JPH09162701A - Quantization noise reduction device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a quantization noise reduction device by which the effect of quantization noise is reduce from even a quantized value made to an analog signal whose amplitude level is relatively smaller than a quantization width. SOLUTION: In the case of reducing quantization noise by converting a sample data string composed of a prescribed word-bit length M into a string with a word bit length N higher than the M, output signals of bit length conversion means 21-24 are divided into plural band components, each level of an input signal obtained for each divided band is compared with each reference level 71-74, and when the signal level is smaller than the reference level, a control signal is outputted to attenuate a corresponding band component, the band components are synthesized 4 again and the result is outputted as a sample data string with a bit length N. Waveform distortion for a very small signal is reduced by bit length extension and a gain at a frequency band whose quantization noise is dominant is decreased, then a signal waveform with more fidelity with an original signal is reproduced even from a signal reproduction device accompanied with digital conversion.

Description

Detailed Description of the Invention

[0001]

[0001]

[0002]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a quantization noise reducing device for reducing quantization noise generated when an analog signal is quantized by a predetermined word bit length, and more specifically, a predetermined word bit length. The present invention relates to a quantization noise reduction device that reduces quantization noise by converting a sample data string composed of M into a sample data string having a word bit length N larger than M.

[0003]

[0002]

[0004]

2. Description of the Related Art In general, when digitizing a voice signal which is an analog signal, a quantization step of sampling the voice signal and quantizing the voice signal at a predetermined sampling frequency is essential. In this quantization step, the sample value which is the amplitude value of the analog audio signal at each sampling timing is quantized at the sampling timing and converted into a quantized value. At this time, since the quantized value can take only a value that is an integer multiple of the predetermined quantization width, there is a so-called quantization error between the quantized value and the sample value, and this quantization error is It becomes quantization noise in the digitization process and is perceived as noise on the sense of hearing.

[0005]

[0003]

[0006]

The quantization noise is audible because the ratio of the quantization noise to the signal becomes small when the amplitude level of the quantized analog signal is relatively larger than the quantization width. Although not so noticeable, when the amplitude level of the analog signal is relatively small, there is a problem that the ratio of the quantization noise to the signal becomes large and the influence of waveform distortion due to the quantization appears.

[0007]

An object of the present invention is to provide a quantizing noise reducing apparatus capable of reducing the influence of quantizing noise even on a quantizing value performed on an analog signal having a relatively small amplitude level. .

[0008]

[0005]

[0009]

SUMMARY OF THE INVENTION In order to solve the above-mentioned problems, the present invention according to claim 1 provides a sample data string having a predetermined word bit length M with a sample having a word bit length N larger than M. A quantization noise reducing device for reducing quantization noise by converting into a data string, the bit length converting means converting a sample data string having the word bit length M into a sample data string having a word bit length N or more. Band dividing means for dividing the output signal of the bit length converting means into a plurality of band components, signal level detecting means for detecting a signal level of the band component, and a predetermined value for setting the detected signal level. In response to the control signal from the level comparing means for comparing and outputting a control signal when the signal level is smaller than the predetermined value, And attenuation means for attenuating the response band component, and combining means for outputting as the sample data string of the word bit length N by combining each band component, and a.

[0010]

According to a second aspect of the invention, the predetermined value is
It is configured to set the value proportional to the theoretical quantization noise level according to the word bit length M.

[0011]

[0007]

[0012]

According to the first aspect of the present invention, the bit length conversion means converts a sample data string obtained by quantizing an analog signal with a word bit length M into a sample data string having a word bit length N larger than M. . The band dividing unit divides the sample data having the word bit length N converted by the bit length converting unit into a plurality of band components. The signal level detecting means provided for each band component detects the signal level of each band component. Each detected signal level is compared with a predetermined value set for each band, and when the signal level is smaller than the predetermined value, the attenuator attenuates the corresponding band component. The band components subjected to the signal processing as described above are combined and output as a monkey data string having a word bit length N. Therefore, it is possible to reduce the waveform distortion of the minute signal due to the bit expansion and to reduce the gain in the frequency band in which the quantization noise generated when the quantization is performed with the word bit length M is dominant.

