JPH08286670A - Electronic musical instrument - Google Patents

Abstract

PURPOSE: To conduct repeated readings without requiring the device to reset a decoder at a loop top during the repeated readings by setting predicting filter coefficients so that every time the results match with each other at the leading portion of the repeated readings during a decoding. CONSTITUTION: When a musical sound signal generating section 5 transmits compressed waveforms read from a waveform memory 6 to a decoder 7, a quantization width control block controls the quantized width in the decoder 7. At that time, a CPU 10 reads the prediction filter coefficient values related to the read-out musical sound waveforms of the section 5 from a prediction filter coefficient memory section 21 and sets them as the coefficients of each multiplier of the decoder 7. Thus, the data stored in a delayed register are read, prescribed computations are executed and desired predicted values are obtained. Thus, no device is required to reset at the loop top because filter coefficients having a high converging speed and a high stability are used and the contents of the delayed registers are matched to each other at the loop top.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention requires a device for resetting a decoding device at a leading read portion in an electronic musical instrument which stores a waveform compressed by a differential PCM system or an adaptive differential PCM system and repeatedly reads the waveform. The present invention relates to an electronic musical instrument that enables repeated reading without doing so.

[0002]

2. Description of the Related Art In electronic musical instruments, how to generate musical sounds that are close to natural sounds is an important issue.
In recent years, the PCM method in which a natural sound is sampled and stored, and a musical sound is generated by reading the natural sound from the stored waveform memory is predominant.

However, the waveform of a natural sound is different for each pitch and volume, and in order to generate a musical sound that is closer to a natural sound by the PCM system, a different waveform for each pitch and volume is used. It is necessary to prepare it, and the waveform memory for storing this is huge.

Therefore, as a method of compressing this memory, a differential PCM system (hereinafter referred to as DPCM system) for storing a differential value from a predicted value using a prediction filter and an adaptive differential P
The CM method (hereinafter referred to as ADPCM method) is used.

Further, in the PCM system, not all the data from the start to the end of the sound generation of the natural sound are stored, but a sound generation start portion and a part of the waveform after that are stored and the sound generation start portion is reproduced. After that, a memory is saved by using a method of forming a musical sound by repeatedly reproducing a part of the stored waveforms.

However, these DPCM system and AD
In the PCM method, it is considered that the repeated reading of the waveform is an effective means, but it has a drawback that errors are accumulated because the data is encoded and processed by using the prediction filter.

Further, in order to read the waveform repeatedly, it is necessary that the contents of the delay register of the prediction filter match at the beginning of the repetition.

Therefore, in the conventional electronic musical instrument, in order to prevent the accumulation of errors and restore the same original waveform every time at the head portion (loop top) of the repeated reading, the contents of the delay register are re-read at the head portion of the repeated reading. Measures were taken to set it up, and equipment was provided for it.

As described above, in order to repeatedly read out the waveform in the conventional electronic musical instrument, since a special device for resetting is required, there is a drawback that the structure is complicated and expensive, and a countermeasure is required. Was there.

[0010]

SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and in an electronic musical instrument that stores a waveform compressed by the DPCM system or the ADPCM system, when decoding is repeatedly performed at the loop top, the decoding is performed. An object of the present invention is to provide an electronic musical instrument that enables repeated reading without requiring a device for resetting the device.

[0011]

According to the present invention, in an electronic musical instrument for decoding a waveform compressed by a differential PCM system or an adaptive differential PCM system and repeatedly reading the waveform, the waveform repeatedly read is repeated at the time of decoding. It is characterized in that it is coded by a prediction filter in which prediction filter coefficients are set so that the results match each time at the beginning of reading.

As the prediction filter coefficient, a prediction filter coefficient calculation means 51 for sampling the original waveform and calculating the prediction filter coefficient, and a prediction filter for storing the prediction filter coefficient calculated by the prediction filter coefficient calculation means 51. A coefficient storage unit 52, a coding unit 61 that codes and compresses an original waveform based on the prediction filter coefficient calculated by the prediction filter coefficient calculation unit 51, and a waveform that is coded by the coding unit 61 are stored. It is characterized in that the predictive filter coefficient and the compressed waveform are set in advance by the waveform storage means 62.

