JPH0683358A - Method and device for recording and reproducing sound data - Google Patents

Abstract

PURPOSE:To obtain the sound data recording and reproducing method and its device which are simplified in the constitution for the addition and removal of dithers. CONSTITUTION:The dither generator 6 of the sound data recording device generates the periodic dithers which periodically vary in amplitude. The periodic dithers are added to sound data prior to requantization (data compressive modulation). The sound data reproducing device removes the dither components included in the sound data from a demodulator 16 through a digital filter 18.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a sound data recording / reproducing method and apparatus. In particular, the present invention relates to a dither addition and removal technique effective in a system for compressing and recording sound information and expanding and reproducing it, and is particularly characterized by simplifying the configuration of dither addition and removal. . The reproducing apparatus of the present invention can be used as a sound source for electronic musical instruments.

[0002]

2. Description of the Related Art A sound data recording / reproducing technique is known in which a dither is added in compression recording of sound data and a dither is removed in decompression reproduction of the data compression sound data. Dither is a harmonic noise generated due to quantization error (especially a problem when the sound signal level is low)
Is a signal added to the original sound data for the purpose of reducing the noise. Quantization (requantization) by adding dither
The number of effective bits in can be increased. FIG. 7 shows the basic configuration of a conventional typical sound data recording / reproducing apparatus including addition and removal of dither. In FIG. 7, the upper half is a sound data recording device and the lower half is a sound data reproducing device.

In the sound data recording device, the sampler 2 samples the analog original sound signal at a predetermined sampling frequency. The output of the sampler 2 (analog sound amplitude sample) and the quantizer 4 (typically a high-bit AD converter) are quantized at a conversion rate equal to the sampling frequency. Amplitude quantized sound data (P
The CM data) is added with the dither from the dither generator 6P in the adder 8. Dither generator 6P is typical
Is a pseudo-random number generator that generates pseudo-random noise with a relatively narrow band. The dithered amplitude quantized sound data output from the adder 8 is requantized by a requantizer 10 or a compression modulator according to an appropriate modulation (encoding) method. Coding)
The generated sound data is stored (recorded) in the appropriate sound data memory 14 as master data.

Sound data memory 14 of the sound data reproducing device
Stores a copy of the sound data recorded in the master memory 12. The sound data memory 14 is typically composed of a ROM. The demodulator 16 subjects the encoded sound data from the sound data memory 14 to decoding, which is an inverse process of encoding, and generates decoded sound data corresponding to the output of the adder 8. This decoded sound data contains a dither component.
This dither component is noise to the reproduced sound. Therefore, it is necessary to remove the dither component. Therefore, conventionally, in the dither generator 22 of the sound data reproducing device,
The same dither as the dither generator 6P of the sound data recording device is generated. Then, the subtractor 24 removes the dither component from the sound data by subtracting the dither from the dither generator 6P from the decoded sound data from the demodulator 16. Digital sound data with dither removed is D
Converted to analog via A converter (DAC) 22
It is output as an analog reproduced sound through a sound system (not shown).

[0005]

Therefore, in the case of the conventional sound data recording / reproducing system, a dither generator which produces the same dither as the dither generator on the recording side on the reproduction side of the sound data in order to remove the dither. Must be provided. Further, in the application of recording and reproducing various sounds, the amplitude or power of the dither generator on the recording side may be adjusted to be optimized for the target sound. In such a case, the dither generator on the reproducing side also needs to adjust the amplitude or power of the dither in accordance with the dither generator, and a configuration for generating a desired dither on the recording side (thus, a configuration for removing the dither). ) Is complicated.

Therefore, an object of the present invention is to record and record sound data in which dither addition and removal are realized by a simpler method.
It is to provide a regeneration method. Further, the object of the present invention is to
Sound data reproduction method that can remove dither components included in sound data with dither added by a simple method and configuration,
It is to provide a device. Further, a specific object of the present invention is a sound data reproducing method capable of effectively removing the dither component in the sound data by the same method and configuration, regardless of the magnitude of the dither power added to the sound data. It is to provide a device.