[0013]

According to the second aspect of the invention, the predetermined value is set to a value proportional to the theoretical quantization noise level due to the word bit length M. Therefore, the quantization noise in the noise dominant band can be attenuated.

[0014]

[0009]

[0015]

BEST MODE FOR CARRYING OUT THE INVENTION Preferred embodiments of the present invention will be described with reference to the drawings. In the following embodiment, the word bit length of the input sample data is 16
A case will be described in which the number of bits is 32 bits, the bit length is converted to 32 bits, and the band is divided into four.

[0016]

FIG. 1 is a block diagram of a quantization noise reducing device according to the present invention. The quantization noise reduction device shown in this figure has an input terminal 1 for inputting sample data with a word bit length of 16 read from a CD (compact disc) or the like, and a bit length of the input sample data with a desired bit length. In order to convert the input sample data into bands and to convert the input sample data into bands, for example, the bit length conversion circuits 21 to 24 including acyclic digital filters shown in FIG. The output data from the circuits 21 to 24 are allowed to pass through at the same signal level, while the control signals are present, the attenuating circuits 31 to 3 which attenuate the signal level of the output data from the bit length converting circuits 21 to 24.
4 and a signal level detection circuit 51 that receives the output data signals of the bit length conversion circuits 21 to 24 and detects each signal level.
To 54, storage circuits 61 to 64 for storing reference levels corresponding to quantization noise levels, and signal level circuits 51 to 51.
The comparator circuits 71 to 74 that generate a control signal by comparing the output of 54 with the reference level stored in the storage circuit, and the sample data that is attenuation-controlled for each band are recombined, and the composite data of the subsequent stage is combined. It is composed of a synthesis circuit 4 for outputting to an A / D converter (not shown). Note that the digital filter can also be configured as a recursive digital filter.

[0017]

Next, the operation will be described.

The 16-bit sample data input to the input terminal 1 is, for example, as shown in FIG. 2, the maximum frequency spectrum of the original signal defined based on the Nyquist interval (when the sampling frequency is Fs, the maximum frequency is Fs / It is 2.)
The bit length conversion circuits 21 to 24 each of which is composed of a digital filter in which each pass band is set in order to divide the band including (including audio signals recorded in a CD, 22.05 KHz) at equal intervals or octave intervals. Sent to. As shown in FIG. 3, the digital filter multiplies the input sample data by an output of the delay memory D having a delay amount equivalent to the Nyquist interval T and the output of the delay memory D of each stage, and a coefficient value a determined according to the filtering characteristic. And a multiplier for Now, if the bit length of the sample data is 16 bits and the bit length of the coefficient value multiplied by the multiplier is 16 bits, the bit length of the output data from the multiplier is 2 ¹⁶ × 2 ¹⁶ = 2 ³² Therefore, it is 32 bits. By defining the bit length of the coefficient value, the bit length of the output data from the multiplier can be converted in units of 2 bits. By this bit length conversion, the output from the bit length conversion circuit is 3
Since it is represented by a quantization width of 2 bits, the resolution is improved. The input data that has undergone band division and bit length conversion is input to the attenuation circuits 31 to 34 and the signal level detection circuits 51 to 54 for each of the divided bands. The signal level detection circuits 51 to 54 convert the band-limited input sample data into an absolute value and output it to the comparison circuits 71 to 74.
Further, the comparison circuits 71 to 74 are also provided with a reference level corresponding to the quantization noise level associated with the 16-bit quantization stored in the storage circuits 61 to 64. Further, the comparison circuits 71 to 74 have a clock circuit for clocking a predetermined time determined for each band, and this clock circuit is a band sent from the signal level detection circuits 51 to 54 at sampling intervals. The clocking operation starts when the signal level of the limited sample data falls below the reference level, and a clocking completion signal is generated when clocking for a predetermined time, and the clocked value is reset When the signal level of the sample data exceeds the reference level, the clock value is reset. This timing operation is performed in order to identify whether each band component is a noise component or a signal component as information. Therefore, it is preferable to set the predetermined time to be measured shorter as the frequency band becomes higher. The comparison circuits 71 to 74 compare the reference level given from the storage circuits 61 to 64 with the signal level of the band-limited sample data given from the signal level detection circuits 51 to 54, and the signal level is smaller than the reference level. When the state continues for a predetermined period of time, that is, when the time counting completion signal is generated from the time counting circuit, a control signal for attenuating the gain of the sample data is output to the attenuating circuits 31 to 34.

