CH709272A2 - Autonomous residual determination and extraction of low-residual additional signals. - Google Patents

Abstract

The invention relates to a method for decoding or upmixing an upmix signal having at least five channels from a downmix signal having at least three channels. The method comprises the steps of performing at least two correlation comparisons between at least two channel pairs of the at least three channels of the downmix signal, each correlation comparison of one of the at least two channel pairs yielding at least one correlation comparison signal, wherein the at least one correlation comparison signal of one of the at least two channel pairs is a common signal of one of the channel pairs and / or a first individual signal of one of the channel pairs and / or a second individual signal of one of the channel pairs, the common signal of one of the at least two channel pairs containing the common signal components of this channel pair and / or the first individual signal of one of the at least two channel pairs, the signal portion specific to the first channel of the channel pair and / or the second individual signal of one of the at least two channel pairs containing signal portions specific to the second channel of the channel pair; Determining a common residual signal based on the common signal portions of the correlation comparison signals of the at least two channel pairs; and correcting at least one of the correlation comparison signals or a signal formed therefrom on the basis of the common signal.

Description

[0001] Multi-channel signals and in particular three-dimensional signals such as audio signals, for example, place high demands on the amounts of data to be transmitted or stored, which can be efficiently reduced.

General known devices or methods for such data reduction are parametric methods here that extract spatial information, for example using the Fast Fourier Transform (FFT) known from the prior art, and then as a permanent data stream, for example together with a mono or stereo signal as a downmix signal. Such an audio technology is known in particular with MPEG Surround and, from a mathematical point of view, represents an adaptive filter method.

WO 2014/072 513, which is exemplified with reference to FIG. 17 is briefly explained, on the other hand, proposes for the first time the use of so-called non-linear inverse coding, this with the additional use of correlation comparison; Since only the parameters of a spatial model have to be transmitted permanently as header information or data pulse, this approach can mathematically be viewed as a static filter method.

However, WO 2014/072 513 does not give an explicit technical solution for such a correlation comparison, since devices or methods known from the prior art exist, such as the UPM1 upmix system from the British company Soundfield, which is also based on FFT , requires a high computational effort overall. The formation of the output signals is, however, amplitude-dependent, which has the disadvantage of moving sound sources and temporally shifting artifacts on the individual channels, and overall leads to significant spectral discoloration, which has a disruptive effect in the area of audio coding.

A simple device or a simple method for such a correlation comparison, which also fulfills the goal of the highest possible psychoacoustic transparency, is disclosed with the unpublished own application CH1696 / 13. It is also demonstrated that when a correlation comparison is applied to a chain of signals, a correction by the mean value of all residuals of this correlation comparison is algebraically the optimal solution. However, with such a correlation comparison with subsequent correction by the mean value of all residuals, this mean value must be added to the compressed signal to be transmitted, which leads to a substantial, ultimately undesirable increase in the bandwidth. The extraction of further signals from elements of the signal chain by means of correlation comparison is also difficult. In the worst case, secondary residuals occur, which simply means that residuals are subjected to a correlation comparison again and the most disruptive artifacts are formed.

The present invention should therefore on the one hand enable an autonomous determination of interfering residuals without having to transmit them, and on the other hand additional signals should be obtained so that on the one hand no secondary residuals form, and on the other hand an optimal masking of the primary residuals or other artefacts occurring through signal processing can be achieved.

[0007] In particular, the autonomous residual determination represents a simple device or a simple method which leads to drastic bandwidth savings with the highest level of transparency.

It is not limited to audio signals, although such a system can be optimally applied to these in particular. For example, video signals can be efficiently compressed and decompressed with the subject matter of the invention - or their residuals can be efficiently minimized.

In general, the subject matter of the invention can be used in all signaling technology, provided that it is based on a Fourier transformation or an inverse Fourier transformation, and leads there, for example, to drastic bandwidth savings or enables efficient data extraction.

With the present invention, for example, in audio coding, the number of downmix channels can be limited to a minimum, since the mixed channels can in turn be isolated by correlation comparison and thus enable efficient storage and transmission of highly complex audio signals overall.

In particular, such highly complex 3D audio signals, as they are known for example with the NHK 22.2 format, can now be summarized in corresponding downmix channels, which in turn can be partially or entirely subjected to such a correlation comparison, and in which there are no residuals with even improved quality must be transferred.

For example, within the top layer of an NHK-22.2 system, see FIG. 21, or a format similar to the so-called Auro formats of the Institute for Broadcast Technology in Munich, see FIG. 11, extract a TpC in purely algebraic form or transmit and extract an LFE signal together with the downmix of such formats using the simplest filter techniques (for example using IRR filters known from the prior art with subsequent all-pass correction).

[0013] The following documents are particularly to be regarded as belonging to the state of the art:

EP 1 850 639 describes a static filter which generates a stereo signal from a mono signal. This filter can also be applied to multi-channel signals.

[0015] WO 2009 138 205 describes a static filter which generates a stereo signal from a mono signal. This filter can also be applied to multi-channel signals.

WO 2011 009 649 describes an expansion of the static filters described in EP1 850 639 and WO 2009 138 205 for adapting the degree of correlation of the stereo signal generated in each case. This extension can also be applied to multi-channel signals.

WO 2011 009 650 describes extensions of the devices or methods described in EP 1 850 639 and WO 2009 138 205 and WO 2011 009 649 in order to optimize the respectively generated stereo signal with regard to the static parameters. These extensions can also be applied to multi-channel signals.

WO 2012 016 992 describes the first practical use of algebraic invariants in general in signaling technology and in particular on EP 1 850 639, WO 2009 138 205, WO 2011 009 649 and WO 2011 009 650.

WO 2012 032 178 describes the time scaling of static filters according to EP 1 850 639, WO 2009 138 205, WO 2011 009 649, WO 2011 009 650 and WO 2012 016 992.

The cited documents show all aspects of what is known as linear inverse coding.

[0021] WO 2014 072 513, which is exemplified with reference to FIG. 17, describes extensions of these static filters for their targeted application to multi-channel signals, this as so-called nonlinear inverse coding, this also using direct correlation comparison, which can take place, for example, using the upmix system ÜPM1 from the British company Soundfield.

The unpublished patent application CH1696 / 13 describes a Fourier transformation or inverse Fourier transformation based, highly efficient system for the correlation comparison of two or more input signals and describes the algebraically optimal correction based on the mean value of the residuals occurring. However, an autonomous residual determination is not possible with this document, and an algebraic extraction of additional signals while avoiding secondary residuals is unknown.

HAMASAKI KIMIO ET AL :. "The 22.2 Multichannel Sound System and Its Application", AES CONVENTION 118, MAY 2005, describes a channel-based playback format of the highest order and spatial resolution.

As a standard, MPEG Surround describes the use of so-called parametric methods for the transmission of multi-channel signals based on a mono or stereo signal.

DISCLOSURE OF THE INVENTION

Introduction

In our own unpublished application CH 1 696/13, see below, an extraction of multi-channel signals using a Fourier transformation and inverse Fourier transformation is described, and an optimization of the resulting signals using the determined mean value of the residuals is proposed. However, this means that a residual reduction is only possible if this mean value is known beforehand, i.e. has been transferred. In particular, the document does not give any explicit technical solution for an autonomous residual determination in the decoder itself, and therefore no devices or methods of this type known from the prior art exist.