The electronic musical instrument according to the present invention has a waveform memory 62 for storing the waveform data obtained by the compression method.
And a prediction filter coefficient memory 52 for storing the prediction filter coefficient, and a difference PC based on the prediction filter coefficient.
A feature of the present invention is that the waveform data compressed by the M method or the adaptive difference PCM method is repeatedly read, and the waveform data is decoded.

[0014]

According to the present invention, in the electronic musical instrument in which the waveform is compressed by the DPCM method or the ADPCM method and the waveform is repeatedly read, the prediction is performed by using the filter coefficient which converges quickly and has high stability, thereby repeatedly reading. The error can be absorbed by the end part (loop end) of.

Therefore, in the present invention, for example, a general-purpose computer is used to previously obtain optimum waveforms and filter coefficients by simulation, and the waveform data and prediction filter coefficients obtained as a result of the simulation are mounted on the electronic musical instrument. Is.

As a result, the contents of the delay register always have the same value at the beginning portion (loop top) of the repeated reading, so that when the end portion (loop end) of the repeated reading moves to the beginning portion (loop top). A device for resetting a new value is not required, and it is possible to provide a low-priced electronic musical instrument with a simple structure and excellent compression efficiency.

[0017]

First, the prediction filter coefficient will be described.
FIG. 8 is a diagram illustrating the relationship between the value of the prediction filter coefficient and the stability. In the figure, FIG. 8A shows an example of a first-order prediction filter, FIGS. 8B and 8C show an example of impulse response of an unstable filter, and FIG. 8D shows a filter with high stability. 3 shows an example of the impulse response of.

For example, assume that data of impulses 1, 0, 0 is input. At this time, it is assumed that the filter coefficient of the prediction filter in FIG. 8A is a = 1.

In this case, 1 is input at first and 1 is output, but 0 is input at the second and subsequent times, and a value a times the previous output value is added from the filter. Therefore, Figure (8)
As shown in B, 1 is always output from the filter.

Further, when a = 2, for example, the output values become 1, 2, and 4, and diverge rapidly as shown in FIG. 8 (C), causing a failure such as overflow.

Therefore, the output value when a = 0.1 is used as the filter coefficient is set as shown in FIG. 8D so that the impulse response converges within the loop period of the waveform in which the repeated reading is performed. , 1, 0.1, 0.01, 0.00
It becomes 1 and the input value rapidly converges to 0.

Therefore, the error in the case of repeatedly reading the waveform is not accumulated at the loop end, and the value of the loop top is correctly set at the time of reading the next waveform.

FIG. 9A is a diagram showing an example of an error accumulated when the same waveform is repeatedly read. When the error is accumulated as shown in the figure, the amplitude of the waveform at the loop end is read with the error added, and this affects the amplitude of the loop top of the waveform read next time.

Therefore, as illustrated in FIG. 9 (A),
Each time the waveform is read, errors are successively accumulated and the waveform shifts upward as shown in the figure. If it is repeated many times, it causes an obstacle such as overflow.

FIG. 9B is a diagram showing an example of output when a filter with high stability is used. As shown in the figure, the error at the loop end is absorbed by the loop end of the next iterative reading, so that the error is not accumulated every time the iterative reading is performed.

For this reason, stable repeated reading of the same waveform as shown in FIG. 9B is possible, and a device for resetting a new value at the loop top is not required.

The present invention has been made paying attention to such a point, and when a waveform compressed by the DPCM method is repeatedly read out, a filter having a fast convergence and a high stability is used, so that the loop interval is increased. The error is not accumulated.

An embodiment in which the present invention is applied to the DPCM system will be described below.

Since the first-order prediction filter is used in the above description, if the prediction filter coefficient is -1 <a <1, it can be converged.

However, the shapes of the waveforms that are actually input vary widely, and it is necessary to use second-order, third-order, and high-order filters in order to improve the accuracy of the filters and reduce noise. Yes, the prediction filter coefficient cannot be determined simply.

Therefore, in the present invention, the waveform previously compressed by the DPCM system or the ADPCM system and the optimum predictive filter coefficient corresponding to the waveform are simulated by using, for example, a general-purpose computer to use the waveform data. The prediction filter coefficient is obtained in advance, and the waveform is mounted on an electronic musical instrument for use.

FIG. 2 is a diagram for explaining a procedure for creating a prediction filter coefficient and a compressed waveform in advance by using a general-purpose computer or the like. A procedure for creating waveform data of a predictive filter coefficient and a compressed waveform according to the present invention will be described below with reference to the drawings.