[0019]

Here, the reference level stored in the storage circuits 61 to 64 will be described.

As the reference level, a level corresponding to X _MAX × 2 ^-15 which is the theoretical maximum quantization noise level of 16-bit sample data input to the input terminal 1 is adopted. X _MAX is the maximum amplitude of the input signal, which is 1 for a CD, for example. That is, in the band-divided input sample data, the band component that does not reach the quantization noise level is regarded as a noise component,
It blocks the passage of that band component. Therefore,
The gain A set by the control signal in the attenuation circuits 31 to 34 is preferably 0, but at least 1
If it is less than twice (0 <A <1), the noise reduction effect can be obtained.

[0021]

Each band component subjected to the gain control as described above is output to the synthesizing circuit 4 to reduce quantization noise,
It is resynthesized as an audio signal with reduced waveform distortion due to bit length extension. The single signal shown in FIG. 4 (for example,
From the frequency spectrum of the output signal from the noise reduction device according to the present invention when a sine wave signal of 1000 Hz) is input, the energy of the quantization noise in the band excluding the band in which the original signal exists (band 1 in this case) is attenuated. You can see how it is being done.

[0022]

As the reference level stored in the storage circuits 61 to 64, instead of the theoretical quantization noise level as described above, in the human auditory perception characteristics shown in FIG.
For example, a signal level proportional to the minimum audible level may be used. That is, the reference level may be set according to the auditory characteristics for each of the divided bands.

[0023]

Further, in the above-described embodiment, an example of converting from 16 bits to 32 bits is shown as an example of the bit length conversion, but the invention is not limited to this, and the bit length may be longer than the bit length of the input sample data. So any number of bits is fine. Further, when the bit length conversion circuit converts the bit length to a word bit length of N bits (32 bits in the above-described embodiment) or more and performs the noise reduction processing described above, the synthesis circuit 4 performs synthesis: Alternatively, it is also possible to perform a rounding operation so as to obtain N word bits after combining.

[0024]

Further, in the above embodiment, the band division is 4
Although the example of division is described, the present invention is not limited to this, and the band may be finely divided if necessary.

[0025]

[0017]

[0026]

As described above, according to the present invention, when the quantization data is reduced by converting the sample data string having the predetermined word bit length M into the sample data string having the word bit length N larger than M. In addition, the output signal of the bit length conversion means is divided into a plurality of band components, the signal level of the input signal obtained for each divided band is compared with a reference level, and when the signal level is smaller than the reference level, the control signal is output. After outputting and attenuating the corresponding band component, each band component is recombined and output as a sample data string of word bit length N, so that the waveform distortion of a minute signal is reduced by bit expansion. Since the gain in the frequency band where the quantization noise is dominant can be reduced, a signal that is more faithful to the original signal, even in a signal reproducing device that involves digital conversion. It can reproduce the form.

[Brief description of the drawings]

FIG. 1 is an overall configuration block diagram of a quantization noise reduction device according to an exemplary embodiment of the present invention.

FIG. 2 is a diagram showing an example of band division in the embodiment of the present invention.

