signal generator is applied. In this case, as the keyboard device, a two-contact type keyboard device in which each key has two key switches that are turned on at different pressing depths is used. Then, the data representing the key range to which the key number of the pressed key is applied is used as the range data. Also, the time from when one key switch is turned on to when the other key switch is turned on is measured in response to a keystroke, and data corresponding to the measured time is used as touch data. Each of these data is, for example, a CPU of an electronic musical instrument.
Is received from the operation panel or keyboard device and sent to the musical tone signal generator.

In the electronic musical instrument having the MIDI interface, the tone color number included in the received program change message can be used as tone color data. Further, the data representing the key range to which the note number included in the note-on message belongs can be used as the range data, and the velocity can be used as the touch data.
Each of these data is stored in the MID
It is extracted from the I message and sent to the tone signal generator.

The first storage means of the present invention comprises a string sound wave memory 10. The string sound waveform memory 10 is assumed to be composed of a ROM in the present embodiment. The string sound wave memory 10 may be composed of a RAM. In this case, the string sound waveform data is loaded into the string sound waveform memory 10 before the tone signal is generated. Loading can be realized, for example, by transferring a plurality of pieces of chordal sound waveform data stored in a storage medium such as a magnetic disk, an optical disk, or an IC card to the chordal sound waveform memory 10 when the power is turned on.

The string sound waveform memory 10 stores string sound waveform data created for each of a plurality of tone ranges for each of a plurality of tone colors. As described above, each string sound waveform data is created based on the sound generated by the keystroke of the strongest touch. The plurality of tones include, for example, a piano sound, a harpsichord sound, and the like. The tone color data is used to select one tone color from the plurality of tone colors. The range of each tone color is, for example, 1 note, 2 notes
~ A range of tones, such as one octave, is used. To select one range from multiple ranges,
Range data is used. As described above, the tone color data and the tone range data are supplied to the musical tone signal generator from the CPU (not shown).

The tone color data and tone range data are used as address data for the string sound waveform memory 10. Then, the sampling data included in the string sound waveform data addressed by the tone color data and the tone range data is sequentially read from the string sound waveform memory 10 and supplied to the multiplier 15.

The second storage means of the present invention comprises a percussion sound waveform memory 11. The impact sound waveform memory 11 is
In the present embodiment, it is assumed that it is composed of a ROM. The striking sound waveform memory 11 may be composed of a RAM. In this case, the above-mentioned string sound waveform memory 10
The same processing is performed as in the case where is composed of a RAM.

The percussion sound waveform memory 11 stores percussion sound waveform data corresponding to each of the plurality of string sound waveform data stored in the string sound waveform memory 10. Each impact sound waveform data is created by the method described above. The tone color data and tone range data are used as address data for the percussion sound waveform memory 11. Then, the sampling data included in the impact sound waveform data addressed by the tone color data and the range data is read from the impact sound waveform memory 11 and supplied to the multiplier 16.

The weighting factor generator 12a generates a weighting factor C1 for the string sound waveform data. The weighting coefficient C1 generated by the weighting coefficient generator 12a is data that changes according to the touch data, as shown in the characteristics for the string sound of FIG. 2, for example. Although FIG. 2 shows only the weighting coefficient C1 for one string sound waveform data, actually, the weighting coefficient C1 for a plurality of string sound waveform data is generated. Weighting coefficient C1 for any string sound waveform data
Whether to generate is determined by the tone color data and the range data.

An example of the relationship between the touch data input to the weighting coefficient generator 12a and the weighting coefficient C1 output from the weighting coefficient generator 12a is shown in the characteristic line for string sounds in FIG. The weighting coefficient C1 is configured to have a characteristic that the increase rate gradually increases from the initial value α as the touch data increases. It should be noted that it is also possible to have a characteristic that it linearly increases from the initial value α and other various characteristics.

The weighting coefficient generator 12a is, for example, RO
M. The weighting coefficient C1 for each value of the touch data is stored in advance in the ROM, for example, in a table format. The weighting coefficient generator 12a is
It can be configured by AM. In this case, the same processing as in the case where the string sound wave memory 10 is configured by a RAM is performed.