[0007]

According to the present invention,
(A) an original sound data adding step of adding digital amplitude data as original sound data; and (B) a dither adding step of adding dither whose amplitude changes periodically to the sound data.
(C) a quantization step of quantizing the sound data to which the dither is added, (D) a recording step of recording the quantized sound data, and (E) a demodulation step of demodulating the recorded sound data, (F) a dither removing step of digitally filtering the demodulated sound data to remove a dither component included in the sound data, and (G) a reproducing step of reproducing the sound data from which the dither is removed. A sound data recording / playback method is provided.

In this configuration, since the dither added to the sound data in the dither adding step is a periodic dither whose amplitude changes periodically, the dither is simpler than the conventional method in which the dither is generated by a pseudo random number. Dither can be generated. Further, in the dither removing step, since the dither component included in the sound data is removed by digital filtering, it is not necessary to generate the same dither on the reproducing side as on the recording side. Further, since the dither component is sufficiently attenuated by digital filtering, this dither removal method realizes effective dither component removal regardless of the magnitude of the dither power added to the sound data on the recording side.

The original sound data applying step includes (a) a sampling step of sampling an analog sound signal of an original sound, and (b) a step of quantizing the sampled analog sound signal to generate digital amplitude data. Can be included.

It is desirable that the amount of sound data recorded on the recording medium is small in order to save the storage capacity. In particular, such data compression is desired for the purpose of reproducing various sounds. For this purpose, a lower bit truncation process is performed prior to the quantization process so that the lower bits (for example, 1
The lower 8 bits of 6-bit data can be truncated. In the (re) quantization process performed on the sound data in which the lower bits are truncated, a large quantization error occurs in response to the truncation, and as a result, a relatively large harmonic noise is generated. The addition of dither is particularly effective for reducing the harmonic noise associated with such truncation of lower bits.

The quantization method used in the quantization step is a high-efficiency coding method such as ADPCM for information compression.
It is preferable to use the (adaptive difference PCM) method. ADP
In the CM quantization step, difference data is formed by taking the difference between the sound data to which the dither is added, and the difference data is AD
Encoding is performed according to the PCM modulation method. In this case, in the demodulation step, the encoded difference data is decoded according to the ADPCM demodulation method, and the decoded difference data is integrated.

According to the sound data reproducing method of the present invention, (A) a demodulating step of demodulating quantized data to which dither of a predetermined frequency is added as sound data, and (B) digitally filtering the demodulated sound data. , A dither removing step of removing a dither component included in the sound data, and (C) a reproducing step of reproducing the sound data from which the dither component is removed.

Further, the sound data reproducing device of the present invention decodes the sound data storing means for storing the coded data to which the dither of the predetermined frequency is added as the sound data, and the coded data from the sound data storing means. Demodulating means for converting the sound data, digital filter means for digitally filtering the sound data decoded by the demodulating means to remove dither components included in the sound data, and sound data from which the dither components have been removed by the digital filter means. And a reproducing means for reproducing as a sound.

The digital filter means preferably has a transfer function having a zero point with respect to the dither frequency. In this case, theoretically, the dither frequency component is completely removed (attenuated). Such a digital filter means can be realized by a moving average filter that takes a moving average of demodulated sound data.

[0015]

Embodiments of the present invention will be described below with reference to the drawings. <Basic Structure (FIG. 1)> FIG. 1 shows the basic structure of the sound data recording / reproducing apparatus according to the present invention. The upper half of FIG. 1 is a sound data recording device, and the lower half is a sound data reproducing device.

