JPH02241217A - Suppressing or attenuating method and device for undersired spectrum in dynamic elongation of digitalized compressed signal - Google Patents

Abstract

PURPOSE: To suppress or damp an undesirable spectrum at the time of performing digital dynamic elongation by using a sampled frequency which is higher than that at the time of performing the subsequent digital processing on elongated signals. CONSTITUTION: Compressed analog signals are converted into digital signals at a frequency fA' higher than a normal sampled frequency fA and the digital signals are elongated at the raised sampled frequency fA'. The signals elongated at the raised sampled frequency fA' are supplied to a decimation filter DF from the output side AO of a dynamic expander EXP. The filter DF lowers the sampled frequency fA' to the normal sampled frequency fA used for the subsequent processing of the digital signals. Therefore, a defective non-harmonic spectrum can be suppressed or damped.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to a method and apparatus for suppressing or attenuating undesired spectra in the dynamic expansion of digitized and compressed signals.

A digital dynamic decompressor is described in the prior art German Patent Application No. 3815079. This dynamic decompressor is used to decompress a compressed analog signal recorded on a memory medium after analog-to-digital conversion. The premise of this method is a digital decompressor which satisfactorily reinstates the compression function used in the previous analogue compression, both statically and dynamically, ie digitally simulating a suitable analogue decompressor.

The decompressor is capable of restoring the original signal formed before compression, with some residual distortion removed. At the output of the analogue expander, such cases manifest themselves, for example, in the acoustic signal by the occurrence of predominantly odd-order harmonic spectral lines, which are not perceived as a significant disturbance in the auditory organ.

This harmonic spectrum causes only a slight change in the timbre of the acoustic signal. Digital decompression, on the other hand, generates a non-harmonic spectrum that can be very disturbing.

OBJECTS OF THE INVENTION It is an object of the present invention to provide a digital dynamic stretching method and device that suppresses or attenuates undesired spectra.

Means for Solving the Problem The above problem is solved by the features described in the characterizing part of claims 1 and 2. Advantageous embodiments are described in the further claims.

Effects of the Invention A disadvantage of digital dynamic stretchers is that they generate a disturbing non-harmonic spectrum. An advantage of the invention is that it is possible to suppress or attenuate such disturbance spectra as much as possible.

DESCRIPTION OF THE PREFERRED EMBODIMENTS The block circuit diagram shown in FIG. 1 shows a digital dynamic stretcher. Branch from input terminal EO 1.・・・
・, Q, A+1. ..., L branches, and the input sides of these branches are El, ..., EQ.

E! Q+1. ..., indicated by EL. Of these branches, only branch α is shown in detail. All these branches have the same configuration in principle, but generally differ due to various parameters that primarily determine their respective dynamic properties. Furthermore, the filters FS, FM provided at the input end for the division of the sub-frequency bands into various branches are different.

Filter FS and/or FM consists of partial filters connected in series or in parallel. This filter is used in canonical form as a cyclic digital filter (1-1, Goeckl
1. Tunable Digital Filters for Acoustic Technology (nLz Archive Vol. 7 (1985), Volume 3, pp. 47-57, Figure 2)) or state-space structure (Zus
(see West German Patent Application No. 3522411, No. 3522412, No. 3522413, No. 3439977).

Each branch branches into a signal path S and a measurement path M at one branch point Ab. In the signal path S, first the filter FS
Branch signal separation is performed by. A signal path multiplier SM acting as a regulating element weights (amplifies, attenuates, etc.) the signal to be expanded, which is here fed from the measuring path M to the second input J side of the signal path multiplier SM. Finally, the adjustment is performed using the adjustment amount SG derived from the signal from the input side EQ. Multiplication with a constant ks in the compensation multiplier MS is performed for compensation.

Finally, all the branch signals, i.e. the expander output A1',
..., AM, ...AQ+1. ..., all signals to AL are summed by a digital adder Ad to form a digitally expanded overall signal A at the output AO.