In CH1696 / 13, such a correlation comparison of two signals Li 'and Ri', in which respectively identical signal components x (t) and y (t), which are for the short-term cross-correlation

have the degree of correlation +1, which on the one hand represents a mathematically exact solution for time-invariant (stationary) signals, and has a specific residual behavior for time-variant (non-stationary) signals (a residual being the difference between the original, non-stationary signal segment and its Fourier transformation represents).

In CH1696 / 13, two channels Li ', Ri', 1 ≤ i ≤ n, which have similar signal components Ci *, are considered, where: Li ́ = Li * + Ci * = Ii ́ (t) = Ii * (t) + ci * (t) Ri ́ = Ri * + Ci * = ri ́ (t) = ri * (t) + ci * (t)

The Fourier series are now determined for the time-dependent signals Ii '(t) and ri' (t). Hence for the synthesis, k =…, –1, 0, 1, ...

and for analysis

and in practice for the discrete Fourier transforms (DFT), from which the Fast Fourier Transforms (FFT) can be derived directly, where now k = 0, ..., N - 1:

The real parts of Li, Ri and Cilassen for stationary signals for all k = 0, ..., N-1 then according to the following rules

Recover: 1. Determine the signs of the real parts of Li '(k) and Ri' (k). 2. If the signs for k are identical, determine - the amounts of the real parts of Li ́ (k) and Ri ́ (k), - the minima or maxima of these amounts of the real parts of Li ́ (k) and Ri ́ (k). - Choose the real part of Li ́ (k) or Ri ́ (k) on which this minimum is based as the real part for Ci (k). - Subtract the real part of Ci (k) from the maximum underlying real part of Li ́ (k) or Ri ́ (k) and, if the real part of Li ́ (k) is based on this maximum, choose the result of this subtraction as the real part for Li (k), otherwise, if the real part of Ri ́ (k) is based on this maximum, the result of this subtraction as the real part for Ri (k). - Set the not yet determined real part of Li (k) or Ri (k) equal to zero. 3. If the signs of the real parts of Li ́ (k) and Ri ́ (k) are not identical, set Ci (k) to zero and set Li (k) = Li ́ (k) and Ri (k) = Ri ́ ( k).

The imaginary parts of Li, Ri and Cil can be recovered for stationary signals for all k = 0, ..., N - 1 according to the following rules: 1. Determine the signs of the imaginary parts of Li '(k) and Ri' (k). 2. If the signs for k are identical, determine - the amounts of the imaginary parts of Li ́ (k) and Ri ́ (k), - the minima and maxima of these amounts of the imaginary parts of Li ́ (k) and Ri ́ (k). - Choose the imaginary part of Li ́ (k) or Ri ́ (k) on which this minimum is based as the imaginary part for Ci (k). - Subtract the imaginary part of Ci (k) from the imaginary part of Li ́ (k) or Ri ́ (k) on which the maximum is based and choose, if the imaginary part of Li ́ (k) is based on this maximum, the result of this subtraction as the imaginary part for Li (k), otherwise, provided the imaginary part of Ri ́ (k) is based on this maximum, the result of this subtraction as the imaginary part for Ri (k). - Set the not yet determined imaginary part of Li (k) or Ri (k) equal to zero. 3. If the signs of the imaginary parts of Li ́ (k) and Ri ́ (k) are not identical, set Ci (k) equal to zero and set Li (k) = Li ́ (k) and Ri (k) = Ri ́ ( k).

In order to finally win Li, Ri and Cizu, for the synthesis for the time-dependent signals, k = ..., -1, 0, 1, ...,

(or in practice for the analysis using discrete Fourier transforms (DFT), k = 0, ..., N - 1,

from which the Fast Fourier Transforms (FFT) can be derived directly) for the synthesis the coefficients fk, gk, hk determined, k =…, -1, 0, 1,…, according to the analysis

or for the synthesis according to the inverse discrete Fourier transform (IDFT), from which the inverse fast Fourier transforms (IFFT) can be derived directly, k = 0, ..., N - 1,

The schematic sequence of such a correlation comparison is shown in FIG. 1 shown as an example:

For each channel of the time-dependent downmix signal Li (t), Ri (t), a Fast Fourier Transform (FFT) is first carried out, and the frequency-dependent complex-valued signal descriptions Li (k) and Ri (k) are thus obtained. The rules for obtaining the real parts and the imaginary parts from Li, Ri and Ciang are now applied to these. An Inverse Fast Fourier Transform (IFFT) is then applied to the resulting signal descriptions Li (k), Ri (k) and Ci (k). The time-dependent signals ci (t), Ii (t) and ri (t) result.

For non-stationary signals, a residual Δ occurs in this form of correlation comparison, which generally has the following behavior:

This residual plays psychoacoustically, provided it is the pure reproduction of Li, Ri and Cinach of an inverse discrete Fourier transform (IDFT), from which the inverse fast Fourier transform (IFFT) can be derived directly, within a normative listening situation ( acts within the "sweet spot") does not matter, since the residual is wiped out. Outside of the “sweet spot”, as occurs in listening situations in everyday life and in non-normative loudspeaker setups, however, clearly audible artifacts can occur which must be avoided.

Depending on the application, it is therefore desirable to determine such residuals, in particular in a decoder, in order to drastically reduce the bandwidth of the signal to be transmitted from the encoder to the decoder, and in particular to drastically discolor the timbre compared to the prior art to minimize.

The residual Δ itself can be approximately frequency-dependent (the Fourier transform for Li or Ri or Ci is already known, and it is therefore only in the same way as described for Li or Ri or Ci, the Fourier transform for Li * or Ri * or Ci *) for each frequency k as follows (with frequency-dependent calculation, an inverse discrete Fourier transform (IDFT) may be required for Δ (k), from which the inverse fast Fourier transform (IFFT) can be derived directly, such as for Li or Ri or Ci, to be carried out): Δ (k) = Li (k) - Li * (k) or Δ (k) = Ri (k) - Ri * (k) or Δ (k) = 1/2 * (C * i (k) - Ci (k))

According to the prior art, to obtain a residual-free signal by means of correlation comparison from two channels Li 'and Ri', 1 ≤ i ≤ n, which have similar signal components Ci *, the associated residual Δ must also be known, which is outside the decoder must be determined, which is a major restriction in audio coding, for example, with regard to optimally reduced bandwidths, since the absolute number of channels to be transmitted to the decoder cannot be reduced to the extent that such a residual must be explicitly added in the data stream for every correlation comparison. Such a residual must be determined as a function of frequency or time, for example if the common signal Ci (t) determined by correlation comparison or the first individual signal Li (t) determined by correlation comparison or the second individual signal Ri (t) determined by correlation comparison is present as a function of time.

In principle, such residual-free signals can be obtained by simple subtraction or addition, both frequency-dependent and time-dependent: Li * = Li - Δ Ri * = Ri-Δ Ci * = Ci + 2Δ

However, it can be shown that for example for adjacent channels Li, Ci1, Ri, Ci2 and Bimit Li ́ = Li * + Ci1 * Ri1 ́ = Ri * + Ci1 * + Ci2 * Ri2 ́ = Ri * + Ci1 * + Ci2 * Bi ́ = Bi * + Ci2 * the residual Δ2, which results from the correlation comparison between Ri2 ́ and Bi ́, cannot be derived in linear form from a residual Δ1, which results from the correlation comparison between Li ́ and Ri1 ́.