Similar to the prior art (see FIG. 7), the sound data recording apparatus of the present invention includes a sampler 2, a quantizer 4, an adder 8, a requantizer 10, and a sound data memory 14. However, the dither added to the sound data by the adder 8 is not a pseudo-random number dither as in the past, but a periodic dither whose amplitude changes periodically. This periodic dither is created by the illustrated dither generator 6. For example, dither generator 6
Is 1/2 of the sampling frequency fs of the sound data to which dither is added by the adder 8 as the simplest periodic dither.
It is possible to generate a periodic dither of the frequency. In other words, for the sample sequence A (0), A (1), A (2), A (3) ... Of the sound data input to the adder 8,
The 1 / 2fs frequency dither generator is + D, -D, + D,-
D, ... pattern (where D is the dither amplitude)
Dithers with. In this case, the output sample sequence B (n) of the adder 8 is: B (0) = A (0) + D B (1) = A (1) −D B (2) = A (2) + D B (3 ) = A (3) -D ……………………………….

The advantage of using such a periodic dither is that the dither removal can be easily realized by a digital filter, and the other is the dither generator 6.
The configuration itself (and hence the dithered configuration) should also be very simple. In fact, in the example above,
Once you determine the dither amplitude D,
Dither addition can be achieved by switching between positive and negative and adding it to the sound data sample.

The frequency of the periodic dither is preferably outside the band of the sound signal. In this sense, the dither of 1/2 fs frequency, which is the simplest periodic dither, is preferable (because the bandwidth of the sound signal can be from 0 to nearly 1/2 fs). However, not limited to 1/2 fs, periodic dither of another frequency may be generated. For example, as a periodic dither of ¼fs frequency, + D,
A dither generator that generates a dither pattern of 0, -D, 0, + D, 0, -D, 0 ... Can be used.

In the sound data reproducing apparatus of the present invention, the sound data memory 14, the demodulator 16 and the DAC 20 may have the same structure as the corresponding elements of the conventional technique (see FIG. 7).
However, in the sound data reproducing apparatus of the present invention, the dither generator 22 and the subtractor 24 as in the prior art have a dither removing structure for removing dither from the sound data to which the dither is added.
Instead of using, the digital filter 18 is used. This simplifies the dither removal configuration. The digital filter 18 has a characteristic of attenuating the dither frequency component included in the decoded sound data supplied from the demodulator 16. In particular, those having a transfer function having a zero point with respect to the dither frequency are preferable.

Such a digital filter 18 can be easily realized by a moving average filter which takes a moving average of a predetermined number of consecutive samples of decoded data. That is, since the dither generator 6 on the recording side generates a periodic dither whose amplitude changes periodically, the moving average of a number of sound data sample strings corresponding to one dither cycle is used to calculate the dither component. Can be canceled.

For example, when a 1/2 fs frequency dither is added to the sound data by the adder 8, a sample string B of the output is added.
(N) is, as already shown, B (0) = A (0) + D B (1) = A (1) −D B (2) = A (2) + D B (3) = A (3 ) -D ……………………………………. Here, if the moving average of two consecutive samples of the sample sequence B (n) is taken, (+ DD) or (-
D + D) shows that the dither component is completely canceled.

In other words, the two moving average filters, which are the dither removing digital filters 18, operate according to a transfer function having a zero point at the dither frequency of 1/2 fs. That is, the inputs B ′ of the two moving average filters 18
Between (n) and the output C (n), there is a relationship of C (n) = {B '(n) + B' (n-1)} / 2 (division 2 may not be performed). . Therefore, the transfer functions of the two moving average filters 18 are H (z) = (1 + z ⁻¹ ) / 2. Here, when z = exp (j · 2πf / fs) and the frequency response characteristic is examined, f is the dither frequency fs.
When / 2 (f = fs / 2), the denominator of H (z) is zero (that is, amplitude = 0). The quantized data B '
The difference between (n) and the unquantized data B (n) is calculated by the requantizer 1
This is a quantization error generated at 0, and of the quantization noise, the dither frequency, that is, the component near the zero point B of H (z) is also attenuated by the filter 18.

