JPH03201715A - Noise shaping circuit - Google Patents

Abstract

PURPOSE:To attain noise shaping in matching with the spectrum of an input signal by controlling the coefficient of a noise filter based on the sum and the difference of at least two consecutive samples of the input signal. CONSTITUTION:The noise shaping processing by so-called error feedback is implemented, in which a quantization noise from a quantizer 12 is fed back to an adder 11 at the input side of the quantizer 12 via a noise filter 30 as a subtraction signal. Then a filter coefficient (a) deciding the filter characteristic of the noise filter 30 is controlled based on the sum and the difference of at least two consecutive samples of the input signal X. Moreover, the filter coefficient (a) at that time is controlled by a coefficient control circuit 40 operated based on the input signal X. Thus, noise shaping in matching more with the spectrum of the input signal is implemented.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a so-called noise shaping circuit that changes the noise spectrum of an input signal.

[Summary of the invention]

The present invention provides a noise shaping circuit in which quantization noise is fed back to a quantizer via a noise filter, in which the coefficients of the noise filter are controlled based on the sum and difference values of at least two consecutive samples of an input signal. By doing so, it is possible to provide a noise shaping circuit that can perform noise shaping that more closely matches the spectrum of an input signal.

[Prior Art] For example, high-efficiency encoding that compresses and encodes the amount of information such as an audio signal can be roughly divided into encoding in the time domain and encoding in the frequency domain.

Furthermore, this time-domain or frequency-domain encoding can be performed using a method of encoding with a predetermined number of quantization bits or a method of encoding with a number of quantization bits that is adaptively changed based on the input audio signal. It can be divided into two methods. For encoding in the time domain, so-called predictive encoding, adaptive predictive encoding, and differential quantization (DP) are used.
CM) method, adaptive differential PCM (adaptive DPCM or AD
PCM) system, delta modulation 11 (ΔM) system, adaptive delta modulation system, variable length coding system, etc. In addition, as coding in the frequency domain, so-called band division coding (sub-band coding: 5BC), which divides and codes a time-axis audio signal into multiple frequency bands, and So-called adaptive transform coding (ATC), which transforms (orthogonal transform) into a signal on the frequency axis, divides it into multiple frequency bands, and adaptively encodes each band, or the above-mentioned S
By combining BC and so-called adaptive predictive coding (APC), the time axis signal is divided into bands, each band's signal is converted to Haasband (low band), and then multi-order linear predictive analysis (
Examples include so-called adaptive bit allocation (APC-AB) in which predictive coding is performed by performing LPC').

Furthermore, when encoding an audio signal, so-called noise shaping may be applied to change the spectrum of quantization noise to substantially reduce noise in the audible band.

[Problem to be solved by the invention]

Signal compression of audio signals is performed using high-efficiency encoding as described above. However, higher signal compression is currently desired. However, as described above, if the signal is compressed at a higher compression rate to further reduce the number of bits, the dynamic range will decrease and the quality of reproduced sound will deteriorate.

Also, even if the bit rate is the same, it is desirable to be able to improve the audible dynamic range of the reproduced sound.
That is, for example, digital audio (for example, so-called compact disc: CD) whose slot word length is defined in advance.

Digital audio tape recorder: DAT, etc.)
In this case, it is more preferable if the perceptual dynamic range of the reproduced sound of this digital audio signal can be increased.

Therefore, the present invention was proposed in view of the above-mentioned actual situation, and for example, even if the bit rate is the same,
The purpose of the present invention is to provide a noise shaping circuit that has a good perceptual dynamic range of reproduced sound and that prevents the quality of reproduced sound from deteriorating even if higher signal compression is performed. be.

[Means to solve the problem]

The noise shaping circuit of the present invention has been proposed to achieve the above-mentioned object, and is configured to feed back quantization noise generated in a quantizer to the input side of the quantizer via a noise filter. In the noise shaping circuit, the coefficients of the noise filter are controlled based on the sum and difference values of at least two consecutive samples of the input signal.

