JPH03244210A - Digital filter circuit and transmitter/receiver - Google Patents

Abstract

PURPOSE:To decrease the number of filters by half and to reduce the amount of a hardware by providing a subtraction circuit to operate the difference of two inputs with the output of the first digital filter and the output of a multiplier circuit as the two inputs, and using the output of the first digital filter and the output of the subtraction circuit as two output signals. CONSTITUTION:A signal passed through a digital filter 13 of an upside channel is sampled by a 1/2 Nyquist frequency fs by a sub sampling circuit 20 and encoded by an encoder circuit 21. The signal of the downside channel is made alpha-times by a multiplier circuit 14 and afterwards, a subtraction circuit 15 calculates difference from the signal passed through the digital filter 1 3. For this signal, sub sampling is executed by a sub sampling circuit 22 similarly to the upper channel and afterwards, this signal is encoded by an encoder circuit 23. The signal of two divided and encoded bands is multiplexed by a multiplexer 24 and multiply transmitted to the reception side with time division by a multiple transmission line 25. On the reception side, the received multiple signal is multiply divided by a multiplexer 26 and two band signals are taken out.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of Industrial Application The present invention relates to a digital filter circuit used for audio transmission or image transmission, and a transmitter/receiver using the digital filter circuit.

[Summary:- The present invention is a system in which an input signal is band-divided using a digital filter, encoded and transmitted, and the filter processing in the band-splitting is performed by converting one side into a difference signal between the original signal and the other filtered signal. By doing so, the number of filters can be reduced and filter design can be made easier.

[Conventional technology]

BACKGROUND ART In recent years, band division coding, in which an input signal is separated into frequency components using a plurality of spatial filters and then coded and transmitted, has been attracting attention as a highly efficient coding method for image signals or audio signals.

An example of an audio signal transmission system using this band division coding is shown in FIG.

In the configuration shown in FIG. 13, on the transmitting side, a band pass filter (BPF) is used to convert the audio signal into multiple frequency bands.
The sub-sampling circuit 20.2 divides each band signal into sub-sampling circuit 20.2.
2, it is subsampled at the Nyquist rate and converted to a low frequency signal, encoded by, for example, adaptive predictive coding (APCM) in the encoding circuits 21 and 23, multiplexed in the multiplexer 24, and sent to the receiving side via the multiplex transmission path 25. Send to. On the receiving side, after code decoding and interpolation are performed by decoding circuits 27.29 and interpolation circuits 28.30,
Pass through band pass filters 35 and 36 and adder circuit 37
The output signals are combined to reconstruct the decoded signal ε(n).

This band division encoding method for audio signals has the advantage of being able to improve overall audio quality by allocating a larger number of quantization bits to bands where audio energy is concentrated. This has the advantage of preventing noise from affecting other bands.

This band division method is performed in units of two divisions, so
FIG. 14 shows and explains the most basic one-dimensional two-band system horizontal aperture.

In the configuration shown in FIG. 14, on the transmitting side, the input signal x (n) is band-divided by two low-pass digital filters (h+ (n)) 6, Then, sub-sampling circuits 20 and 22 sample and transmit the signals, and interpolation circuits 28 and 30 perform O-value interpolation on the receiving side, and then the band-divided and input signals are sent to digital filter 6 A low-pass digital filter (g+ (
n)) 8. Bypass dicicle filter (g2(n)
)9, the signals are combined in an adder circuit 37 to reconstruct the decoded signal x(n).

Here, it is impossible to realize a filter with ideal frequency characteristics for boat fishing with a finite number of taps, and three types of distortion occur in the decoded signal: aliasing distortion, amplitude distortion, and phase distortion. However, if the relationship (1) holds between the filters, it becomes possible to make the decoded signal x (n) completely match the original signal x (n).

For this reason, based on this equation (1), a filter that can completely reconstruct a decoded signal is being studied.

As an example of the configuration of this complete reconstruction filter, first, we will use the QMF (Quadrature Mirror) proposed by Crochiere.
Filter).