The weight coefficient generator 12a can also be constituted by a function generator. In this case, a function that inputs touch data and outputs a weighting coefficient C1 according to the characteristics for string sounds as shown in FIG. 2 is used. Weighting coefficient C generated by this weighting coefficient generator 12a
1 is supplied to the multiplier 15.

The first reading means of the present invention is composed of the waveform reading circuit 13. The waveform reading circuit 13 is a circuit for reading the waveform data from the string sound waveform memory 10 in a loop method. The waveform data read by the waveform reading circuit 13 is supplied to the multiplier 15.

The multiplier 15 multiplies the string sound waveform data from the string sound waveform memory 10 by the weight coefficient C1 from the weight coefficient generator 12a. By this multiplication, string sound waveform data having an amplitude corresponding to the touch is generated. The output of the multiplier 15 is supplied to the filter 17.

The weighting factor generator 12b generates a weighting factor C2 for the impact sound waveform data. The weighting coefficient C2 generated by the weighting coefficient generator 12b is data that changes according to the touch data, as shown in the characteristics for the hitting sound of FIG. 2, for example. Although only the weighting coefficient C2 for one hitting sound waveform data is shown in FIG. 2, the weighting coefficient C2 for a plurality of hitting sound waveform data is actually generated. Which impact sound waveform data the weighting coefficient C2 is generated for is determined by the tone color data and the tone range data.

An example of the relationship between the touch data input to the weighting coefficient generator 12b and the weighting coefficient C2 output from the weighting coefficient generator 12b is shown in the characteristic line for impact sound in FIG. The weighting coefficient C2 is configured to have a characteristic that the increase rate gradually increases from the initial value β as the touch data increases. It should be noted that it is also possible to have a characteristic that it linearly increases from the initial value β and other various characteristics.

The weighting factor generator 12b is, for example, RO
M. The weighting coefficient C2 for each value of the touch data is stored in advance in the ROM, for example, in a table format. The weight coefficient generator 12b is
It can be configured by AM. In this case, the same processing as in the case where the string sound wave memory 10 is configured by a RAM is performed.

The weighting coefficient generator 12b can also be constituted by a function generator. In this case, a function that inputs touch data and outputs a weighting coefficient C2 according to the characteristics for a hitting sound as shown in FIG. 2 is used. Weighting coefficient C generated by this weighting coefficient generator 12b
2 is supplied to the multiplier 16.

The second reading means of the present invention is composed of the waveform reading circuit 14. The waveform reading circuit 14 is a circuit for serially reading the waveform data from the impact sound waveform memory 11. The waveform data read by the waveform reading circuit 14 is supplied to the multiplier 16.

The multiplier 16 multiplies the impact sound waveform data from the impact sound waveform memory 11 by the weight coefficient C2 from the weight coefficient generator 12b. By this multiplication, impact sound waveform data having an amplitude corresponding to the touch is generated.
The output of the multiplier 16 is supplied to the adder 18.

The weighting factor generators 12a and 12 are described above.
b can be configured by sharing the same table. In this case, the weighting coefficient C1 and the weighting coefficient C2 for each touch data value may be stored in one table.

The filter 17 is composed of, for example, a low pass filter. The filter 17 filters the string sound waveform data from the multiplier 15 according to the filter parameter from the parameter generator 20 and outputs it.

The parameter generator 20 generates filter parameters for the filter 17. The target of the filtering is the string sound waveform data. The parameter generator 20 generates a filter parameter for changing the cutoff frequency according to the touch data, for example. That is, the parameter generator 20 increases the cutoff frequency when the touch data increases, and conversely,
If the touch data becomes small, a filter parameter that reduces the cutoff frequency is generated. With this configuration, the number of overtones included in the musical sound increases in the case of a hard hit
On the contrary, in the case of a weak hit, the number of overtones contained in the musical tone is reduced, and a musical tone signal having a sound quality closer to that of a natural musical instrument can be generated.

The filter 17 and the parameter generator 20 are not essential in the present invention. The filter 17 and the parameter generator 20 can be removed in a tone signal generator that does not need to control the number of overtones according to the touch strength. The output of the filter 17 is supplied to the adder 18.

The mixing means of the present invention comprises an adder 18. The adder 18 adds the string sound waveform data from the filter 17 and the impact sound waveform data from the multiplier 16. By this addition, musical tone waveform data in which the string sound waveform data and the impact sound waveform data are mixed at a ratio according to the touch strength is created. The output of the adder 18 is supplied to the multiplier 19.