[Effect]

According to the present invention, the coefficients of the noise filter are controlled based on the sum and difference values of at least two samples. For example, when the difference value is large, the characteristics of the noise filter are controlled to emphasize high frequencies, When the sum value is large, control is performed to emphasize the low frequency range, and noise shaping is performed to match the spectrum of the input signal.

〔Example〕

Embodiments to which the present invention is applied will be described below with reference to the drawings.

FIG. 1 shows a schematic configuration of an encoding device provided with a noise shaping circuit according to an embodiment of the present invention.

In the encoding device shown in FIG. 1, a digital input signal X such as an audio signal supplied to an input terminal l is input to an adder
1, the signal is quantized by a quantizer 12, and then output as a coded output Y from an output terminal 2. By the example noise shaping circuit, the above quantizer 1
This process performs noise shaping processing for the quantization noise generated in step 2.

Here, the noise shaping circuit of this embodiment transmits the quantization noise from the quantizer 12 to the adder 1 on the input side of the quantizer 12 via the noise filter 30.
1, and the filter coefficient a that determines the filter characteristics of the noise filter 30 is the sum of at least two consecutive samples of the input signal X. It is controlled based on the minute and difference values. The filter coefficient a at this time is controlled by a coefficient control circuit 40 that operates based on the input signal X.

The quantization noise is obtained by the adder 13 subtracting the input signal X supplied to the quantizer 12 from the quantized output of the quantizer 12. Also,
The noise filter 30 receives the output from the adder 13 (that is, quantization noise) for a predetermined period of time (for one sample).
A delay device 14 that delays (z-') and a filter coefficient a controlled by the coefficient control circuit 40 are transmitted to the delay device 1.
This is a first-order filter composed of a coefficient multiplication circuit 36 that multiplies the output of 4.

Incidentally, the present applicant has previously proposed a configuration in which the prediction filter characteristics for the input signal are switched, and the characteristics of the noise filter are also switched. However, in this case, the characteristics of the noise filter are only switched in conjunction with switching of the characteristics of the prediction filter, and are not switched in accordance with the input signal as in the present invention.

FIG. 2 shows a more specific configuration of an embodiment of the present invention. In FIG. 2, components that operate in the same manner as in FIG. 1 are given the same reference numerals as in FIG. 1. That is, the input signal X is supplied to the input terminal l, and the noise shaping circuit performs noise shaping processing on the quantization noise generated in the quantizer 12.

Here, the noise filter 30 in FIG. 2 includes a delay device 14 that delays the quantization noise from the adder 13 for a predetermined time, a coefficient memory 16 that stores a plurality of filter coefficients, and an output of the delay device I4. It is a first-order filter having a multiplier 15 that multiplies the filter coefficient a supplied from the coefficient memory 16 by the filter coefficient a supplied from the coefficient memory 16. At this time, the filter coefficient a from the coefficient memory 16 is is selected from among the filter coefficients based on the control signal of the coefficient control circuit 40 and output. In addition,
The coefficient memory 16 and multiplier 15 in FIG. 2 become the coefficient multiplication circuit 36 in FIG.

In order to control the filter coefficient a as described above, the coefficient control circuit 40 generates the control signal corresponding to the spectrum of the input signal X, that is, the input signal
A difference value and a sum value obtained from at least two consecutive samples are compared, and a control signal corresponding to, for example, the magnitude of the difference value and sum value is outputted. here,
For example, when the above-mentioned difference value is large, it indicates that there are many high-frequency components of the input signal
The filter coefficient a is controlled. That is, the control signal output from the coefficient control circuit 40 is such that the coefficient memory 16 outputs a filter coefficient a that emphasizes the characteristics of the noise filter 30 in high frequencies. For example, when the sum value is large, it indicates that the input signal X has many low frequency components, so the control signal in this case is a filter that emphasizes the low frequency characteristics of the noise filter 30 The coefficient a becomes a signal output from the coefficient memory 16. By doing this,
Quantization noise becomes audibly inconspicuous. Expressing the above in equation (1), when input X, encoded output Y, 1 quantization characteristic Q, filter coefficient a, and delay amount z -1, Y=X+, (1-a z-' ) Q (1)
For example, when the difference value is small, the filter coefficient a is made small, and when the sum value is small, the filter coefficient a is made large.