Literature Crochere, “Subband Corp. Inc.”
Bell Systems Technical Journal No. 60 (
September 1981〉, pp. 1633-1653 (R,”-
E. Crochiere, “5ubband Co.
ding,” B, S, T, J.

60, pp, 1633-1653 Sep, 19
81) In this method, the filter h+(n) is a linear phase FIR filter with symmetrical coefficients, and the other filter h2(
n), g+(n), and g2(n) (h2(n)
= (1)”h + (n)1g1(n)=2・hl(
n) (2) (gz (n)=-2-(-t)"h+(n), complete removal of aliasing distortion and phase distortion in the decoded signal is achieved. However, the amplitude Complete removal of distortion is limited to cases where the number of taps of the filter is 2, which is infinite. Under other circumstances, an approximation method is used.

Therefore, when used for image processing, a filter with 16 taps or more is required.

On the other hand, Le Ga1l et al. proposed a 5SKF filter based on the viewpoint that a long-tap filter like QMF is undesirable when considering band division processing for image processing. Symmet
rlc 5hort to ernel Filter)
proposed a filter configuration called .

Literature Lugol, Tabata Pai, r-subband coding of digital images using
Symmetric short kernel filter configuration
and Arithmetic Coding Techniques J, IEεε Eyecasp 88 (June 1988),
Pages 761-764 (D, L, Ga1l and A,
Tabatabai, “5ubband Coding
of Digital Images [Ising
Symmetric 5hort KernelFa
lters and Arithmetic Codi
ng Techinques, “Proc, of
IEEE ICASSP'88. Pp, 761-764
(June, 1988)) This method relaxes the constraints between filters compared to the QMF described above, and the design problem of a complete reconstruction filter is F(2)-
This results in the problem of factorizing the polynomial F(2) of 2 that satisfies (4) into Hl (z) and H2(-2).

In this case, F (z) is a polynomial in which there is only one term with a non-zero coefficient among its odd-order terms, but Lugol et al. It was shown that it is possible to find a concrete solution to 5SKF by imposing a coefficient symmetry condition centered on the term and performing factorization under that constraint.

The 5SKF proposed by Lugol et al. can completely eliminate aliasing distortion and phase distortion, and also
- MF also solves the problem of amplitude distortion, which is only an approximation. Furthermore, a concrete complete reconstruction filter is defined as a coefficient symmetric FI with the number of taps at most 3 and 5, or 4 and 4.
We have shown that this can be achieved using an R filter.

Furthermore, Yasuda et al. have proposed the following method regarding a technique related to hierarchical encoding that has characteristics similar to band division encoding.

Literature Yasuda, Takagi, Kato, Awano, "Stepwise transmission and display of still images using hierarchical encoding method" Transactions of the Institute of Electronics and Communication Engineers (B), Vol. J63-B No. 4 (April 1980), No. 379 ~386 pages Literature Yasuda, Kato, "Still image coding and its applications" Journal of the Institute of Electronics, Information and Communication Engineers Vol. 71. No. 7. pp. 66
9-675 July 1988 This method constructs a hierarchical structure of the image by recursively using low-pass filters and subsampling, and transmits the image from the upper layer corresponding to the low frequency components in the initial stage of transmission. This will help you understand the overview.

This method aims to reduce the transmission rate by calculating the difference between layers, and this operation is equivalent to frequency division in band division coding. However, unlike band division coding, this method has a problem in that the number of pixels to be coded increases as the number of layers increases.

[Problem to be solved by the invention]

As described above, the conventional hierarchical encoding method for images has a problem in that the number of pixels to be encoded increases as the number of hierarchical processing increases. For example, in the case of 1 layer processing, 1.25 (1+%) times, 2 layer processing
In the case of 1.3125 (1+1/4 +1/16) times, the number becomes 1.3125 (1+1/4 +1/16) times, and as the number of layer processing increases, it becomes asymptotic to 4×3 times.

On the other hand, in conventional band division coding, there is no problem that the number of pixels to be coded increases, but as mentioned above, QM
When F is used, amplitude distortion is eliminated by approximation, so it is difficult to design a filter that eliminates amplitude distortion, making it impossible to construct a perfect reconstruction filter.