In the figure, 50 is a prediction block, and 6
0 is an encoding unit. The prediction block 50 includes a prediction filter coefficient calculation unit 51 and a prediction filter coefficient memory unit 52, and the encoding unit 60 includes a DPCM encoding unit 61 and a waveform memory 62.

The predictive filter coefficient calculating section 51 extracts a plurality of samples from the input original waveform such as natural sound and predicts them, and calculates and determines the optimum predictive filter coefficient. The determined coefficient is stored in the prediction filter coefficient memory unit 52 and also the DPCM encoding unit 6
Sent to 1.

The predictive filter coefficient memory unit 52 is composed of, for example, a ROM, and stores the predictive filter coefficient determined by the predictive filter coefficient calculating unit 51. The prediction filter coefficient stored in the prediction filter coefficient memory unit 52 is used as data stored in the prediction filter coefficient memory unit 21 of the electronic musical instrument.

The DPCM encoder 61 quantizes the input original waveform, and encodes and compresses the waveform by the DPCM method based on the prediction filter coefficient sent from the prediction filter coefficient calculator 51. The compressed waveform is stored in the waveform memory unit 62.

The waveform stored in the waveform memory unit 62 is used as data in the waveform memory 6 of the electronic musical instrument. It should be noted that the prediction filter coefficient stored in the prediction filter coefficient memory unit 52 and the compressed waveform data stored in the waveform memory unit 62 form a pair, and related data therefor is also a predetermined area of the ROM (not shown). Memorized in.

Next, referring to FIG.
The operation of the encoder in the following case will be described. In encoding, the prediction filter coefficient is first determined.

In determining the prediction filter coefficient, the sample data of the input original waveform is taken out by the prediction filter coefficient calculation section 51, and based on the sample data, the prediction filter coefficient calculation section 51 uses, for example, Levinson-D.
The prediction filter coefficient is obtained by a method such as the urbin method or the Burg method.

The obtained prediction filter coefficient is stored in the prediction filter coefficient memory section 52 and sent to the encoding section 60.

The encoding unit 60 includes a delay register 61.
1, 612, multipliers 613, 614, 615. Adder 6
16, 617 and a quantization control block 618.

In such a configuration, the prediction filter coefficient sent from the prediction block 50 is set in K0 and K1 as the prediction filter coefficient of the primary and secondary filters.

Next, when a part of the original waveform is fetched by the prediction filter block 60, the values that were inputted two times before last time and fetched by the delay registers 611, 612 are recalled, and the respective multipliers 613, 614 give predetermined values. The coefficients are multiplied and then added by the adder 616.

Further, after the sign is inverted by the multiplier 615, it is added to the original waveform input by the adder 617 to obtain the difference value. Then, the quantization width control block 61
Quantized by 8. As a result, a waveform to which a predetermined compression has been applied is output, and the waveform is stored in the waveform memory 62.

Since the waveform according to the DPCM method is created in this manner, the manufacturer, when manufacturing the electronic musical instrument, obtains the waveform compressed by the above method and the prediction filter coefficient, and the data for associating the waveform with the prediction filter coefficient. By incorporating it into an electronic musical instrument, it is possible to provide an electronic musical instrument whose waveform memory is compressed.

Moreover, as in the prior art, when the loop end is reached, it is not necessary to provide a special device for resetting the value of the loop top of the waveform that is repeatedly read next time, and a low-priced electronic musical instrument can be provided. Become.

Next, an electronic musical instrument equipped with the waveform and the prediction filter created according to the above embodiment will be described with reference to the drawings. FIG. 1 is a schematic block diagram illustrating the overall configuration of an electronic musical instrument according to the present invention.

In the figure, 10 is a CPU, 11 is a ROM, 12 is a RAM, and 13 is a display section. Also, 1
Is a keyboard, 2 is a keyboard scan circuit, 3 is an operation panel, 4 is a panel scan circuit, 5 is a tone signal generator, and 6 is a waveform memory.

Further, 7 is a decoder, 8 is a digital control filter (hereinafter referred to as DCF), 9 is a digital control amplifier (hereinafter referred to as DCA), 14 is a digital / analog converter (hereinafter referred to as D / A converter), 15
Is an amplifier and 16 is a speaker.