In the measuring path M, a further bandlimiting takes place in the filter FM after the branch point Ab. The output signal of the filter FM is transmitted to a measurement path multiplier M provided in the control circuit R.
one of the inputs of M. For compensation, the output signal of the measurement path multiplier MM is multiplied by a constant km in the compensation multiplier Mm.
weighted by Next, an absolute value forming operation (suppression of positive and negative signs) is performed in the absolute value forming unit B, and then a correction constant is added in the adder Al, and an amplitude limiter SI is provided in the control circuit R. and amplitude limiter S
L is followed by a logarithmizer LM. The transition from linear to logarithmic representation (αin/Qog) makes it possible to realize that the stretched characteristic is linear in the logarithmic domain.

The first operation in the logarithmic domain is the nonlinear transfer element NL
Level-dependent amplification in V. Then follows an adder A2 for the addition of the value 1o, and finally an integrator with an amplitude limiter Sz in the integrator control section and a delay element T' in the return path Rf to the adder A3. .

In the feedback path of the control circuit R, which is fed back to the measuring path node MM, the output signal of the integrator is weighted by -a.

For weighting, a multiplier is used as means M1. This is followed by an antilogizer DG, which inversely transforms the signal back into the linear value domain (Qog/Qin conversion). This is done in reverse to the operation for the linear/logarithmic transformation in the logarithmizer LM. The output signal of the antilogizer DG is used as a regulating variable in the control circuit R in the measuring path M.
That is, it is supplied to the measurement path multiplier MM.

Further, from the output signal of the integrator and thus of the control circuit R, an adjustment variable SG for the signal path S is weighted by a·(n-m)/m in the means M2 and then in the measuring path M. The antilogizer DG in the feedback path of the control circuit R of
is derived by performing antilogization in the antilogizer D2 corresponding to . Parameters n and m are static characteristics,
That is, determine the slope of the amplification characteristic curve of the digital expander/m. The dynamic characteristics of the digital expander can be determined by selecting the parameter a.

This device is designed such that at a constant input level 0 at the input side EQ, a signal is generated at the output side of the control circuit R that remains constant from sample value to sample value, that is, the adjustment amount SG also remains constant. works. However, these signals are level dependent, and therefore the signal path multiplier SM causes a larger signal level to be boosted more strongly in the signal path S than a smaller signal level (or a smaller signal level is boosted more strongly than a larger signal level). level). This is the purpose of the stretcher. In the level-up or level-down jump at the input 112, the control interval A certain transition characteristic must appear at the output of R. In this case, it must be taken into account which transition characteristics the compressor that generated the analog/digital converted input signal arriving at the inlet EO has. This transition characteristic of the compressor typically differs depending on whether it is an up-jump or a down-jump. In upward jumps the transition time is very short (in the millisecond range) and in downward jumps it is relatively long (in the second range).

In the control interval, first and foremost, a non-linear amplification effect that is level-dependent in the logarithmic domain is used, such that the expander shown in FIG. 1 exhibits dynamic properties that are matched to those of the corresponding compressor. A transfer element NLV is provided. The nonlinear transfer element NLV cooperates with a downstream integrator to determine the dynamic behavior of the stretcher in the level jump. The dynamic characteristic or damping in the downward jump is primarily performed by the constant 10 supplied to adder A2.

As an example, for example, West German Patent Application No. 38150
The description has been made on the basis of an embodiment for the digital dynamic stretcher on which the '79 patent is based. The following measures for suppressing or attenuating the disturbance spectrum can also be used in the dynamic stretcher described above, but also in the embodiments that are being modified from this.

In the signal processing in the digital dynamic expander, nonlinear operations such as the absolute value former B and the nonlinear amplification NLV in the embodiment shown in FIG. 1 are also performed. Such non-linear operation is responsible for the generation of harmonic oscillations which have non-harmonic oscillation characteristics and therefore have a disturbing effect on the used signal spectrum. The second method determines which disturbance spectrum is generated by the digital dynamic stretcher.
This will be explained using the spectrum diagram S shown in the figure. In this case fA
indicates the sampling frequency used to process the signal as usual after the digital dynamic stretcher.