In addition, in the 5SKF method, when the polycyclic formula F (2) becomes a higher-order formula, its factorization becomes very complicated, so
There was a problem in that it became difficult to extend good frequency characteristics to a multi-tap filter, making filter design difficult.

Furthermore, this S S K F! The method according to : factorizes the polycyclic F (Z).
The frequency characteristics of h , (n), h 2 (n)) are given only as a result of factorization, and there is a problem in that the filter coefficients cannot be determined according to the desired frequency characteristics.

Furthermore, in the conventional method, a filter (h(n), g(n)) is required for each band to be divided.
There was a problem in that the number of filters increased in proportion to the number of band divisions, resulting in an increase in the amount of hardware.

[Means to solve the problem]

The digital filter circuit of the present invention includes a first digital filter h (n) through which an input digital signal passes;
A multiplication circuit (α) which receives the input digital signal as an input, and a subtraction circuit which takes the output of the first digital filter and the output of the multiplication circuit as two inputs and calculates the difference between the two inputs, The present invention is characterized in that the output of the first digital filter and the output of the subtraction circuit are two output signals.

Further, the digital filter circuit of the second invention includes a first digital filter (h
(n) ), a multiplication circuit (α) which receives the input digital signal as input, and a subtraction circuit which takes the output of the first digital filter and the output of the multiplication circuit as two inputs and calculates the difference between the two inputs. an adder circuit which receives the output of the first digital filter and the output of the subtractor circuit as two inputs and calculates the sum of the two inputs; an adder circuit through which the output signal of the adder circuit passes and A second digital filter (g(n)) having a passband substantially opposite to that of the digital filter, and a synthesis circuit that combines the output of the second digital filter and the output of the first digital filter. It is characterized by:

Further, the transmitting/receiving device of the present invention includes a first digital filter h (n) through which an input digital signal passes, a multiplier circuit (α) that receives the input digital signal, an output of the first digital filter, and a multiplier circuit (α) that receives the input digital signal. a subtraction circuit that takes the output of a multiplier circuit as two inputs and calculates the difference between the two inputs; a transmission means that transmits the output of the first digital filter and the output of the subtraction circuit to two subband transmission lines; An adder circuit that takes the signals received from these two subband transmission paths as two inputs and calculates the sum of the two inputs, and an adder circuit through which the output signal of this adder circuit passes and which is almost opposite to the first digital filter. The present invention is characterized in that it includes a second band digital filter (g (n)), and a synthesis circuit that combines the outputs of these two digital filters and the output of the first digital filter.

Note that the first and second digital filters can also be configured as two-dimensional filters.

[Effect]

In the present invention, conventional digital filters that divide into two bands are hl (n) and h2 (n), and digital filters that decode the signal are g+(n) and gz.
(n), the first digital filter h(n) is used for one band.
A multiplication circuit with a coefficient α is provided in the other band, the output of this first digital filter is used as the output signal in one band, and the difference between the output of the multiplier circuit and the output of the first digital filter is used in the other band. The configuration is such that the output is taken as an output signal, and both outputs are sampled and encoded.

FIG. 1 shows a one-dimensional two-band band division system configuration which is the basis of the present invention, and its operation will be explained based on this diagram.

In FIG. 1, reference numeral 13 denotes a digital filter having a filter coefficient h (n), and 14 denotes an input signal x (n) with α
15 is a subtraction circuit that takes the difference between the output of the multiplication circuit 14 and the output of the digital filter 13, 20.22 is a sub-sampling circuit indicated by a downward arrow, and 28.30 is indicated by an upward arrow. An interpolation circuit for zero-value interpolation performed in response to subsampling, 32 an adder circuit, 33 a digital filter with a filter coefficient g(n), 34 an upper channel signal and digital filter 3
This is a synthesis circuit that performs synthesis with the output signal of No. 3.

When a digital filter circuit is configured as shown in Figure 1,
The system configuration shown in Figure 14 is H+ (2)
= H(2) Equivalent when set as 1.