FIG. 3 is a block diagram showing a digital filter that constitutes a bit length conversion circuit according to an embodiment of the present invention.

4 is a diagram showing a frequency spectrum of an output signal from the quantization noise reduction device of FIG.

FIG. 5 is a diagram showing human audible characteristics.

[Explanation of symbols]

 1 ... Sample data input terminal 21-24 ... bit length conversion circuit 31-34 ... attenuation circuit 4 ... synthesis circuit 51-54 ... signal level detection circuit 61-64 ... memory Circuits 71 to 74 ... Level comparison circuit

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

[Procedure amendment]

[Submission date] December 16, 1996

[Procedure amendment 1]

[Document name to be amended] Statement

[Correction target item name] Detailed description of the invention

[Correction method] Change

[Correction contents]

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a quantization noise reducing device for reducing quantization noise generated when an analog signal is quantized by a predetermined word bit length, and more specifically, a predetermined word bit length. The present invention relates to a quantization noise reduction device that reduces quantization noise by converting a sample data string composed of M into a sample data string having a word bit length N larger than M.

[0002]

2. Description of the Related Art In general, when digitizing a voice signal which is an analog signal, a quantization step of sampling the voice signal and quantizing the voice signal at a predetermined sampling frequency is essential. In this quantization step, the sample value which is the amplitude value of the analog audio signal at each sampling timing is quantized at the sampling timing and converted into a quantized value. At this time, since the quantized value can take only a value that is an integer multiple of the predetermined quantization width, there is a so-called quantization error between the quantized value and the sample value, and this quantization error is It becomes quantization noise in the digitization process and is perceived as noise on the sense of hearing.

[0003]

The quantization noise is audible because the ratio of the quantization noise to the signal becomes small when the amplitude level of the quantized analog signal is relatively larger than the quantization width. Although not so noticeable, when the amplitude level of the analog signal is relatively small, there is a problem that the ratio of the quantization noise to the signal becomes large and the influence of waveform distortion due to the quantization appears.

An object of the present invention is to provide a quantizing noise reducing apparatus capable of reducing the influence of quantizing noise even on a quantizing value performed on an analog signal having a relatively small amplitude level. .

[0005]

SUMMARY OF THE INVENTION In order to solve the above-mentioned problems, the present invention according to claim 1 provides a sample data string having a predetermined word bit length M with a sample having a word bit length N larger than M. A quantization noise reducing device for reducing quantization noise by converting into a data string, the bit length converting means converting a sample data string having the word bit length M into a sample data string having a word bit length N or more. Band dividing means for dividing the output signal of the bit length converting means into a plurality of band components, signal level detecting means for detecting a signal level of the band component, and a predetermined value for setting the detected signal level. In response to the control signal from the level comparing means for comparing and outputting a control signal when the signal level is smaller than the predetermined value, And attenuation means for attenuating the response band component, and combining means for outputting as the sample data string of the word bit length N by combining each band component, and a.

According to a second aspect of the invention, the predetermined value is
It is configured to set the value proportional to the theoretical quantization noise level according to the word bit length M.

[0007]

According to the first aspect of the present invention, the bit length conversion means converts a sample data string obtained by quantizing an analog signal with a word bit length M into a sample data string having a word bit length N larger than M. . The band dividing unit divides the sample data having the word bit length N converted by the bit length converting unit into a plurality of band components. The signal level detecting means provided for each band component detects the signal level of each band component. Each detected signal level is compared with a predetermined value set for each band, and when the signal level is smaller than the predetermined value, the attenuator attenuates the corresponding band component. The band components subjected to the signal processing as described above are combined and output as a monkey data string having a word bit length N. Therefore, it is possible to reduce the waveform distortion of the minute signal due to the bit expansion and to reduce the gain in the frequency band in which the quantization noise generated when the quantization is performed with the word bit length M is dominant.

According to the second aspect of the invention, the predetermined value is set to a value proportional to the theoretical quantization noise level due to the word bit length M. Therefore, the quantization noise in the noise dominant band can be attenuated.