An ideal approximate determination, which also represents the drastic reduction in the number of residuals to be transferred, consists in the following consideration:

If n residuals Δ1, Δ2, Δ3, Δ4, ..., Δn were determined, see for example FIG. 2 or 22, and apply to the differences Δ2– Δ1 = η3– ηn Δ3– Δ2 = η4– η1 Δ4 - Δ3 = η5- η2 ... Δn-Δn-1 = η1- η2-2 Δ1- Δn = η2- ηn-1 then let the relationships Δ1 = η2 - ηn-1 + Δn Δ2– Δ1 = Δ2– (η2– ηn – 1 + Δn) = η3– ηn or. Δ2 = η3– ηn + (η2– ηn – 1 + Δn) Δ3– Δ2 = Δ3– (η3– ηn + (η2– ηn – 1 + Δn)) = η4– η1 or Δ3 = η4– η1 + (η3– ηn + (η2 - ηn – 1 + Δn)) Δ4 - Δ3 = Δ4 - (η4– η1 + (η3 – ηn + (η2– ηn – 1 + Δn))) = η5– η2 or. Δ4 = η5– η2 + (η4– η1 + (η3– ηn + (η2– ηn – 1 + Δn))) Δ1- Δn = (η2- ηn-1 + Δn) -Δn = η2- ηn-1 or. Δ1 = (η2– ηn – 1 + Δn) derive.

Thus, for example, (η2− ηn − 1 + Δn) is a term contained in all residuals. The same consideration can be applied to every Δ1, i = 1,…, n. Since n remains small in practice, it follows that with each term determined in this way, each residual can be approximated with high accuracy.

If one now sets Δ2– Δ1 = η3– ηn = a1 Δ3– Δ2 = η4– η1 = a2 Δ4– Δ3 = η5– η2 = a3 Δ5– Δ4 = η6– η3 = a4 Δ6– Δ5 = η7– η4 = a5 Δ7– Δ5 = η8– η5 = a6 Δ8- Δ7 = η9- η6 = a7 Δ9- Δ8 = η10- η7 = a8 Δ10- Δ9 = η11- η8 = a9 ... Δ1– Δn = η2– ηn – 1 = an surrendered η3 = ηn + a1 η6 = ηn + a1 + a4 η9 = ηn + a1 + a4 + a7 ... or. η4 = η1 + a2 η7 = η1 + a2 + a5 η10 = η1 + a2 + a5 + a8 or. η5 = η2 + a3 η8 = η2 + a3 + a6 η11 = η2 + a3 + a6 + a9 ...

From the relationships η2 = η2 + a1 + a4 + a7 + ... + an – 2 or ηn – 1 = ηn – 1 + an + a3 + a6 + ... + an – 3 or. ηn – 2 = ηn – 2 + an – 1 + a2 + a5 + ... + an – 4 the "difference rule for residuals" (a1 + a4 + a7 + ... + an – 2) + (an + a3 + a6 + ... + an – 3) + (an – 1 + a2 + a5 + ... + an – 4) = 0 derive, which means that a1, a2, a3, ..., anpsychoacoustically within a normative listening situation (within the “sweet spot”) show an ideal residual behavior in that they cancel each other out.

The following “addition theorem for residuals” can be derived directly from the above “difference rule”, namely Δ1 + Δ2 + Δ3 + ... + Δn = n * (η2– ηn – 1 + Δn) which means that the mean value of all residuals is contained in these at the same time and can also be calculated easily, for example within an encoder.

The principle of a residual correction, for example for the top layer of an NHK-22.2 signal, is shown in FIG. 19th

If the residual correction is now made using the mean value instead of the residuals Δ1, Δ2, Δ3, ..., Δnnunmehr, discoloration of the timbre and other unmasking effects can be minimized and a significant widening of the sweet spot can be achieved, but this method is possible a residual determination outside the decoder is necessary.

According to CH1696 / 13 it is approximately according to FIG. 2 possible, whose circumscribed triangle with its corners denotes the number of downmix channels and whose inscribed, dashed triangle denotes the number of channels additionally extracted by means of correlation comparison (which are then subtracted from the associated downmix channels in order to approximately obtain all the original channels of the circumscribed triangle), when transmitting the mean of all residuals, to extract a maximum of six channels from at least three channels, which have significantly fewer artifacts or discoloration of the timbre and other unmasking effects. In other words, the corner points of the circumscribed triangle describe the three downmix channels of a multi-channel signal with six channels. A corner of the inscribed triangle describes a channel of the multi-channel signal that is added to the two adjacent downmix channels. Such a further channel lying between two downmix channels can be obtained by a correlation comparison between the two adjacent downmix channels (the corner points of the circumscribed triangle) by extracting this signal contained in both downmix channels, as well as the sum of the two original, outer corner signals the downmix with its neighboring signal, which lies in the middle of the side of the triangle closest to the respective corner signal (mind you, not on the side on which the first correlation comparison was carried out!). If a correlation comparison is now also carried out for the two downmix channels of this newly viewed page, the signal contained in both downmix channels is again extracted. This can be subtracted from the closest sum of the first correlation comparison and then results in the original corner signal. If this is done for all three adjacent pairs of downmix channels, the six channels of the multi-channel signal are obtained again. Since, in addition to the downmix signals, if not stationary, three residuals would have to be transmitted in addition to the three downmix channels in order to precisely calculate the six multi-channel signals, the amount of data to be transmitted would again be the same as the transmission of the multi-channel signal. Therefore, the mean value of all three residuals is now transmitted and the six channels of the multi-channel signal obtained from the correlation comparison are corrected on the basis of this averaged residual.

The FIG. 2 has a purely combinational meaning and should not be confused with specific loudspeaker positions.

FIG. 18 geometrically illustrates the combinational application of such a correlation comparison to a corresponding downmix with four channels, which is illustrated by the circumscribed square. The same considerations apply to them as to FIG. 2.

The system described can be extended to any n-gon to infinity, although the calculated mean value of all residuals increasingly leads to artifacts or discoloration of the timbre and other unmasking effects due to the above considerations.

The downmix channels or residuals, which are intended to store and transmit audio data as efficiently as possible, for example between an encoder and a decoder, can be converted into a corresponding lossless or lossy base audio codec known from the prior art (examples of a Such Base Audio Codecs are Opus or the MPEG standards MP3, AAC, HE-AAC, HE-AAC v2 and USAC) additionally compress and decompress, whereby the Base Audio Codec used in each case can be additionally optimized with regard to the overall underlying spatial coding method (« Tuning »).

In particular, since a number of audio codecs for lossless or lossy compression of audio signals already make use of the Fourier transform or Fast Fourier Transform (FFT), the rules described above for obtaining the real parts or imaginary parts from can be used with little computing effort Signals are integrated directly into these audio codecs, or signals are derived from such audio codecs that can be subjected to these rules for obtaining the real parts or imaginary parts. The total computational effort required can thus be significantly reduced.

Autonomous residual determination by Fourier transformation

Two signals that have been subjected to the same correlation comparison described are "pseudo-stationary" after this correlation comparison due to the residuals formed, which means that after

Correlation comparison their behavior for the same frame as it was used for the Fourier transformation or inverse Fourier transformation is identical to that of a stationary signal. The reason for this is the previous representation in the frequency domain, which in practice only allows insufficient temporal resolution for non-stationary signals and at the same time leads to undesirable residual formation. Thus, no second correlation comparison described above can be carried out on the output signals of such a correlation comparison. Pseudo-stationary signals have become “immune” to this.