Therefore, the complete reconfiguration condition for the system configuration shown in Figure 1 is Σ (1-(-1)'') h(n)z-'=2*
z-)' (7) is required to hold.

however Therefore, the filter coefficient is As long as the following is satisfied, complete reconstruction can be achieved.

On the other hand, if T is the sampling period between pixels, then the filter coefficient h'(n) of an odd-tap FIR filter that aims to realize half-band low-pass characteristics can be calculated as ω(n) by determining its impulse response. indicates the window function.

All even-numbered terms in the equation for α are 0 except for n=(], and as long as the delay for odd-numbered taps is taken into account, it can be directly applied to the filter coefficient h (n).

In addition, when the digital filter (h(n)N3) functions as a low-pass filter, the lower channel signal, which is given as the difference between the original signal and the signal passed by the digital filter (h(n)N3), has the direct current of the original signal. It is desirable that no component be included.For this purpose, it is necessary to set the parameter α of the multiplier circuit to an appropriate value.This parameter α can be determined as follows.

This is equivalent to α-H(ωT)-〇 holds when ωT=0, and for convenience, let h(0〉 be the odd term that is forced to 1 in h(n)), then the filter's The amplitude characteristic is H(e””)l = h(0)+2Σh (2n+ 1)
It is given as cos (2n+1) ωTα blade. In this equation (2), since it can be expressed as α-h(0)+2Σh(2n+1) n+0 04). On the other hand, in order for the filter (h (n)) to function as a low-pass filter and to realize H(π)-0, it is necessary that the relationship between the filter coefficients be satisfied. Therefore, α-2-,,Q6) is obtained.

In this way, even if the original signal is multiplied in one band-divided channel to remove its DC component, and the signal obtained by taking the difference from the filtered signal of the other channel is sampled, it is still possible to As with the method, the decoded signal can be reconstructed.

〔Example〕

The present invention will be described in detail below with reference to the drawings.

FIG. 2 shows an example of a band division image transmission system using the digital filter circuit shown in FIG.

This embodiment shows an example in which image information is divided into two bands and transmitted using the digital filter circuit of the present invention.

The input analog signal x (t) is filtered by a low-pass filter 10
The DC component is extracted by
is converted to This digital signal j=converted image signal is input to the digital filter circuit 12 of the present invention and is divided into two bands.

In Figure 2, the upper channel digital filter (
The signal passing through h(n)) 13 is subsampled by a subsampling circuit 20 at a percentage of the Nyquist frequency fS, and encoded by an encoding circuit 2. In addition, the lower channel signal is multiplied by α in the multiplier circuit 14 and then sent to the subtracter circuit 1.
5, the difference between the signal and the signal that has passed through the digital filter 13 is taken. This signal is subsampled by the subsampling circuit 22 in the same way as the upper channel, and then encoded by the encoding circuit 23. These band-divided and encoded two-band signals are multiplexed by a multiplexer 24 and time-division multiplexed transmitted to the receiving side via a multiplex transmission path 25.

On the receiving side, the received multiplexed signal is demultiplexed by a demultiplexer 26 to extract two band signals. The upper channel signal is decoded by a decoding circuit 27, and this signal is subjected to O-value interpolation by an interpolation circuit 28 using the Nyquist frequency fs. The lower channel signal is also decoded by the decoding circuit 29 and subjected to O value interpolation by the interpolation circuit 30.

The signals subjected to O-value interpolation by the interpolation circuits 28 and 30 are input to the digital filter circuit 31 on the receiving side of the present invention. That is, the signal subjected to O-value interpolation in the lower channel is added to the upper channel signal by the adder circuit 32, and this signal passes through the digital filter (g(n)) 33.

The signal that has passed through the digital filter 33 and the upper channel signal are combined by a combining circuit 34 to restore the digital signal x (n). This restored digital signal x (n) is converted to a digital-to-analog conversion circuit (D
/A) is converted into an analog signal x (t) by 38, and the image signal of the restored image is extracted. In this embodiment, the upper channel digital filter 13 is a low-pass filter, and the lower channel digital filter 33 is a low-pass filter.
is configured as a no\pass filter.