Furthermore, in the coefficient control circuit 40 of the noise shaping circuit of this specific example, when comparing these difference values and sum values, the digital input signal X, that is, the straight PCM
(Linear pulse encoding) The value of the signal is used as a value (reference value) for comparison between the difference value and the sum value, and n samples of each of the difference value and sum value are used for more reliable comparison. A comparison is being made.

That is, in the coefficient control circuit 40, the delay output from the delay device 17 that delays the stray PCM signal by a predetermined time (one sample) and the input signal A value is required. Further, the above-mentioned difference value is obtained by subtracting the above-mentioned delayed output and the input signal X in a subtracter 19. The sum value and the difference value are each sent to a divider 20.21, where they are each divided by the delayed output (ie, the reference value). This divider 20.2
The output from 1 is accumulated every n samples in an n sample accumulator 22.23, and then output every n samples. The output for every 11n samples is transmitted to the coefficient determining circuit 24, which compares the sum value divided by the reference value with the difference value to determine the magnitude relationship between them. A control signal is output to the coefficient memory 16 according to the comparison result of the magnitude relationship. Note that the filter coefficient from the coefficient memory 16 obtained by comparing the sum value and the difference value at this time is not obtained by switching a small number of coefficients prepared in advance, but also an intermediate value. It is possible to take analog coefficient values such as

Furthermore, this filter coefficient is not necessarily output intermittently every n samples, but may be obtained every l sample by interpolating between the coefficient values obtained every n samples. can. Furthermore, for example, by shifting the comparison every n samples by one sample, it is also possible to obtain a filter coefficient for every l sample.

For this reason, if the noise shaping circuit of this embodiment is applied to an encoding device, more precise noise shaping can be performed on the spectrum of the input signal X. Further, by performing noise shaping according to the spectrum of the input signal X in this way, the masking effect on the frequency axis at the same time is enhanced. For this reason, for example, conventional digital audio (for example, the so-called compact disc), in which the slot word length is determined, can only obtain a dynamic range that corresponds to the slot word length.

Even in the case of a digital audio tape recorder (such as a digital audio tape recorder), it is possible to increase the perceptual dynamic range of the reproduced sound of this digital audio signal. Also,
Even if higher signal compression is performed, the quality of reproduced sound will not deteriorate.

will be obtained. Furthermore, even if higher signal compression is performed, it is possible to prevent the quality of reproduced sound from deteriorating.

[Brief explanation of drawings]

FIG. 1 is a block circuit diagram showing a schematic configuration of an encoding device provided with a noise shaping circuit according to an embodiment of the present invention, and FIG. 2 is a block circuit diagram showing a specific configuration of the embodiment. (Effects of the Invention) In the noise shaping circuit of the present invention, the coefficients of the noise filter are controlled based on the sum and difference values of at least two consecutive samples of the input signal. More consistent noise shaping can be performed, and for example, the playback sound of a digital audio signal has an audible dynamic range that is greater than the dynamic range determined by the slot word length.Adder 2...・・・・・・・・・・・・・・・Quantizer 4.
17・・・・・・・・・Delay device 5・・・・・・
・・・・・・・・・・・・Multiplier 6・・・・・・・・・・・・
・・・・・・Coefficient memory 9・・・・・・・・・・・・
・・・・・・Subtractor 0.21・・・・・・・・・・・・
- Divider 2.23......n sample accumulator 4...Coefficient determination circuit 30...・・・・・・・・・・・・・・・Noise filter 36・・・・・・・・・・・・・・・・・・
Coefficient multiplication circuit 40...
・Coefficient control circuit
・・・・・・Input signal Y・・・・・・・・・・・・・・・
・・・・・・Encoded output patent holder

Claims (1)

[Claims] In a noise shaping circuit that feeds back quantization noise generated in a quantizer to the input side of the quantizer via a noise filter, A noise shaping circuit characterized in that the coefficients of the noise filter are controlled.