Likewise, the determination of the mean value of all residuals formed by means of correlation comparison, viewed algebraically, depends on the knowledge of the original signals. In particular, there is no method of algebraically reconstructing such a mean without knowing the original signals. Thus, according to the state of the art, it is necessary to transmit the mean value together with the signals.

The invention enables this also by means of Fourier transformation, although not the absolute determination of this mean value of all residuals formed, but can, depending on the frequency, obtain those signal components which have the degree of correlation +1 from signals that cannot be assigned to the same correlation comparison. Thus, depending on the frequency, the above-described signal component with a degree of correlation +1 of two residuals Δ1 and Δ2 can be extracted absolutely for each sample, and the resolution of the recovered signals is essentially only dependent on the resolution of the Fourier transformation or the subsequent inverse Fourier transformation. It is an analogy to the mean value of all formed residuals, but here the maximum value Δmax is extracted for each sample of two compared signals, so to speak the analytical analogue to the mean value of all formed residuals - with the only difference that the maximum value Δmax is determined without knowledge of the previous signals can be.

A correlation comparison is therefore to be applied to two output signals of non-identical correlation comparisons i, j which have a correlating component Δijmax in their two residuals Δi and Δj. In practice this means that for Li = Li * + Δi Ri = Ri * + Δi Ci = Ci * - 2Δi and Lj = Lj * + Δj Rj = Rj ́ + Δj Cj = Cj * - 2Δj preferably one of the following signal pairs should be subjected to a further correlation comparison: Liund Lj, or Riund Rj, or Liund Rj, or Riund Lj, or Ciund Cj, or - Liund –Lj, or - Riund –Ri, or - Liund -Rj, or - Riund –Lj, or -Ci and -Cj, or - * Liund 0.5 * Cj, or - * Riund 0.5 * Cj, or Liund –0.5 * Cj, or also Riund -0.5 * Cj.

Each of the aforementioned correlation comparisons provides the same result Δijmaxbzw. -Δijmax.

For example, for a further correlation comparison k that is not identical to the correlation comparisons, the pairs can now be identified Liund Lk, or Riund Rk, or Liund Rk, or Riund Lk, or Ciund Ck, or –Liund –Lk, or –Riund –Rk, or –Li and –Rk, or –Riund –Lk, or –Ci and –Ck, or - * Liund 0.5 * Ck, or - * Riund 0.5 * Ck, or Liund –0.5 * Ck, or also Riund -0.5 * Ck form as well Ljund Lk, or Rjund Rk, or Ljund Rk, or Rjund Lk, or Cjund Ck, or –Ljund –Lk, or –Rjund –Rk, or –Ljund –Rk, or –Rjund –Lk, or –Cjund –Ck, or - * Ljund 0.5 * Ck, or - * Rjund 0.5 * Ck, or Ljund –0.5 * Ck, or also Rjund -0.5 * Ck.

The correlating components are thus obtained for the correlation comparison of the output signals of the correlation comparisons i, k

[0065] Δikmax or –Δikmax and for the correlation comparison of the output signals of the correlation comparisons j, k, the correlating components

[0066] Δjkmax or –Δjkmax.

On Δijmax and Δikmax or. -Δijmax and -Δikmax or also to Δijmax and Δjkmax or -Δijmax and -Δijmax can now be carried out in the same way a correlation comparison, resulting in the correlating component Δijmax, which represents the signal component with a degree of correlation +1 for the three previous correlation comparisons.

This system can, as FIG. 3 and 4 show, expand ad infinitum, and furthermore allows a multitude of equivalent combinatorial possibilities with regard to the post-connection of correlation comparisons described here, as FIG. 3, 4 and 7 illustrate.

Instead of correcting the signals obtained by means of correlation comparisons by the mean value of all residuals in the manner described above, the result of the subsequent connection of correlation comparisons described here, for example in the case of FIG. 2 Δijkmax, inserted into the systems of equations for residual correction, see above. In the case of FIG. 18 this process is illustrated, for example, by FIG. 19 illustrated.

The result provides very little discoloration, signals extracted by means of correlation comparison, in particular for the sweet spot.

In the case of signals with very high-frequency components, this can lead to an unpleasant amplification of the waveforms. This can either be countered by an increased resolution of the discrete Fourier transformation used, or also by its opposite, namely a discrete Fourier transformation, due to the dimensions of which the harmonic frequencies are clearly below the high-frequency components mentioned. In the second case, the procedure is similar to that for the mean value of all residuals. Thus, for example, the length of a Fast Fourier Transform (FFT) can be used to precisely control the signal behavior.

Such a system can be integrated into a decoder with very little computational effort, and a transmission of the mean value of all residuals can thus, as already noted, be dispensed with.

While the present methodology delivers surprisingly natural results, especially in the sweet spot, the residual correction by the mean value of all residuals is an alternative in particular when the largest possible listening area (sweet spot) is to be created - but with higher bandwidths when transmitting Signals from the encoder to the decoder.

Extraction of low-residual additional signals

In the case of multi-channel signals, additional signals (such as the TpC of an NHK-22.2 signal or the TpC of an Auro signal developed by the Institute for Broadcasting Technology (IRT)) can occur which, according to the state of the art, are not advantageous due to optimized residual behavior with regard to occurring Have artifacts masked. Practical examples of such signals are the TpC signal of an MHK-22.2 signal or a similar TpC signal of an Auro format developed by the Institute for Broadcasting Technology - or so-called LFE signals, which only differ due to the high-energy, low frequencies and special setup difficult to mask.

In the first case, however, due to a surprising algebraic relationship, the summation of the correlation comparisons of the two diagonals in FIG. 5 derived TpC signals, this with additional optional residual corrections, and the final subtraction of the TpC thus formed in each case with one of the resulting center signals of the correlation comparisons of the four sides of the square, see FIG. 5, this with an additional optional residual correction, see below, gain signals that mutually mask each other with regard to the primary residuals occurring.