Next, specifically, the configuration of the digital filter circuit 12 is explained in the third section.
This will be explained with reference to the diagram.

FIG. 3 is a block diagram showing the configuration of the digital filter circuit 12, and the digital filter 13 includes six delay elements (D) 131.132.133.134.135.
．． 136, multipliers 142 and 143 with a coefficient a, multipliers 141 and 144 with a coefficient A, and an adder 151.

The output of the adder 151 for the upper channel signal is led to the subsampling circuit 20. The lower channel signal input to the multiplication circuit 14 with coefficient α is transmitted to the delay element 131.
.

Here, the values of the filter coefficients a and b of the multipliers 141 to 144 are determined by changing the window function ω(n) in the α-th equation described above to ω(n
)=o, 54-0.46 cos(2πn/M)
A Hamming window defined by αD, and this α
This can be determined by setting the parameter M in equation 1 to 8. The Teja filter coefficients of the digital filter circuit shown in FIG. 3 are a=0.545 and b--0,045.

A digital filter 13 using these filter coefficients a and b
FIG. 4 shows a comparison of the frequency characteristics of the 5SKF and the conventional 5SKF.

The digital filter shown in FIG. 4 has 11 taps, and its characteristics are shown by solid lines. 5SKF
Cases where the number of taps is 3 and 5 are listed, and are represented by broken lines and dotted lines, respectively. However, the number of non-zero filter coefficients in the frequency characteristics shown in FIG. becomes 8.

On the other hand, 5SKF configured as a pair of low-pass filters and high-pass filters with tap numbers of 3 and 5 or 5 and 3 also has 8 multiplications per pixel, so
Comparing the two based on the number of multiplications, it can be seen that the digital filter circuit of the present invention exhibits better frequency characteristics.

FIG. 5 is a block diagram showing the configuration of the digital filter circuit 31 on the receiving side in the embodiment.

The digital filter circuit 31 on the receiving side is also configured to have an opposite pass band corresponding to that on the transmitting side, and the digital filter 33 has delay elements 331, 332, 333, 334, 335, and the output of the adder v&32. .336 to the taps and coefficient symmetric multiplier 341.
Adder 351 via 342.343.344.345
The decoded signal x (n) is reconstructed by combining the upper channel signal inputted to the upper channel signal and delayed by the delay elements 361 and 362 and the combining circuit 34 .

In this way, in this embodiment, the digital filter (h(
n)) 13 and (g(n)) 33 can be constructed by coefficient symmetric filters, and their filter coefficients a1b can be easily obtained, making it possible to easily design a complete reconstruction filter.

Next, an example will be described in which band division of a linear two-band configuration is extended to two dimensions.

In order to process image signals by band division, it is desirable to extend the one-dimensional configuration to two-dimensional signal processing.

This expansion method includes a method of using a repartition type filter configuration or a method of using a two-dimensional filter configuration.

FIG. 6 shows, corresponding to FIG. 1, a configuration expanded to a two-dimensional system using a repartition filter configuration.

In this re-division filter configuration, first, a digital filter (h) 130, a multiplication circuit 140, a subtraction circuit 15,
A digital filter circuit 12 consisting of and a sub-sembling circuit 20. , 22. Filter processing and subsampling processing are performed in the horizontal direction. Subsequently, a digital filter circuit 122 consisting of a digital filter (h(n)) 132, a multiplication circuit 142, and a subtraction circuit 15□ and a subsampling circuit 202.22
□ performs filter processing and subsampling processing on the upper channel in the vertical direction. Also, for the vertical direction of the lower channel, the digital filter (h(n)) 1
33, a digital filter circuit 123 consisting of a multiplication circuit 142 and a subtraction circuit 153, and subsampling circuits 203 and 223 perform filter processing and subsampling processing. The four signals thus obtained are each encoded and transmitted to the receiving side.

On the receiving side, interpolation processing and filter processing are performed in the vertical direction, contrary to those on the transmission side, and then interpolation processing and filter processing are performed in the horizontal direction to obtain the final decoded signal x (m, n).

In this batch type configuration, the band is divided into four bands, and the image signal is transmitted using the four bands as the transmission path.