The CPU 10 controls each section of the electronic keyboard instrument according to a control program stored in a program memory section (not shown) of the ROM 11, and controls to read out predetermined data according to a key pressing portion of the keyboard 1. It is a thing.

The ROM 11 stores tone color data and various other fixed data in addition to the program for operating the CPU 10 described above. The filter coefficients directly related to the present invention are stored in the predictive filter coefficient memory unit 21 of the ROM 11.

The RAM 12 has a work area for the CPU 10,
Various registers, counters, flags, buffers, etc. for controlling the electronic musical instrument are defined, and ROM1
1 has a data area in which necessary data is transferred and temporarily stored.

Further, a plurality of registers in which data necessary for sound emission is set corresponding to the setting states of the keys and switches of the operation panel 3 and the tone generating circuits (DCO) of the tone signal generating section 5 are not used. The RAM 12 is also provided with an assigner memory for storing data to be assigned to channels, a storage area for storing tone information, and the like.

The keyboard 1 is used for designating a musical sound to be generated, and is composed of a plurality of keys and a key switch which opens and closes in conjunction with key pressing / release operations of these keys. The keyboard scanning circuit 2 is used by the performer to press and release keys.
The detected signal is sent to the tone signal generator 5 under the control of the CPU 10.

The performance information generated by pressing or releasing the keyboard 1 is temporarily stored in a predetermined area of the RAM 12 and read by the CPU 10 as required.

The keyboard scanning circuit 2 detects the key-pressing / key-releasing operation of the player, that is, the ON / OFF state of the key, and transmits the detected ON / OFF information to the tone signal generating section 5 together with the key number. . The CPU 10 stores the ON / OFF information of this key in the RAM 12.

In addition to the power switch, the operation panel 3 is provided with various switches and indicators such as a tone color selection switch, a mode designation switch, a melody selection switch and a rhythm selection switch.

The set / reset state of each switch of the operation panel 3 is detected by the panel scan circuit 4 included therein, and the data on the set state of the switch detected by the panel scan circuit 4 is stored in the CPU 10. It is stored in a predetermined area on the RAM 12 under control.

In addition to the various switches described above, the operation panel 3 is provided with a display 13 for displaying various information.

The tone signal generator 5 reads the tone waveform data and envelope data corresponding to the signal output from the CPU 10 from the waveform memory 6, adds the envelope to the read tone waveform data, and outputs the tone signal. is there.

The musical tone signal generator 5 has a number of digital control oscillators (hereinafter referred to as "DC
"O"), and a tone signal is generated by a different DCO for each sound and sent to the decoder 7.

A waveform memory 6 for storing waveform data and envelope data is connected to the tone signal generator 5.

The decoder 7 decodes the waveform data sent from the tone signal generator 5, and the predictive filter coefficient of the present invention is used when the decoder 7 decodes the waveform data. It

The DCF 8 is a digital control filter, which adds a tone color change to the musical tone waveform sent from the decoder 7. The DCA 9 is a digital control amplifier, which applies amplitude modulation to the tone waveform transmitted from the DCF 8.

The amplifier 15 amplifies the analog musical tone signal supplied from the D / A converter 14 with a predetermined gain. The output of the amplifier 8 is supplied to the speaker 16.

The speaker 16 converts an analog musical tone signal as an electric signal sent from the amplifier 15 into an acoustic signal. That is, the musical tone is emitted according to the generated musical tone signal.

With such a structure, when the performance is started,
The key depression / key release data input from the keyboard 1 connected via the key depression detection unit 2 and the tone generation condition set by the operation panel 3 connected via the panel scan circuit 4 are RA.
It is temporarily stored in M12.

At a predetermined timing, RA
The keyboard data and panel event data stored in M12 are read out by the CPU 10, processed and given to the tone signal generator 5, and the compressed tone signal is read out and given to the decoder 7 for decoding. Then, the original waveform is generated.

Next, the decoding flow of the decoder 7 of the electronic musical instrument of the present invention will be described with reference to FIG.

The portion directly related to the decoding of the present invention is a waveform memory 6 for storing a compressed waveform and a prediction filter for storing a prediction filter coefficient of a filter read when decoding the compressed waveform. A coefficient memory unit 21,
The DPCM decoding unit 22 that performs DPCM decoding, and the association data storage unit (not illustrated) that stores the data that associates the waveform data read from the waveform memory 6 with the prediction filter coefficient.