In the above example, the moving average filter when the dither frequency is fs / 2 has been described, but even when the dither frequency is other than fs / 2, the moving average is obtained for demodulated sound data samples for one cycle of dither. By taking it, we can construct a digital filter with a zero point for the dither frequency,
Thereby, the dither component can be effectively removed from the sound data.

<Specific Configuration (FIGS. 2 to 5)> FIGS. 2 and 3
The more specific configurations of the sound data recording device and the sound data reproducing device are shown in FIG. In the sound data recording device of FIG. 2, after the analog original sound signal is passed through the LPF 30, the sampling and holding (S / H) circuit 32 has a sampling frequency fs of a sufficient rate for the audible signal, for example, fs =
It is sampled at 42432 Hz. Output S / H to S / H
AD converter (ADC) 3 which has a conversion speed equal to fs and has a relatively high amplitude resolution (for example, 16-bit resolution).
Quantize (digitize) at 4. The output of the ADC 34 is once recorded in an appropriate memory (not shown), and then the original sound data generation unit 36 processes the waveform. The result (referred to as original sound data) is subjected to the sound data compression processing described below, and the result (compressed sound data) is recorded in the waveform memory 46 (compressed sound data master).

The sound data is compressed in both time and amplitude directions. First, the decimation filter 38
At, data compression in the time axis direction, that is, undersampling is performed. For example, the decimation filter 38
It consists of 1/8 decimation and 16-point moving average filter. This filter is a filter input, that is, the original sound data generation unit 2
When the original sound data from 6 is represented by orig (n) and the filter output is represented by deci (n), deci (0) = 0 deci (1) = orig (0) + ... + orig (7) deci (2) = Orig (0) + ... + orig (16) deci (3) = orig (8) + ... + orig (24) .... As a result, the sampling frequency of the output data of the decimation filter 38 is 1 / of the original sound data.
8 or fs = 5304 Hz. 1
Due to the effect of 6-point moving averaging, the components around 2.5 KHz are reduced.

With respect to the output data from the decimation filter 38, the dither addition unit 40 has 1/2 fs, that is, 265.
Add a 2 Hz periodic dither. That is, the dither addition unit 40 executes the following operation: deci (0) = deci (0) + D deci (1) = deci (1) -D deci (2) = deci (2) + D. Is the dither value).

Next, the lower bit truncation unit 42 truncates the lower bits of the dithered 16-bit sound data, for example, the lower 8 bits. As a result, the sound data has 8 bits. Next, for this 8-bit sound data, ADPCM
Re-quantization is performed by the modulator 44 and amplitude-compressed data, for example, 4-bit sound data is formed. In detail,
The ADPCM modulation section 44 outputs the 2 bits from the lower bit cutoff section 42.
The difference between two 8-bit data is taken, and the previous quantization error is subtracted from the difference to form difference data a representing (current sample value of 8-bit input data-previous prediction value). (Bit data) is quantized (encoded) with an adaptive quantization width, and 4-bit encoded difference sound data is output. This encoded difference sound data is recorded in the waveform memory 46 as a master of compressed data. In the ADPCM modulator 44, the encoded difference sound data is decoded, and the difference data a before encoding is subtracted from the decoded difference data a ′ to obtain the quantization error. This quantization error is used as the previous quantization error when creating the next difference data a. The adaptive quantization width is updated appropriately based on the previous adaptive quantization width.

The lower bit truncation unit 42 may be omitted, and instead, the ADPCM modulation unit 44 may quantize the 16-bit dither-added sound data input into 4-bit encoded sound data. For demodulation, a conversion table that is looked up with the encoded difference value as the first argument and the line number related to the quantization width as the second argument and returns the decoded difference value (for example, for 4-bit encoding) In case,
An ADPCM modulation unit of a type that uses MAX as the maximum data value of each line and P as the first argument, and uses a conversion table that stores the data D represented by D = INT {(MAX · P / 7) +0.5} When 44 is used, the line number is updated based on the previous encoded difference value (ADPCM conversion value) instead of updating the adaptive quantum width itself. Also,
The encoded difference value P is calculated from the maximum value MAX (having a size related to the quantization width) of the data of the line of the conversion table indicated by the line number and the difference data a.
For example, in the case of 4-bit encoding, it is encoded according to P = INT {(a · 7-3.5) /MAX+0.5}.