[0009]

BEST MODE FOR CARRYING OUT THE INVENTION Preferred embodiments of the present invention will be described with reference to the drawings. In the following embodiment, the word bit length of the input sample data is 16
A case will be described in which the number of bits is 32 bits, the bit length is converted to 32 bits, and the band is divided into four.

FIG. 1 is a block diagram of a quantization noise reducing device according to the present invention. The quantization noise reduction device shown in this figure has an input terminal 1 for inputting sample data with a word bit length of 16 read from a CD (compact disc) or the like, and a bit length of the input sample data with a desired bit length. In order to convert the input sample data into bands and to convert the input sample data into bands, for example, the bit length conversion circuits 21 to 24 including acyclic digital filters shown in FIG. The output data from the circuits 21 to 24 are allowed to pass through at the same signal level, while the control signals are present, the attenuating circuits 31 to 3 which attenuate the signal level of the output data from the bit length converting circuits 21 to 24.
4 and a signal level detection circuit 51 that receives the output data signals of the bit length conversion circuits 21 to 24 and detects each signal level.
To 54, storage circuits 61 to 64 for storing reference levels corresponding to quantization noise levels, and signal level circuits 51 to 51.
The comparator circuits 71 to 74 that generate a control signal by comparing the output of 54 with the reference level stored in the storage circuit, and the sample data that is attenuation-controlled for each band are recombined, and the composite data of the subsequent stage is combined. It is composed of a synthesis circuit 4 for outputting to an A / D converter (not shown). Note that the digital filter can also be configured as a recursive digital filter.

Next, the operation will be described. The 16-bit sample data input to the input terminal 1 is, for example, as shown in FIG. 2, the maximum frequency spectrum of the original signal defined based on the Nyquist interval (when the sampling frequency is Fs, the maximum frequency is Fs / 2. .) Is included in the band (22.05 KHz for an audio signal recorded on a CD) at equal intervals or octave intervals so that each pass band is set in a bit length conversion circuit including a digital filter. 21 to 24. As shown in FIG. 3, the digital filter multiplies the input sample data by an output of the delay memory D having a delay amount equivalent to the Nyquist interval T and the output of the delay memory D of each stage, and a coefficient value a determined according to the filtering characteristic. And a multiplier for now,
If the bit length of the sample data is 16 bits and the bit length of the coefficient value multiplied by the multiplier is 16 bits, the bit length of the output data from the multiplier is 2 ¹⁶ × 2 ¹⁶ = 2 ^32. , 32 bits. By defining the bit length of the coefficient value, the bit length of the output data from the multiplier can be converted in units of 2 bits. By this bit length conversion, the output from the bit length conversion circuit is 3
Since it is represented by a quantization width of 2 bits, the resolution is improved. The input data that has undergone band division and bit length conversion is input to the attenuation circuits 31 to 34 and the signal level detection circuits 51 to 54 for each of the divided bands. The signal level detection circuits 51 to 54 convert the band-limited input sample data into an absolute value and output it to the comparison circuits 71 to 74.
Further, the comparison circuits 71 to 74 are also provided with a reference level corresponding to the quantization noise level associated with the 16-bit quantization stored in the storage circuits 61 to 64. Further, the comparison circuits 71 to 74 have a clock circuit for clocking a predetermined time determined for each band, and this clock circuit is a band sent from the signal level detection circuits 51 to 54 at sampling intervals. The clocking operation starts when the signal level of the limited sample data falls below the reference level, and a clocking completion signal is generated when the clocked for a predetermined time, and the clocked value is reset When the signal level of the sample data exceeds the reference level, the clock value is reset. This timing operation is performed in order to identify whether each band component is a noise component or a signal component as information. Therefore, it is preferable to set the predetermined time to be measured shorter as the frequency band becomes higher. The comparison circuits 71 to 74 compare the reference level given from the storage circuits 61 to 64 with the signal level of the band-limited sample data given from the signal level detection circuits 51 to 54, and the signal level is smaller than the reference level. When the state continues for a predetermined time, that is, when the time counting signal is generated from the time counting circuit, a control signal for attenuating the gain of the sample data is output to the attenuating circuits 31 to 34.