In such a configuration, when it is detected that the keyboard 1 is pressed, for example, under the control of the CPU 10, the musical tone signal generator 5 reads the corresponding waveform data from the waveform memory 6 to generate a musical tone signal, Send to the decoder 7.

On the other hand, the CPU 10 reads the association data for determining the prediction filter coefficient corresponding to the waveform read by the tone signal generator 5, and based on this, the prediction filter coefficient memory unit 21 outputs a predetermined prediction filter coefficient. It is read and sent to the DPCM decoding unit 22.

The DPCM decoding section 22 performs decoding based on this to generate a musical tone waveform. The decoding operation will be described in detail with reference to FIG.

FIG. 5 is a diagram for explaining the configuration of the decoder 7. As shown in the figure, the decoder 7 includes a quantization width control block 2
01, the delay registers 202 and 203, the multiplier 204,
205 and adders 206 and 207.

In such a configuration, the compressed waveform read out from the waveform memory 6 by the tone signal generator 5 is decoded by the decoder 7
To the quantization width control block 20 in the decoder 7.
1 controls and sends the quantization width.

At this time, at the same time, the CPU 10 reads from the prediction filter coefficient memory unit 21 the prediction filter coefficient value associated with the musical tone waveform read by the musical tone signal generation unit 5, and sets K0, as the coefficient of each multiplier of the decoder 7. K1
Set to.

As a result, the delay registers 202 and 203
The data stored in the
After a predetermined calculation is performed in 205, it is sent to the adder 206 and added to obtain a desired predicted value.

The predicted value is sent to the adder 207 and added to the input data to generate the original tone waveform, which is sent to the DCF 8.

Note that these operations are performed by the tone signal generator 5
This is repeated every time the waveform data is sent, and a part of the waveform added and sent by the adder 207 is sequentially taken into the delay registers 202 and 203, and the delay register 202 is used as the data of the waveforms of the previous time and the previous time. , 20
3 is stored.

In such an operation, if the filter coefficient of the present invention with fast convergence and high stability is used, the contents of the delay registers 202 and 203 always match at the loop top.
No device is needed to reset at the loop top.

It should be noted that the encoder and the decoder of the present invention are 2
Although the following case is described as an example, the present invention is not limited to this. Also, the prediction filter coefficient memory is set to RO
Although provided in M, the memory may be provided in another position.

Next, a case will be described in which the filter coefficient of the present invention, which converges quickly and has high stability, is applied to a waveform compressed by the ADPCM system.

The ADPCM system is a differential PCM system using adaptive quantization. Since the quantization width is adaptively controlled according to the amplitude value, higher-performance coding is possible as compared with the DPCM system. Known as.

However, when the present invention is applied to the ADPCM system, the actual structure and operation of the device are as follows:
This is almost the same as the case of applying the above method, and the only difference is the signal encoding and decoding part. Therefore, the description of the parts that perform the same operation will be omitted, and only the differences will be described.

FIG. 6 is a diagram for explaining a waveform compression procedure when the present invention is applied to the ADPCM system. As shown in the figure, the difference from the DPCM method when determining the filter coefficient is that the DPCM encoding unit 61 of the prediction filter block 60 performs AD processing.
Only the PCM encoding unit 63 has been changed.
The method of encoding by the method is a known technique.

Therefore, as in the case of the DPCM system, the filter coefficient is arbitrarily set, the waveform is coded based on the ADPCM coding system, and the operation of reproducing and checking the waveform is repeated to obtain a suitable prediction filter coefficient. Just select it.

Also, when the waveform according to the present invention is mounted on an electronic musical instrument, the DPCM encoder 2 as shown in FIG.
2 is changed to the ADPCM encoding unit 23, and A
The encoding method by the DPCM method is a known technique,
The operation procedure is the same as in the case of DPCM.

According to the present invention, even in the waveform compressed and coded by the ADPCM system, it is not necessary to reset the waveform when moving from the loop end to the loop top, and furthermore, the quality is higher than that of the DPCM coding. It is possible to obtain a high waveform.

In the present embodiment, the case where the predictive filter coefficient is determined for each waveform has been described as an example, but the present invention only needs to be able to absorb the error within the cycle of the waveform repeatedly read out. The prediction filter coefficient may be changed for each group.