The reason for this is on the one hand the fact that, for example, four corner speakers TpFL, TpFR, TpBL and TpBR, which are arranged symmetrically around the head and which reproduce a TpC signal with -6dB each, lead to a sound field in the case of crosstalk and to a localization of this signal in the Area of the TpC, and on the other hand the following derivation, which is based on targeted algebraic, maximized elimination of interfering signal components and residuals according to the following scheme (see associated table in FIG. 6): 1. Carry out a correlation comparison between the downmix signals TpFL ́ and TpBR ́ to obtain an approximate TpC signal from (line 1 in the associated table of FIG. 6). 2. Carry out a correlation comparison between the downmix signals TpFR ́ and TpBL ́ to obtain an approximate TpC signal (line 2 in the associated table of FIG. 6). 3. If necessary, determine the autonomous residual of steps 1 and 2 according to FIG. 7th Or alternatively: Read the mean of all residuals that result from steps 6 and 8 as well as 13 and 15. 4. Perform a residual correction on the result of step 1. 5. Perform a residual correction on the result of step 2. 6. Carry out a correlation comparison between the downmix signals TpBL ́ and TpFL ́ to obtain an approximate (TpSiL + TpC) signal (line 6 in the associated table of FIG. 6). 7. If necessary, carry out a residual correction to the result of step 6. 8. Carry out a correlation comparison between the downmix signals TpFR ́ and TpBR ́ to obtain an approximate (TpSiR + TpC) signal (line 4 in the associated table of FIG. 6). 9. If necessary, carry out a residual correction on the result of step 8. 10. Add the signals obtained from steps 1 and 2 together. 11. Add up the signals obtained from steps 6 to 9. 12. Subtract the signals obtained in step 11 from the signals obtained in step 10. The result is the signal with minimized residuals r ́: -0.5 * TpSiL - r ́ (0.5 * TpSiL) - 0.5 * TpSiR - r ́ (0.5 * TpSiR). 13. Carry out a correlation comparison between the downmix signals TpFL ́ and TpFR ́ to obtain an approximate (TpFC + TpC) signal (line 3 in the associated table of FIG. 6). 14. If necessary, carry out a residual correction on the result of step 13. 15. Carry out a correlation comparison between the downmix signals TpBR ́ and TpBL ́ to obtain an approximate (TpBC + TpC) signal (line 5 in the associated table of FIG. 6). 16. If necessary, carry out a residual correction on the result of step 15. 17. Add the signals obtained from steps 13-16. 13. Subtract the signals obtained in step 17 from the signals obtained in step 10. The result is the signal with minimized residuals r ́: -0.5 * TpFC - r ́ (0.5 * TpFC) - 0.5 * TpBC - r ́ (0.5 * TpBC). 19. Add the signals obtained using steps 11 and 17 with the signals obtained using steps 12 and 18. The TpC signal we are looking for results with minimized primary residuals: TpC - 2 * r ́ (TpFL) - 2 * r ́ (TpFR) - 2 * r ́ (TpBL) - 2 * r ́ (TpBR) - 2 * r ́ (0.5 * TpFC) 2 * r ́ (0.5 * TpSiR) 2 * r ́ (0.5 * TpBC) - 2 * r ́ (0.5 * TpSiL). 20. Subtract half of the signal obtained in step 19 from the signals obtained in steps 6 to 9 and 13 to 16, respectively.

The overall procedure describes in a simplified manner (without steps 11, 12 and 17, 18 and 19) the busbar diagram of FIG. 16, in which each input channel marked in bold and each resulting channel are shown separately. For the input channels TpFL ́, TpFR ́, TpBR ́, TpBL ́ and the determined mean value of the residuals of the extraction of the channels of the square or alternatively the autonomous residual Δmax of FIG. 3 or equivalent 4, the following busbars are determined for the extraction of the TpC: • The (0.5 * TpFC ́) to be assigned busbar signal is determined by a correlation comparison of TpFL ́ and TpFR ́ with subsequent residual correction by the mean value of all correlation comparisons of the square in FIG. 5 or, alternatively, the autonomous residual of these correlation comparisons (Δmax). • The (0.5 * TpSiR ́) to be assigned busbar signal is determined by a correlation comparison of TpFR ́ and TpBR ́ with subsequent residual correction by the mean of all correlation comparisons of the square in FIG. 5 or, alternatively, the autonomous residual of these correlation comparisons (Δmax). • The busbar signal to be assigned to (0.5 * TpBC ́) is determined by a correlation comparison of TpBR ́ and TpBL ́ with subsequent residual correction by the mean value of all correlation comparisons of the square in FIG. 5 or, alternatively, the autonomous residual of these correlation comparisons (Δmax). • The (0.5 * TpSiL ́) to be assigned busbar signal is determined by a correlation comparison of TpBL ́ and TpFL ́ with a subsequent residual correction by the mean value of all correlation comparisons of the square in FIG. 5 or, alternatively, the autonomous residual of these correlation comparisons (Δmax). • The busbar signal to be assigned to 0.5 * TpC ́ is calculated by the sum of the correlation comparison of TpFL ́ and TpBR ́ or of TpFR ́ and TpBL ́, added with the doubled mean value of all correlation comparisons of the square in FIG. 5 or, alternatively, after subtracting the doubled autonomous residual of these correlation comparisons (Atop), see FIG. 7 is formed, whereby (with steps 11, 12 and 17, 18 and 19) 0.5 * TpC - r ́ (TpFL) - r ́ (TpFR) - r ́ (TpBL) - r ́ (TpBR) - r ́ (0.5 * TpFC) - r ́ (0.5 * TpSiR) - r ́ (0.5 * TpBC) - r ́ (0.5 * TpSiL) results.

While the TpC obtained in this way is retained as such, with minimized primary residuals there is now a sound field in the case of crosstalk and "phantom sound sources" and thus "masking" according to FIG. 8th .

With a spherical arrangement of the loudspeakers, the “phantom sound sources” of these minimized primary residuals appear as shown in FIG. 9 and 10, the factor can be calculated for NHK-22.2 signals as follows: If the top layer is based on the elevation ε, currently mostly 35 °, the level increase for the shortening of the distance for the «phantom sources» is also calculated 1 / (sin (π / 4) * cos ε)

[0080] and thus mathematically with 2.4414. Due to the increase in level with the square of the distance, this result must also be squared, which gives 5.9604. In addition, there is a psychoacoustic increase in the “phantom sound sources” by around + 3dB, i.e. a factor of 1.4142. Mathematically, this results in a total factor of 8.4292 or +18.5157 dB!

The fact that a practical correction of the TpC by + 1.6dB is sufficient to fully compensate for this factor, shows how efficient the TpC extraction shown is. This correction can be fallen below or below, and it is recommended overall for the calibration of the decoder an additional check of this parameter under listening conditions according to ITU-R Recommendation BS.1116-1.

If the above consideration is to be applied to the Auro formats developed by the Institute for Broadcasting Technology (IRT) according to FIG. 11 are applied, the table in FIG. 12. If the TpC with -6dB each is added to the speakers TpFL, TpFR, TpBR, TpBL during the downmix, it is extracted with a correlation comparison of the resulting channels TpFL ́ and TpFR ́ together with the TpFC with -6dB. Thus, the TpC can be calculated by simply adding the result of an optionally by autonomous residual correction, see FIG. 7, supported correlation comparison of the downmix signals TpFL ́ - TpBR ́ and TpFR ́ - TpBL ́.

If the result is now subtracted from the TpFC determined by correlation comparison and from the TpBL signal and TpBR signal, a masking sound image with few residuals also results.

However, here too there is a corresponding amplification of the in FIG. 13 illustrated «phantom sound sources» of the sound field formed in the case of crosstalk:

Their correction factor is calculated according to FIG. 14 and FIG. 15 with the elevation ε for the top layer of Auro 11.1 according to the formula 1 / (sin (π * η / 36) * cos ε)

Which for ε = 7 * η / 36 gives a value of 2.1281. This factor has to be squared, it results in 4.5288; there is also an increase in the “phantom sound sources” by + 3dB, ie a factor of 1.4142. This gives a factor of 6.4046; this corresponds to +16.1298 dB.

However, a separate correction factor of -3dB must be provided for TpFC, since this is intended to simulate a mono signal reproduced on TpFL and TpFR and would appear too loud in terms of residual + 3dB despite the increase in the mid-range sound sources!