Figure 7 shows the linear digital filter (
h(n), g(e)) with a two-dimensional digital filter (h
(m, n), g (L n) ) to create a two-dimensional system configuration. In this case, unlike the batch type configuration shown in FIG. 6, the band is divided into two bands.

Therefore, when the two-dimensional sampling pattern is taken as an discrete pattern as shown in FIG. 8 (O indicates a pixel that is sampled and linked, and x indicates a pixel that is not sampled), the perfect reconstruction condition is given by: With a sharp cut,
As in the case of - dimension, the filter coefficient is 1, arbitrary (m, n: even number or odd number 09) g (m, n) = 1/α (-1>'(-1)" h
As long as (m, n) − [) is satisfied, complete reconstruction can be achieved.

On the other hand, the filter coefficient h' (m, n) that realizes the frequency characteristics shown in FIG. 4 is given as (21) by determining its impulse response. Note that the window function ω(m, n) is omitted.

Therefore, (m, n) is an even number other than (0, 0),
Or, in the case of a combination of odd numbers, h' (m,
n) is always 0, and as in the case of − dimension, h'
(m, n) as filter coefficient h (m, n)
It can be used as In addition, like the − dimension,
Considering that the DC component of the original signal should not be included in the information of the lower channel, if the filter coefficient is set so that H(π, π)−〇 holds, then α−2 is obtained.

In addition, when taking a sampling pattern as shown in FIG. 8 and applying this two-dimensional configuration processing to hierarchical encoding, the method of multiplication of filter coefficients in even-numbered hierarchical processing,
The sampling pattern needs to be a pattern obtained by rotating the odd-numbered hierarchical processing pattern by 45°.

The configurations of the digital filters 13 and 33 when using this two-dimensional filter are shown in FIGS. 9 and 10.

FIG. 9 is a block diagram showing the configuration of the two-dimensional transmitting side digital filter 13, which includes a vertical one-pixel delay element D.
A delay circuit is configured by a combination of H and a horizontal one-pixel delay element, and the tap output of this delay circuit is multiplied by a coefficient of a or b in multipliers 145 and 146, and added in an adder circuit 156. This is the configuration to be taken out. 10 is a block diagram showing the configuration of the receiving side digital filter 33. Corresponding to the configuration of the two-dimensional transmitting side digital filter 13 in FIG. It is configured to have

The coefficients a and b of this digital filter 13.33 are a = 0.297. b = −0,025.

This can be determined from the filter coefficients a and b of the one-dimensional digital filter shown in FIGS. 3 and 4 by setting a=axa and b=axb.

Next, FIG. 11 shows an example of a configuration on the transmitting side when the configuration shown in FIG. 1 is applied to hierarchical encoding.

This shows one-dimensional hierarchical processing on the sending side.

This hierarchical encoding process is performed using digital filter circuits 12-2, 12-3 and sub-sampling circuit 20- for the upper channel signal from which low frequency components of the original signal are extracted.
2.22-2.20-3.22-3 shows how filter processing and subsampling processing are performed one after another.

FIG. 12 shows a configuration for eliminating phase distortion when only upper channel information is displayed on the receiving side when the configuration shown in FIG. 1 is applied to hierarchical encoding.

That is, when displaying in stages starting from upper channel information, there is a problem that phase distortion occurs in the decoded image. This means that although it is possible to configure both the filters (h(n), g(n)) as linear phase filters, the filter H2, which is equivalent to the information processing of the lower channel on the transmitting side and the upper channel on the receiving side, This is because linear phase characteristics cannot be realized with (Z) and G + (z).

In order to solve this problem, as shown in Figure 12, another path is provided on the output side, and a digital filter (h(n)
) 39 and the image signal passed through this digital filter 39 is preferably used for stepwise display of the image.

In this way, it is possible to display images in stages similar to conventional hierarchical encoding, and the problem of phase distortion is avoided. Even if this separate path is provided, the effect of the present invention can be achieved because the number of filters is reduced compared to the conventional band division coding method.