[0090]

As described above in detail, according to the present invention, D
In an electronic musical instrument equipped with a waveform compressed by the PCM coding method or the ADPCM coding method, it is not necessary to reset when shifting from the loop end of the waveform to be repeatedly read to the loop top, and it has a high compression rate and a structure. A simple and low-priced electronic musical instrument can be provided.

[Brief description of drawings]

FIG. 1 is an overall configuration diagram of an electronic musical instrument equipped with the waveform data of the present invention.

FIG. 2 is a diagram illustrating an operation of calculating a filter coefficient using a general-purpose computer or the like in DPCM waveform compression.

FIG. 3 is a diagram illustrating an encoding operation in a general-purpose computer.

FIG. 4 is a diagram for explaining an operation of a decoder of an electronic musical instrument having a waveform based on the DPCM method.

FIG. 5 is a block diagram illustrating a configuration of a decoder of an electronic musical instrument having a waveform based on the DPCM method.

FIG. 6 is a diagram illustrating an operation of calculating a filter coefficient by using a general-purpose computer or the like in ADPCM waveform compression.

FIG. 7 is a diagram for explaining the operation of a decoder of an electronic musical instrument that has a waveform based on the ADPCM method.

FIG. 8 is a diagram illustrating a relationship between a filter coefficient and an impulse response.

FIG. 9 is a diagram (continuation) illustrating the relationship between filter coefficients and impulse response.

[Explanation of symbols]

Electronic musical instrument 1 Keyboard 2 Keyboard scan circuit 3 Operation panel 4 Panel scan circuit 5 Music signal generator 6 Waveform memory 7 Decoder 8 DCF 9 DCA 10 CPU 11 ROM 12 RAM 13 Display 14 D / A converter 15 Amplifier 16 Speaker 21 Prediction Filter coefficient memory 22 DPCM decoding unit 23 ADPCM decoding unit 201 Quantization width control block 202, 203 Delay register 204, 205 Multiplier 206, 207 Adder general-purpose computer 50 Prediction block 51 Prediction filter coefficient calculation unit (prediction filter coefficient calculation) Means) 52 Prediction filter coefficient memory (prediction filter coefficient storage means) 60 Prediction filter block (encoder) 61 DPCM coding section (coding means) 62 Waveform memory (waveform storage means) 63 ADPCM coding section (coding means) ) 611, 612 Delay register 613, 614, 615 Multiplier 618 Quantization block

─────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] March 29, 1996

[Procedure Amendment 1]

[Document name to be amended] Statement

[Correction target item name] 0065

[Correction method] Change

[Correction content]

The amplifier 15 amplifies the analog musical tone signal supplied from the D / A converter 14 with a predetermined gain. The output of the amplifier 15 is supplied to the speaker 16.

[Procedure Amendment 2]

[Document name to be corrected] Drawing

[Name of item to be corrected] Fig. 4

[Correction method] Change

[Correction content]

[Figure 4]

[Procedure 3]

[Document name to be corrected] Drawing

[Name of item to be corrected] Figure 7

[Correction method] Change

[Correction content]

[Figure 7]

Claims (3)

[Claims] 1. An electronic musical instrument which performs repeated reading by compounding a waveform compressed by a differential PCM method or an adaptive differential PCM method, wherein the repeated reading waveform has a result obtained at the beginning of the repeated reading at the time of decoding. An electronic musical instrument characterized by being encoded by a prediction filter in which prediction filter coefficients are set so as to match. 2. The predictive filter coefficient, a predictive filter coefficient calculating means for sampling an original waveform and calculating a predictive filter coefficient, and a predictive filter coefficient memory for storing the predictive filter coefficient calculated by the predictive filter coefficient calculating means. And a coding unit that codes and compresses an original waveform based on the prediction filter coefficient calculated by the prediction filter coefficient calculation unit, and a waveform storage unit that stores the waveform coded by the coding unit. 2. The electronic musical instrument according to claim 1, wherein the predictive filter coefficient and the compressed waveform are set in advance. 3. A waveform memory for storing the waveform data obtained by the compression method according to claim 1, and a prediction filter coefficient memory for storing a prediction filter coefficient, and based on the prediction filter coefficient. An electronic musical instrument characterized by performing decoding of waveform data in which a waveform compressed by a differential PCM system or an adaptive differential PCM system is repeatedly read.