DESCRIPTION OF THE DRAWINGS

Various embodiments of the present invention are described below by way of example, reference being made to the following drawings: FIG. 1 shows the principle of the illustrated correlation comparison, this in accordance with the rules for obtaining the real and imaginary parts of signals, see disclosure of the invention. A Fast Fourier Transform (FFT) is carried out before this correlation comparison, followed by an Inverse Fast Fourier Transform (IFFT). FIG. 2 geometrically illustrates the combinatorial application of such a correlation comparison to a corresponding downmix with three channels, which is illustrated by the circumscribed triangle. For a corresponding downmix with four channels, see FIG. 18th FIG. 3 illustrates the principle of the autonomous residual determination by Fourier transformation for four input signals. FIG. 4 illustrates the example of an alternative, equivalent embodiment, which illustrates the principle of the combinatorial possibilities of equivalent embodiments. FIG. 5 geometrically illustrates the combinatorial application of six correlation comparisons within the top layer of an NHK-22.2 signal for the algebraic determination of a TpC without secondary residual formation. FIG. 6 shows FIG. 5 associated table of the resulting signals or residuals. FIG. 7 shows the autonomous residual determination for the extraction of a TpC channel in the top layer of an NHK-22.2 system or an Auro-11.1 system. FIG. 8 shows the masking of an algebraic extraction of the TpC in the sound field formed in the case of crosstalk using “phantom sound sources” in the top layer of an NHK-22.2 arrangement. FIG. 9 shows the cross-section of an NHK-22.2 system, which is the calculated factor of an optimal masking according to FIG. 8 can be seen. FIG. 10 shows the floor plan of an NHK-22.2 system, which is the calculated factor of an optimal masking according to FIG. 8 can be seen. FIG. 11 FIG. 11 geometrically illustrates the combinatorial application of three correlation comparisons within the top layer of an Auro-11.1 signal for the algebraic determination of a TpC without secondary residual formation. FIG. 12 shows FIG. 11 associated table of the resulting signals or residuals. FIG. 13 shows the masking of an algebraic extraction of a TpC in the sound field formed in the case of crosstalk using “phantom sound sources” in the top layer of an Auro 11.1 arrangement. FIG. 14 shows the cross section of an Auro-11.1 system, which is based on the calculated factors of an optimal masking according to FIG. 13 can be removed. FIG. 15 shows the floor plan of an Auro-11.1 system, which is based on the calculated factors of an optimal masking according to FIG. 13 can be removed. FIG. 16 shows in a simplified representation the algebraic extraction of a TpC channel of the top layer of an NHK-22.2 signal, which subsequently leads to a masking of the residuals that occur. FIG. 17 shows the example of a non-linear inverse coding according to the unpublished application CH 02 300/12. FIG. Figure 18 geometrically illustrates the combinational application of a correlation comparison to a corresponding downmix with four channels, which is illustrated by the circumscribed square. FIG. 19 shows a decoder that uses the entire downmix and the transmitted mean value of all residuals calculated by the encoder modules by means of a correlation comparison, this in accordance with the rules for obtaining the real and imaginary parts of signals, see introduction to the disclosure of the invention, and by means of summation and difference formation the original input signals of the encoder are approximately calculated. Finally, the Inverse Fast Fourier Transforms (IFFT) are determined. The residual correction by the transmitted mean value of all residuals calculated by the encoder modules by means of correlation comparison can be replaced directly by their autonomous residual Δmax. In this case, Δ (k) is to be set equal to Δmax (k) in the frequency domain, while Δ (t) is equal to Δmax (t) in the time domain. Contrary to Δ (k), Δmax (k) can be determined directly in the decoder and therefore does not have to be transmitted from the encoder to the decoder, which is equivalent to a drastic bandwidth reduction. FIG. 20 shows the principle of the illustrated correlation comparison, this in accordance with the rules for obtaining the real and imaginary parts of signals, see introduction to the disclosure of the invention. A Fast Fourier Transform (FFT) is carried out before this correlation comparison, followed by an Inverse Fast Fourier Transform (IFFT). FIG. 21 shows an NHK-22.2 arrangement.

DETAILED DESCRIPTION

Application of the autonomous residual determination to an NHK-22.2 top layer signal (see FIGS. 5 to 10):

The first example of an application of the subject matter of the invention, here according to FIG. 5 to 10, to an NHK-22.2 top layer signal, see FIG. 21, provides the summation for the channels TpFL, TpFR, TpFC, TpC, TpBL, TpBR, TpSiL, TpSiR, TpBC (the «downmix»)

TpFL '= TpFL + 0.5 * TpFC + 0.5 * TpSiL + 0.25 * TpC

TpFR '= TpFR + 0.5 * TpFC + 0.5 * TpSiR + 0.25 * TpC

TpBL '= TpBL + 0.5 * TpBC + 0.5 * TpSiL + 0.25 * TpC

TpBR '= TpBR + 0.5 * TpBC + 0.5 * TpSiR + 0.25 * TpC

Where TpFL ', TpFR', TpBL ', TpBR' the corner points of the circumscribed square of FIG. 5 correspond. A correlation comparison can now not only be carried out for each side of the square in accordance with the rules for obtaining the real and imaginary parts of signals, see the introduction to the disclosure of the invention, but also for the two diagonals, i.e. TpFL ́ and TpBR ́ or TpFR ́ and TpBL ́.

By correlation comparison of TpFL 'and TpBR' according to the rules for obtaining the real and imaginary parts of signals, see introduction to the disclosure of the invention, a signal 0.25 k (TpC - Δ5) is now extracted (which is then given the factor 2 is multiplied). See also FIG. 5 and 6 and 16 as well as the following section on the application of the low-residual algebraic extraction of a TpC signal to an NHK-22.2 top-layer signal.

By correlation comparison of TpFR 'and TpBL' according to the rules for obtaining the real and imaginary parts of signals, see introduction to the disclosure of the invention, a signal 0.25 * (TpC - Δ6) is now extracted (which is then given a factor of 2 is multiplied). See also FIG. 5 and 6 and 16 as well as the following section on the application of the low-residual algebraic extraction of a TpC signal to an NHK-22.2 top-layer signal.

[0098] Now, according to FIG. 7 another frequency-dependent correlation comparison is carried out, i.e. that no inverse Fourier transformation takes place beforehand. The result is –Δ56max. Both 0.25 * (TpC - Δ5) and 0.25 * (TpC - Δ6) are multiplied by 4 and these are each added by 4 * Δ56max. See also the section below on applying low-residual algebraic extraction of a TpC signal to an NHK-22.2 top-layer signal.

By correlation comparison of TpFL 'and TpFR' according to the rules for obtaining the real and imaginary parts of signals, see introduction to the disclosure of the invention, a signal 0.5 * (TpFC + 0.5 * TpC - Δ1) is now extracted (the subsequently is multiplied by a factor of 2) and two signals (TpFL +0.5 * TpSiL + Δ1) and (TpFR + 0.5 * TpSiR + Δ1). See FIG. 5, 16, 18, 19 and 20.

By correlation comparison of TpFR ́ and TpBR ́ according to the rules for obtaining the real and imaginary parts of signals, see introduction to the disclosure of the invention, a signal 0.5 * (TpSiR + 0.5 * TpC - Δ2) is now extracted (the subsequently is multiplied by a factor of 2) and two signals (TpFR + 0.5 * TpFC + Δ2) and (TpBR + 0.5 * TpBC + Δ2). See FIG. 5, 16, 18, 19 and 20.

By correlation comparison of TpBR 'and TpBL' according to the rules for obtaining the real and imaginary parts of signals, see introduction to the disclosure of the invention, a signal 0.5 * (TpBC + 0.5 * TpC-Δ3) is now extracted (the subsequently is multiplied by a factor of 2) as well as two signals (TpBR + 0.5 * TpSiR + Δ3) and (TpBL + 0.5 * TpSiL + Δ3). See FIG. 5, 16, 18, 19 and 20.