Note that the filter coefficient of the added filter in FIG. There is a need. In this example, α-2
Therefore, the normalization factor is 0.5.

〔Effect of the invention〕

As explained above, in the present invention, the number of filters is halved compared to the conventional band division encoding method,
The amount of hardware can be reduced.

In addition, filter coefficients that realize complete reconstruction of the decoded signal can be easily obtained, making filter design easy and I
;Ru.

[Brief explanation of drawings]

FIG. 1 is a diagram explaining the system configuration of the present invention. FIG. 2 is a diagram showing a transmission method using the digital filter circuit of the present invention. FIG. 3 is a block diagram showing the configuration of a transmitting side digital filter. FIG. 4 is a diagram comparing the frequency characteristics of the embodiment digital filter with that of a conventional example. FIG. 5 is a block diagram showing the configuration of a receiving side digital filter. FIG. 6 is a diagram showing a two-dimensional configuration system using a repartition filter. FIG. 7 is a diagram showing a two-dimensional configuration system using a two-dimensional filter. FIG. 8 shows a sub-sampling pattern in the embodiment shown in FIG. FIG. 9 and FIG. 10 are block diagrams showing the configuration of a two-dimensional digital filter in a two-dimensional configuration system. FIG. 11 is a diagram showing the configuration of a transmitting side that uses a one-dimensional configuration system for hierarchical encoding. FIG. 12 is a diagram illustrating a configuration for eliminating phase distortion in hierarchical encoding. FIG. 13 is a diagram showing a conventional band division transmission system. FIG. 14 is a diagram explaining a conventional band division encoding system. 8-9.13.33.39...Digital filter,
lO...Low pass filter, 11...Analog-digital conversion circuit, 14...Multiplication circuit, 15...Subtraction circuit, 20.22-0. Subsampling circuit, 21.
23... Encoding circuit, 24... Multi-Flexer, 2
5... Multiplex transmission line, 26... Demultiplexer, 2
7.29...Decoding circuit, 28.3o...Interpolating circuit, 32.37...Adding circuit, 34...Synthesizing circuit, 3
8...Digital-to-analog conversion circuit.

Claims (1)

[Claims] 1. A first digital filter (h(n)) through which an input digital signal (x(n)) passes; a multiplication circuit (α) that receives the input digital signal; a subtraction circuit that takes the output of one digital filter and the output of the multiplication circuit as two inputs and calculates the difference between the two inputs, and the output of the first digital filter and the output of the subtraction circuit are used as two outputs. Digital filter circuit for signal. 2. A first digital filter (h(n)) through which the input digital signal (x(n)) passes, a multiplication circuit (α) that receives the input digital signal as input, and an output of the first digital filter. and a subtraction circuit that uses the output of the multiplication circuit as two inputs and calculates the difference between the two inputs, and uses the output of the first digital filter and the output of the subtraction circuit as two inputs, and a subtraction circuit that uses the output of the first digital filter and the output of the subtraction circuit as two inputs. an adder circuit that calculates a sum; a second digital filter through which the output signal of the adder circuit passes and whose pass band is substantially opposite to that of the first digital filter; and an output signal of the second digital filter and the first digital filter. 1. A digital filter circuit comprising: a synthesis circuit for synthesizing the output of the digital filter; 3. A first digital filter (h(n)) through which the input digital signal (x(n)) passes, a multiplication circuit (α) that receives the input digital signal as input, and an output of the first digital filter. and a subtraction circuit that takes the output of the multiplication circuit as two inputs and calculates the difference between the two inputs, and a transmission means that transmits the output of the first digital filter and the output of the subtraction circuit to two subband transmission lines. and an adder circuit which takes signals received from these two subband transmission paths as two inputs and calculates the sum of the two inputs, and an adder circuit through which the output signal of this adder circuit passes and which is almost opposite to the first digital filter. A transmitting/receiving device comprising: a second digital filter having a passband of , and a synthesizing circuit that synthesizes the outputs of the two digital filters and the output of the first digital filter. 4. The transmitting/receiving device according to claim 3, wherein the first digital filter and the second digital filter are two-dimensional filters that pass two-dimensional signals.