Here, the reference level stored in the storage circuits 61 to 64 will be described. As this reference level, a level corresponding to X _MAX × 2 ⁻¹⁵ which is the theoretical maximum quantization noise level of 16-bit sample data input to the input terminal 1 is adopted. X _MAX is the maximum amplitude of the input signal, which is 1 for a CD, for example. That is, in the band-divided input sample data, the band component that does not reach the quantization noise level is regarded as a noise component, and the passage of the band component is blocked. Therefore, the gain A set by the control signal in the attenuation circuits 31 to 34 is preferably 0, but if it is at least less than 1 time (0 <A <1),
A noise reduction effect can be obtained.

Each band component subjected to the gain control as described above is output to the synthesizing circuit 4 to reduce quantization noise,
It is resynthesized as an audio signal with reduced waveform distortion due to bit length extension. The single signal shown in FIG. 4 (for example,
From the frequency spectrum of the output signal from the noise reduction device according to the present invention when a sine wave signal of 1000 Hz) is input, the energy of the quantization noise in the band excluding the band in which the original signal exists (band 1 in this case) is attenuated. You can see how it is being done.

As the reference level stored in the storage circuits 61 to 64, instead of the theoretical quantization noise level as described above, in the human auditory perception characteristics shown in FIG.
For example, a signal level proportional to the minimum audible level may be used. That is, the reference level may be set according to the auditory characteristics for each of the divided bands.

Further, in the above-described embodiment, an example of converting from 16 bits to 32 bits is shown as an example of the bit length conversion, but the invention is not limited to this, and the bit length may be longer than the bit length of the input sample data. So any number of bits is fine. Further, when the bit length conversion circuit converts the bit length to a word bit length of N bits (32 bits in the above-described embodiment) or more and performs the noise reduction processing described above, the synthesis circuit 4 performs synthesis: Alternatively, it is also possible to perform a rounding operation so as to obtain N word bits after combining.

Further, in the above embodiment, the band division is 4
Although the example of division is described, the present invention is not limited to this, and the band may be finely divided if necessary.

[0017]

As described above, according to the present invention, when the quantization data is reduced by converting the sample data string having the predetermined word bit length M into the sample data string having the word bit length N larger than M. In addition, the output signal of the bit length conversion means is divided into a plurality of band components, the signal level of the input signal obtained for each divided band is compared with a reference level, and when the signal level is smaller than the reference level, the control signal is output. After outputting and attenuating the corresponding band component, each band component is recombined and output as a sample data string of word bit length N, so that the waveform distortion of a minute signal is reduced by bit expansion. Since the gain in the frequency band where the quantization noise is dominant can be reduced, a signal that is more faithful to the original signal, even in a signal reproducing device that involves digital conversion. It can reproduce the form.

Claims (2)

[Claims] 1. A quantization noise reduction device for reducing quantization noise by converting a sample data string having a predetermined word bit length M into a sample data string having a word bit length N larger than M, wherein: A bit length converting means for converting a sample data string having a word bit length M into a sample data string having a word bit length N or more; a band dividing means for dividing an output signal of the bit length converting means into a plurality of band components; Signal level detecting means for detecting a signal level of the band component, level comparing means for comparing the detected signal level with a set predetermined value, and outputting a control signal when the signal level is smaller than the predetermined value, Attenuating means for attenuating the corresponding band component in accordance with the control signal from the level comparing means, and the word bit by combining the respective band components. And a synthesizing means for outputting as a sample data string of length N. 2. The quantization noise reduction device according to claim 1, wherein the predetermined value is set to a value proportional to a theoretical quantization noise level based on the word bit length M.