By correlation comparison of TpFL 'and TpBL' according to the rules for obtaining the real and imaginary parts of signals, see introduction to the disclosure of the invention, a signal 0.5 * (TpSiL + 0.5 * TpC - Δ4) is now extracted (the subsequently multiplied by a factor of 2) as well as two signals (TpFL + 0.5 * TpFC + Δ4) and (TpBL + 0.5 * TpBC + Δ4). See FIG. 5, 16, 18, 19 and 20.

With the signals extracted in this way (TpFL + 0.5 TpSiL + Δ1), (TpFR + 0.5 * TpFC + Δ2), (TpBR + 0.5 * TpSiR + Δ3), (TpBL + 0.5 * TpBC + Δ4), if neither the residuals Δ1, Δ2, Δ3, Δ4 nor their mean value are known, now apply an autonomous residual determination:

Thus, from the extracted signals (TpFL + 0.5 * TpSiL + Δ1) and (TpFR + 0.5 * TpFC + Δ2) before an inverse Fourier transformation is carried out by correlation comparison according to the rules for obtaining the real and imaginary parts of signals , see introduction to the disclosure of the invention, now a signal Δ12max extracted.

Thus, from the extracted signals (TpFR + 0.5 * TpFC + Δ2) and (TpBR + 0.5 * TpSiR + Δ3) before an inverse Fourier transformation is carried out by correlation comparison according to the rules for obtaining the real and imaginary parts of signals , see introduction to the disclosure of the invention, now a signal Δ23max extracted.

Thus, before performing an inverse Fourier transformation, the extracted signals (TpBR +0.5 * TpSiR + Δ3) and (TpBL + 0.5 * TpBC + Δ4) are converted by correlation comparison according to the rules for obtaining the real and imaginary parts of signals , see introduction to the disclosure of the invention, now a signal Δ34max extracted.

The extracted signals are (TpBL +0.5 * TpBC + Δ4) and (TpFL + 0.5 * TpSiL + Δ1) before an inverse Fourier transformation is carried out by correlation comparison in accordance with the rules for obtaining the real and imaginary parts of signals , see introduction to the disclosure of the invention, now a signal Δ41max extracted.

A signal Δ123max is now extracted from the extracted signals Δ12max and Δ23max before an inverse Fourier transformation is carried out by correlation comparison according to the rules for obtaining the real and imaginary parts of signals, see introduction to the disclosure of the invention.

A signal Δ1234max is now extracted from the extracted signals Δ123max and Δ34max before an inverse Fourier transformation is carried out by correlation comparison according to the rules for obtaining the real and imaginary parts of signals, see introduction to the disclosure of the invention.

A signal Δmax is now extracted from the extracted signals Δ1234max and Δ41max before an inverse Fourier transformation is carried out by correlation comparison according to the rules for obtaining the real and imaginary parts of signals, see introduction to the disclosure of the invention.

[0111] With Δmax the following residual corrections, denoted in bold, result:

TpFL ≅ (TpFL + 0.5 * TpSiL + Δ1) - 0.5 * (TpSiL + 0.5 * TpC - 2 * Δ4) - 2 * Δmax≅ (TpFL + 0.5 * TpFC + Δ4) - 0.5 * (TpFC + 0.5 * TpC - 2 * Δ1) - 2 * Δmax

TpFR ≅ (TpFR + 0.5 * TpSiR + Δ1) - 0.5 * (TpSiR + 0.5 * TpC - 2 * Δ2) - 2 * Δmax≅ (TpFR + 0.5 * TpFC + Δ2) - 0.5 * (TpFC + 0.5 * TpC - 2 * Δ1) - 2 * Δmax

TpBR ≅ (TpBR + 0.5 * TpBC + Δ2) - 0.5 * (TpBC + 0.5 * TpC - 2 * Δ3) - 2 * Δmax≅ (TpBR + 0.5 * TpSiR + Δ3) - 0.5 * (TpSiR + 0.5 * TpC - 2 * Δ2) - 2 * Δmax

TpBL ≅ (TpBL + 0.5 * TpBC + Δ4) - 0.5 * (TpBC + 0.5 * TpC - 2 * Δ3) - 2 * Δmax≅ (TpBL + 0.5 * TpSiL + Δ3) - 0.5 * (TpSiL + 0.5 * TpC - 2 * Δ4) - 2 * Δmax

TpFC + 0.5 * TpC ≅ (TpFC + 0.5 * TpC - 2 * Δ1) + 2 * Δmax

TpSiR + 0.5 * TpC = (TpSiR + 0.5 * TpC - 2 * Δ2) + 2 * Δmax

TpBC + 0.5 * TpC = (TpBC + 0.5 * TpC - 2 * Δ3) + 2 * Δmax

TpSiL + 0.5 * TpC = (TpSiL + 0.5 * TpC - 2 * Δ4) + 2 * Δmax

Compared to signals even with correction by the mean value of the residuals Δ1, Δ2, Δ3, Δ4, this leads to significantly reduced artifacts or discoloration of the timbre and other unmasking effects without the mean value of the residuals Δ1, Δ2, Δ3, Δ4, for example from an encoder to a decoder. This results in a drastic reduction in bandwidth.

The effect of Δmax on the overall signal can be precisely controlled via the length of the discrete Fourier transform or inverse discrete Fourier transform used and represents a precise tool for controlling the final timbre.

An NHK-22.2 top layer signal can thus be used with regard to data to be stored or transmitted, for example between an encoder and a decoder, in the sense of FIG. 19 again very significantly with the clearest gain in quality.

Alternative autonomous residual determinations:

The devices for autonomous residual determination are shown in FIGS. 3, 4 and 7 by a dashed box. In these examples the autonomously determined residual is determined by the cascaded application of correlation comparisons of signal pairs. This determines the common or the correlated signal component of all input signals of the device for autonomous residual determination. Alternatively, however, to determine the autonomous residual it is also possible to use any other method for determining the common or correlated signal component of all input signals of the device mentioned. For example, the algorithm presented in FIG. 1 could be extended to three or more signals. For example, to determine the correlated component in n signals, the signs of the signal components per frequency of the n signals are compared. If all n signal components of this frequency have the same sign, the signal component of the signal with the minimum absolute signal component of this frequency is selected for the correlated signal component of this frequency. Otherwise this signal component is zero. This method is preferably used both for the real part and for the imaginary part. However, many other methods for determining the common or correlated signal component could also be used in order to determine the residual of the input signals autonomously.

One exemplary embodiment thus provides the cascaded determination of the correlated signal of the input signals. The cascaded determination includes at least the determination of a first correlated signal from a first signal pair of the input signals and at least one further determination of a second correlated signal from a signal pair having a signal based on the first correlated signal and a signal based on a third input signal.

A further embodiment relates to all other methods for determining the correlated signal of the input signals, i. those methods that do not have a cascaded determination of the correlated signal of the input signals.

Application of the low-residual algebraic extraction of a TpC signal to an NHK-22.2 top-layer signal (see FIGS. 5 to 10):

All operations lead to signals which both in the table to FIG. 6 contained basic signals as well as the mentioned residuals. These are to be targeted by the additional autonomous residual determination described above in combination with algebraic considerations according to FIG. 5 or FIG. 6 associated table can be eliminated.

Both 0.25 * (TpC - Δ5) and 0.25 * (TpC - Δ6) are multiplied by 2, analogously to the previous section, and added with 2 * Δ56max each. Both resulting signals are now summed.