This sampling frequency fA will be referred to below as the normal sampling frequency. The used spectrum, ie the spectrum of the signal to be expanded, is indicated by fl.

For example, if a signal with frequency f1 is applied to the input side Eff of the dynamic stretcher, this frequency fl
In the measurement path M, the harmonic frequency 21<fl (k= 1,
2°...) occurs. These spectra contained in the output signal SG of the measuring path M are processed in the multiplier SM of the signal path node at the frequency f1, which is not yet disturbed.
input signal. This allows the spectrum fh=
2kfl+f1=(2k +1)fl is generated. In the case of an analog dynamic stretcher, these spectra are only a slight disturbance, since they are essentially harmonics of the used spectrum f1. However, in digital dynamic decompression, if the sampling frequency in the decompressor is non-harmonic with respect to frequency f1, then the non-harmonic image frequency (f
A 3f1. 2fA-5fl,...) is obtained due to the periodic nature of the spectrum of the sampled signal. That is, a harmonic (2 + 1) fl, which is mainly an odd number, of the input signal frequency fl appears as a mirror image at (1/2 or 1 times) the sampling frequency or an integer multiple thereof. The harmonic spectrum (2 + 1) fl located mirror image above 1/2 of the sampling frequency is transformed into a non-harmonic spectrum. Measures should be taken to suppress or attenuate as strongly as possible these interfering non-harmonic spectra. In the present invention, such means provide a sampling frequency f higher than the normal sampling frequency fA for the digital dynamic stretcher.
It consists of using A'.

The following aspects are important in determining the ratio fx'/fA.

1. For practical reasons of implementation, the ratio fx'/fA is generally (but not necessarily) an integer >l.

2, fA'/fA is 1/2 of the sampling frequency (fA'/fA
2) and the image disturbance non-harmonic spectra at integral multiples of 1 (fA') must be selected such that the disturbances in these spectra have amplitude values so small that the disturbances are tolerably or negligible.

The assumption is that the amplitude of harmonic oscillations generally decreases with increasing order.

In the block circuit diagram shown in Fig. 3, first, the still analog-compressed signal is converted into a digital signal by an analog/digital converter A/D at a frequency fA' higher than the normal sampling frequency fA. and this digital signal is expanded by an increased sampling frequency fA'.

The circuit arrangement shown in FIG. 1, for example, can be used as the digital dynamic expander EX, P in this case. From the output AO of the dynamic expander EXP, a signal expanded at an increased sampling frequency fA' is supplied to a decimation filter DF. This decimation filter serves to reduce the sampling frequency fA' to the normal sampling frequency fA used during subsequent digital signal processing.

As shown in FIG. 4, the compressed analog signal is processed at the analog/digital converter A/D at a sampling frequency fAr that is less than, greater than, or equal to the normal sampling frequency fA.
It is also possible to convert with J. A subsequent interpolation filter IF increases the sampling frequency fA'' to fA', and thus the expansion of the signal to fA' is performed in a digital dynamic expander.The output signal of the expander EXP increases at that sampling frequency. , is reduced from fA' to fA by the aforementioned decimation filter DF.

The interpolation filter and decimation filter are R, E,
Crochiere, L. R1Rabiner1
'Multi-rate digital signal processing〃(Prent, jc
e 1llall Inc.

Co., Ltd., Engelwood Cliffs, N.J., 1983.
(annual).

In the digital dynamic stretcher shown in FIG. 1, it is advantageous to increase the sampling frequency from fAJJ to f'(fA'>fA) only after the input horn filter FS. This is because the lower frequencies of the signals to be processed by the filter require lower filter costs. Therefore, as shown in Figure 5,
An interpolation filter IF is connected downstream to the input filter FS.