The envelope generator 21 generates envelope data for generating an envelope having a shape peculiar to a tone color and a tone range and having an amplitude corresponding to the touch strength. The output of the envelope generator 21 is supplied to the multiplier 19.

The multiplier 19 multiplies the waveform data from the adder 18 and the envelope data from the envelope generator 21. By this multiplication, musical tone waveform data in which the envelope unique to the tone color and the tone range is added to the waveform data unique to the tone color and the tone range is generated. This tone waveform data is output to the outside as the output of the tone signal generator.

Next, the operation in the case where the weighting factor generators 12a and 12b generate the weighting factor according to the characteristic line as shown in FIG. 2 will be described. For example, when tone color data and tone range data designating a piano sound in a predetermined key range is supplied from the CPU (not shown) and touch data representing the strongest touch is supplied, a weighting factor C1
And C2, maximum values are generated respectively. The ratio between the weighting factors C1 and C2 is "1: 1". When a musical sound is generated based on the musical tone waveform data generated in this case, the piano sound (original sound) tapped with the strongest touch is reproduced, and the volume of the entire piano sound becomes maximum.

On the other hand, when tone color data and tone range data designating a piano sound in a predetermined key range is supplied and touch data representing the weakest touch is supplied, the weighting factors C1 and C are obtained.
The ratio with 2 is, for example, “1: 2”. That is, in the case of the weakest touch, the ratio of the volume of the striking sound to the volume of the entire generated piano sound is maximum, the ratio of the volume of the string sound is minimum, and the volume of the entire piano sound is minimum. Therefore, when a musical sound is generated based on the musical tone waveform data generated in this case, a piano sound having a large percussion sound component can be obtained although the volume of the entire piano sound is small, as compared with the case of the strongest touch.

Further, as the touch data gradually increases from the weakest touch, the volume of the entire piano sound gradually increases, and the ratio of the impact sound to the volume of the entire piano sound gradually decreases. By the control as described above, it is possible to generate a sound close to an actual piano sound that appropriately includes a hitting sound component for each touch strength.

[0055]

As described in detail above, according to the present invention,
It is possible to provide a musical tone signal generating device and a musical tone signal generating method capable of reducing the amount of waveform data and faithfully reproducing a sound including a hitting sound.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of an embodiment of a musical tone signal generating apparatus according to the present invention.

FIG. 2 is a diagram for explaining a weighting coefficient used in the embodiment of the present invention.

FIG. 3 is a diagram showing an example of a spectrum of an original sound used in the embodiment of the present invention.

FIG. 4 is a diagram showing an example of a spectrum of a hitting sound used in the embodiment of the present invention.

FIG. 5 is a diagram showing an example of a spectrum of a string sound used in the embodiment of the present invention.

FIG. 6 is a diagram for explaining a loop method according to the present invention and a conventional technique.

[Explanation of symbols]

 10 string sound wave memory 11 striking sound wave memory 12a, 12b weighting coefficient generator 13, 14 waveform readout circuit 15, 16, 19 multiplier 17 filter 18 adder 20 parameter generator 21 envelope generator

Claims (2)

[Claims] 1. A first storage means for storing waveform data obtained by sampling a sound from which an impact sound component has been removed from an original sound at a predetermined sampling frequency and which has been subjected to loop processing. A second storage unit for storing waveform data obtained by sampling a percussion sound removed from an original sound at a sampling frequency lower than the predetermined sampling frequency; and waveform data from the first storage unit in a loop method. First reading means for reading, second reading means for reading the waveform data from the second storage means, and mixing means for mixing the respective waveform data from the first reading means and the second reading means And a musical tone signal generating device based on the waveform data mixed by the mixing means. 2. A sound obtained by removing a striking sound from an original sound is sampled at a predetermined sampling frequency, subjected to loop processing and stored, and the original sound is sampled at a sampling frequency lower than the predetermined sampling frequency. Waveform data obtained by sampling the percussion sound removed from the original sound is stored, the waveform data of the sound from which the percussion sound is removed from the original sound is read by a loop method, and the waveform data of the percussion sound is read. A method of generating a musical tone signal, which comprises mixing the read waveform data and generating a musical tone signal based on the mixed waveform data.