As a result of the above data compression processing, the sampling frequency fs = 5304 Hz, 4
Bit / sample sound data will be recorded.

In the sound data reproducing device of FIG. 3, the same data as that recorded in the waveform memory 46 of the sound data recording device (FIG. 2) (sampling frequency of 5304 Hz and code of 4 bits / sample) is stored in the waveform memory 50. Difference sound data) is recorded. The coded difference sound data from the waveform memory 50 is decoded by the ADPCM demodulation unit 52. The ADPCM demodulation unit 52 is the ADP of the sound data recording device.
The same output as the decoded differential sound data output of the demodulation unit included in the CM modulation unit 44 is output. To this end, the ADPCM demodulator 52 adapts the quantization width in the same manner as the ADPCM modulator 44. For example, in the case of the ADPCM modulation unit 44 using the conversion table as described above, the conversion table is looked up by the decoded differential sound data from the waveform memory 50 and the current line number, and the decoded differential sound data is extracted. , The line number is appropriately updated (decrement or increment) by the same method as the line number updating method in the ADPCM modulator 44.
To do.

The maximum number of effective bits of the decoded differential sound data output from the ADPCM demodulation unit 52 is 9. By simply integrating the decoded difference sound data, the sound data corresponding to the input of the ADPCM modulator 44 is restored.

The integration / interpolation filter 54 shown in FIG. 4 performs such integration processing as well as companding data in the time axis direction,
That is, interpolation processing is performed for oversampling. For example, when 4 is selected as the oversampling ratio, the sampling frequency of the sound data in the waveform memory 50 is 5304 Hz, so the sampling frequency of the output data of the filter 54 becomes 21216 Hz due to this 4-fold oversampling. In this case, the integral quadruple oversampling interpolation filter which is the integration / interpolation filter 54 has a filter 54 input, that is, an ADPCM demodulation unit 52 output is dif (n) and a filter 54 output is so.
For example, sound (4n) = sound (4n−1) + dif (n) sound (4n + 1) = sound (4n) + dif (n) sound as und (4n + k) (k = 0,1,2,3). (4n + 2) = sound (4n + 1) + dif (n) sound (4n + 3) = sound (4n + 2) + dif (n) The integral / four-point linear interpolation calculation can be performed. When indicating this transfer of the filter function H (z), the H (z) = (1 + z -1 + z -2 + z -3) 2/4. As a result, the sound data output of the integration / interpolation filter 54 has a maximum of 10 bits and a sampling frequency 2121.
It becomes 6 Hz. The data at the beginning of sound () is initialized to 0 to prevent a clip sound during reproduction.

Next, the moving average digital filter 56 removes the dither component from the sound data. In the numerical examples described so far, the periodic dither frequency is 2652 Hz, and the sampling frequency of the input data of the moving average digital filter 56, that is, the output data of the integration / interpolation filter 54 is 21216 Hz (8 times the dither frequency). So
An 8-point moving average filter is used as the moving average digital filter 56. In this case, the filter 56 executes the processing of sound (0) = sound (0) + sound (1) + ... sound (7) sound (1) = sound (1) + sound (2) + ... sound (8). To do.

The logical configuration of the 8-point moving average filter is shown in FIG. In the figure, 561 to 567 are 1-sample delay units, 56
Reference numeral 8 is an adder. Also here
The input of the 8-point moving average filter 56 is x (n), and the output is y.
It is indicated by (n).