Reducing the sampling frequency of the decompression horn signal from fA' to fA in the corresponding decimation filter DF preferably reduces the output J signal corresponding to the partial frequency band of many parallel connected decompression units. This is done on the output side of the adder Ad that adds the . In this way, the outputs A 1 ′, ..., AQ, of the individual stretcher units
AQ+ 1. ...It is not necessary to provide one decimation filter on each output side of the AL.

A further improvement with respect to the suppression of the disturbance spectrum is that the sampling frequency for the control circuit R in the measurement path M can be changed in particular to the signal path S.
It is obtained by increasing the sampling frequency fA'' to fA''>fA' compared to the sampling frequency fA' at .

As shown in FIG. 6, for this purpose an interpolation filter IFI is connected downstream of the filter FM in the measuring path M, which reduces the sampling frequency fA' to fA''. The filter DPI again causes the increased sampling frequency fA9 to be reduced again to the lower sampling frequency fA' used in the stretcher.

[Brief explanation of drawings]

Fig. 1 is a block circuit diagram of a digital dynamic expander, Fig. 2 is a spectrum diagram, and Fig. 3 is a spectral diagram. 4 and 5 are block diagrams of various variants for changing the sampling frequency in a dynamic stretcher;
The figure shows a block circuit diagram of a stretcher-measuring path with means for changing the sampling frequency. AO...Output side, AI, A2. A3...adder. Ab...branch point, Ad...digital adder, B...
- Insulation value former, D2. DC: Antilogizer, DF: Decimation filter, Eo: Input terminal, EIEL
...Input side, EXP...Digital dynamic expander + fl...Spectrum of the signal to be expanded. fA...sampling frequency, IF...interpolation filter, I. ...value, km...constant, LM...logarithmizer 1M.
...Measuring path, Mm...Compensation multiplier, MM...Measuring instrument multiplier. NLV...nonlinear transfer element, R...control circuit, Rf
...Feedback path, S...Signal path, sr, sz...Amplitude limiter, SM...Signal path multiplier, SG...Adjustment amount, r'...Delay element. Agent: Satoshi Yano, Patent Attorney I.

Claims (1)

[Claims] 1. A method for suppressing or attenuating undesired spectra in dynamic stretching of a digitized and compressed signal, comprising: in digital dynamic stretching a higher sampling frequency than in digital subsequent processing of the stretched signal; A method for suppressing or attenuating undesired spectra in dynamic stretching of digitized and compressed signals, characterized in that it is used. 2. In a device for suppressing or attenuating undesired spectra in a digital dynamic expander (EXP), the frequency of the signal to be expanded is First means of increasing the sampling frequency (IF)
and a second means (DF) for reducing the sampling frequency (f_A') to the normal sampling frequency (f_A) after stretching. . 3. Digital dynamic expander (EXP) connects the signal path (
S) and a measurement path (M), the measurement path (M) branching from the signal path (S) has a control circuit (R);
) is connected to the signal path node (SM) provided in the middle of the signal path (S), and the output side of filter (FS
), adjust the pass region of the filter (FS) to the frequency band where dynamic expansion is performed, and
3. The device for suppressing or attenuating undesired spectra in a digital dynamic stretcher according to claim 2, characterized in that means (IF) for increasing the sampling frequency are provided between and the branching point (Ab). 4. Claim characterized in that the signal fed to the filter (FS) has a sampling frequency (f_A") of equal magnitude or lower or higher than the normal sampling frequency (f_A) 3. Device for suppressing or attenuating undesirable spectrum in the digital dynamic stretcher according to 5. In the middle of the measurement path (M), the sampling frequency (f_A′″) of the signal supplied to the control circuit (R) is Signal path (S)
The measurement path (M) connected to the signal path node (SM) is higher than the sampling frequency (f_A′) used in
4. An undesirable signal in a digital dynamic expander according to claim 3, characterized in that the reduction of the sampling frequency (f_A''') to the sampling frequency (f_A') is carried out in the signal path (S) on the output side of the Spectral suppression or attenuation device.