The transfer function H (z) of the 8-point moving average filter 56 is H (z) = (1-z- ⁸ ) / (1-z- ¹ ). This transfer function has a zero point at a frequency of 1/8 fs, that is, a dither frequency of 2652 Hz. Therefore, the dither component mixed in the sound data in the recording system can be completely removed. Of the quantization noise due to ADPCM modulation and the interpolation noise of the filter 54, 2652 Hz
Attenuates nearby components.

The output data of the 8-point moving average filter 56 is 13 bits. The amplifying unit (multiplying unit) 58 further amplifies this to have an amplitude corresponding to the full scale of the DAC 60, for example, 16-bit full amplitude. The digital sound data thus expanded and reproduced is input to the DAC 60 having a conversion frequency of 21216 Hz and 16-bit resolution, converted into an analog signal, and further passed through the LPF 62 to reproduce the analog sound.

The sound data reproducing apparatus as described in FIG. 3 can be used as a sound source of an electronic musical instrument. In this case, the waveform memory 50
It is desirable that a large number of sound data be prepared in order to enrich the timbre. In addition, the ADPCM demodulation unit 52, integration /
Interpolation filter 54, moving average filter 56, amplifier 58
The processing in the above can be realized by operating the CPU under program control. A main part of the hardware configuration when the sound data reproducing device is incorporated in an electronic musical instrument is shown by reference numeral 100 in FIG. The program is stored in the program ROM 102. The waveform data ROM 108 stores data of various sounds. The RAM 110 includes a work area for sound reproduction processing. The interrupt timer 106 is D
A cycle corresponding to the conversion cycle of AC114, for example, 1 / 20KH
In the cycle of z, the CPU 104 is requested to interrupt and the CPU 104
To execute a sound reproduction processing routine (interruption routine).

In operation, when a key is pressed on the keyboard 112, the CPU 104 detects it and assigns a reproduction channel for sound generation for the key press, and the waveform data memory 10
The beginning of the corresponding sound data in 8 is located. Thereafter, in the corresponding channel reproduction routine in the sound reproduction processing routine following the interrupt timer 102, sound data is read from the waveform data memory 108, ADPCM demodulation processing, integration / interpolation filtering processing, moving average filtering processing, and amplification processing are performed. Execute to generate channel playback data. The generated channel reproduction data is transferred to the DAC 114 in a format in which all reproduction channel data are added at the beginning of the sound reproduction processing routine. DAC1
14 converts this into an analog signal and passes it through the sound system 116 to reproduce the sound.

FIG. 6 shows the flow of the channel reproduction routine. In the ADPCM demodulation / integration / interpolation routine 200, the corresponding coded difference sound data is taken out from the sound data memory at a rate of once every four passes for 4-fold oversampling, and is decoded. Each time the decoding result D is added to SD, integration is performed. In the moving average amplification routine 300, a new integration value SD (four times oversampling linear interpolation integration value) from the routine 200 is added.
Then, the past seven integrated value samples in BUF () are added together to perform 8-point moving average filtering, and as a result, AV is amplified (multiplied) by the amplification factor a. In addition, routine 2
The integrated value sample SD from 00 is stored in the storage location n on BUF (), and the storage location is updated for the next pass. The execution result of the channel reproduction routine (channel reproduction sound data) is indicated by AV.

Although the description of the embodiment has been completed, various modifications and applications are possible within the scope of the present invention.

[0041]

As described above in detail, according to the present invention, since the dither to be added to the sound data at the time of compressing and recording the sound data uses the periodic dither whose amplitude changes periodically, the structure for dither generation is provided. To be simplified. Furthermore, use of such periodic dither facilitates removal of dither components in decompressed reproduction of sound data to be achieved by simple digital filtering such as moving average. Therefore, it is not necessary to provide the same dither generating means as the recording side on the reproducing side for removing the dither component, and the configuration for removing the dither component is simplified. Moreover, such digital filtering can achieve effective dither removal regardless of the magnitude of the dither power that can be adjusted to secure the quality of the reproduced sound.

[Brief description of drawings]

FIG. 1 is a block diagram showing a basic configuration of a sound data recording / reproducing apparatus according to the present invention.

FIG. 2 is a block diagram showing a more specific configuration of the sound data recording device.

FIG. 3 is a block diagram showing a more specific configuration of the sound data reproducing device.

FIG. 4 is a block diagram showing a logical configuration of an 8-point moving average digital filter for dither removal.

FIG. 5 is a block diagram showing a hardware configuration of an electronic musical instrument incorporating the sound data reproducing device according to the present invention.

6 is a flowchart of a channel sound reproduction routine executed by the CPU of FIG.

FIG. 7 is a block diagram showing a basic configuration of a conventional representative sound data recording / reproducing apparatus including addition / removal of dither.

[Explanation of symbols]

 6 periodic dither generator 8 adder 10 requantizer 16 demodulator 18 dither removal digital filter

Claims (11)

[Claims] 1. An original sound data applying step of (A) applying digital amplitude data as original sound data, (B) a dither adding step of adding dither whose amplitude changes periodically to the sound data, ) A quantizing step of quantizing the sound data to which the dither is added, (D) a recording step of recording the quantized sound data, (E) a demodulating step of demodulating the recorded sound data, and (F) ) A dither removing step of digitally filtering the demodulated sound data to remove a dither component included in the sound data, and (G) a reproducing step of reproducing the sound data from which the dither is removed. Sound data recording / playback method. 2. The sound data recording / reproducing method according to claim 1, wherein the original sound data providing step includes (a) a sampling step of sampling an analog sound signal of the original sound, and (b) a sampled analog A sound data recording / reproducing method comprising: a step of quantizing a sound signal to generate digital amplitude data. 3. The sound data recording / reproducing method according to claim 1, further comprising a lower bit truncation step of truncating lower bits of the dither-added sound data prior to the quantization step. Data recording / playback method. 4. The sound data recording / reproducing method according to claim 1, wherein the quantizing step forms a difference data by taking a difference between the sound data to which dither is added, and the difference data is ADPCM modulated. And a decoding step of decoding the encoded difference data according to the ADPCM demodulation method, and an integration step of integrating the decoded difference data. A sound data recording / reproducing method characterized by comprising: 5. A demodulating step of demodulating (A) quantized data to which dither of a predetermined frequency is added as sound data,
(B) a dither removing step of digitally filtering the demodulated sound data to remove a dither component contained in the sound data; and (C) a reproducing step of reproducing the sound data from which the dither component has been removed. Sound data reproduction method characterized by. 6. The sound data reproducing method according to claim 5,
The sound data reproducing method, wherein the dither removing step includes a step of taking a moving average of the demodulated sound data. 7. The sound data reproducing method according to claim 5,
The sound data reproducing method, wherein the dither removing step includes a step of performing digital filtering according to a transfer function having a zero point with respect to the predetermined frequency which is a dither frequency. 8. Sound data storage means for storing coded data to which dither of a predetermined frequency is added as sound data, demodulation means for decoding the coded data from the sound data storage means, and the demodulation means. Digital filter means for digitally filtering the sound data decoded by the above to remove dither components contained in the sound data, and playback means for playing back the sound data from which the dither components have been removed by the digital filter means as a sound, A sound data reproducing device having. 9. A sound data reproducing apparatus according to claim 8, wherein said reproducing means has DA converting means for converting sound data which is digital data into an analog sound signal. 10. Sound data storage means for storing encoded digital sound data, decoding means for decoding digital sound data from the sound data storage means, and digital data decoded by the decoding means. A digital filter means for digitally filtering the sound data in accordance with a transfer function having a zero at a predetermined frequency; and a reproducing means for reproducing the sound data digitally filtered by the digital filter means as a sound. Data playback device. 11. A sound data reproducing apparatus according to claim 10, wherein the digital filter means is a moving average filter means for taking a moving average of a predetermined number of consecutive samples of the decoded digital sound data. Characteristic sound data reproducing device.