This sum is now the sum of the signals obtained above

TpSiR + 0.5 * TpC ≅ (TpSiR + 0.5 * TpC - 2 * Δ2) + 2 * Δmax and TpSiL + 0.5 * TpC ≅ (TpSiL + 0.5 * TpC - 2 * Δ4) + 2 * Δmax or the sum of the signals obtained above TpFC + 0.5 * TpC ≅ (TpFC + 0.5 * TpC - 2 * Δ1) + 2 * Amax and TpBC + 0.5 * TpC ≅ (TpBC + 0.5 * TpC - 2 * Δ3) + 2 * Δmax subtracted. The signals result - TpSiL - r ́ (TpSiL) - TpSiR - r ́ (TpSiR) or. -TpFC - r ́ (TpFC) - TpBC - r ́ (TpBC), which, multiplied by the factor 0.5, are now summed up.

The signals formed in the previous section are now used TpFC + 0.5 * TpC ≅ (TpFC + 0.5 * TpC - 2 * Δ1) + 2 * Δmax TpSiR + 0.5 * TpC ≅ (TpSiR + 0.5 * TpC - 2 * Δ2) + 2 * Δmax TpBC + 0.5 * TpC ≅ (TpBC + 0.5 * TpC - 2 * Δ3) + 2 * Δmax TpSiL + 0.5 * TpC ≅ (TpSiL + 0.5 * TpC - 2 * Δ4) + 2 * Δmax multiplied by the factor 0.5 and also added.

The addition of the two last-mentioned sums results in the sought-after TpC signal with minimized primary residuals TpC - 2 * r ́ (TpFl) - 2 * r ́T (TpFR) - 2 * r ́ (TpBL) - 2 * r ́ (TpBR) - 2 * r ́ (0.5 * TpFC) - 2 * r ́ (0.5 * TpSiR) - 2 * r ́ (0.5 * TpBC - 2 * r ́ (0.5 * TpSiL), which, after multiplication by the factor 0.5, can be deducted from the above formulas in such a way that TpFC, TpSiR, TpBC and TpSiL are not only shown in excellent approximation, but also together with the TpC formed an ideal residual masking according to the considerations in FIG. 8 to 10, see above, results.

Application of the low-residual algebraic extraction of a TpC signal to an Auro-11.1 signal (see FIGS. 11 to 15):

The same considerations can also apply to the so-called Auro formats of the Institute for Broadcast Technology (IRT).

The following downmix results for the top-layer signal: TpFL ́ = TpFL + 0.5 * TpFC + 0.5 * TpC TpFR ́ = TpFR + 0.5 * TpFC + 0.5 * TpC TpBL ́ = TpBL + 0.5 * TpC TpBR ́ = TpBR +0.5 * TpC

In the decoder, by correlation comparison of TpFL 'and TpFR' according to the rules for obtaining the real and imaginary parts of signals, see introduction to the disclosure of the invention, now the sum of TpFC and TpC according to FIG. 12 associated table extracted with –6dB.

Likewise, in the decoder by correlation comparison of TpFL 'and TpBR' according to the rules for obtaining the real and imaginary parts of signals, see introduction to the disclosure of the invention, now the TpC signal according to FIG. 12 associated table with also -6dB extracted.

Likewise, in the decoder by correlation comparison of TpFR 'and TpBL' according to the rules for obtaining the real and imaginary parts of signals, see introduction to the disclosure of the invention, now the TpC signal according to FIG. 12 associated table with also -6dB extracted.

A signal -Δmax is now extracted from the extracted TpC signals before an inverse Fourier transformation is performed by correlation comparison according to the rules for obtaining the real and imaginary parts of signals, see introduction to the disclosure of the invention.

A residual correction is now carried out by adding 2 * Δmax with one of the two extracted doubled TpC signals.

In a further step, the TpC signal is calculated as a good approximation by simply adding the results and then multiplying by 0.5.

If half of the result is now subtracted from the TpFC determined by correlation comparison, excellent residual masking can be achieved with the following level correction:

The TpFC signal is lowered by -3dB. This results in an ideal masking; This is particularly the case if either an increased resolution of the discrete Fourier transform used to determine -Δmax is selected, or by its opposite, namely a discrete Fourier transform, whose harmonic frequencies are significantly below the high-frequency signal components due to their dimensions. In the second case, the procedure is similar to that for the mean value of all residuals. Thus, for example, the length of a Fast Fourier Transform (FFT) can be used to precisely control the signal behavior.

In the following, various exemplary embodiments for decoding and encoding the TpC signal are presented: 1. A method for decoding an audio multi-channel signal from an audio downmix signal, the audio downmix signal having a front right channel (TpFR), a front left channel (TpFL), a rear left channel (TpBL) and a rear right channel (TpBR), the method comprising the following steps; Performing a first correlation comparison between the front left channel (TpFL) and the rear right channel (TpBR); Performing a second correlation comparison between the front right channel (TpFR) and the rear left channel (TpBR); Determining an upper central channel (TpC) based on the first correlation comparison and the second correlation comparison. 2. The method of embodiment 1, wherein the upper central channel (TpC) is determined based on the sum of the output of the first correlation comparison and the second correlation comparison. 3. The method according to embodiment 1 or 2, wherein the upper central channel (TpC) is determined on the basis of the first correlation comparison and the second correlation comparison and on the basis of a residual signal of the upper central channel. 4. The method of embodiment 3, wherein the upper central channel (TpC) is determined on the basis of subtracting the residual signal of the upper central channel from the sum of the output of the first correlation comparison and the second correlation comparison. 5. The method according to embodiment 3 or 4, wherein the residual signal of the upper central channel is determined by correlation comparison of the output of the first correlation comparison and the second correlation comparison. 6. The method according to one of the exemplary embodiments 1 to 5, wherein a front central channel (TpFC) of the audio multi-channel signal is determined by correlation comparison of the front right channel (TpFR) and the front left channel (TpFL) of the audio downmix signal, the specific front central channel being determined by correcting the determined upper central canal. 7. The method according to one of the embodiments 1 to 6, wherein a rear central channel (TpBC) of the audio multi-channel signal is determined by correlation comparison of the rear right channel (TpBR) and the rear left channel (TpBL) of the audio downmix signal, the specific rear central channel being determined by correcting the determined upper central canal. 8. The method according to any one of the exemplary embodiments 1 to 7, wherein a right central channel (TpSiR) of the audio multi-channel signal is determined by correlation comparison of the front right channel (TpFR) and the rear right channel (TpBR) of the audio downmix signal, the specific right central channel being corrected by the specific upper central channel, and / or a left central channel (TpSiL) of the audio multi-channel signal is determined by correlation comparison of the front left channel (TpFL) and the rear left channel (TpBL) of the audio downmix signal, the left side central channel being corrected by the determined upper central channel. 9. The method according to one of the exemplary embodiments 6 to 8, wherein the front left channel of the audio multi-channel signal is determined on the basis of the front left channel of the audio downmix signal and an adjacent central channel derived from the audio downmix signal, and wherein the front right channel of the audio multi-channel signal is determined on the basis of the front right channel of the audio downmix signal and an adjacent central channel derived from the audio downmix signal wherein the rear left channel of the audio multi-channel signal is determined on the basis of the rear left channel of the audio downmix signal and an adjacent central channel derived from the audio downmix signal, and wherein the rear right channel of the audio multi-channel signal is determined on the basis of the rear right channel of the audio downmix signal and an adjacent central channel derived from the audio downmix signal. 10. The method according to embodiment 9, wherein the right central channel (TpSiR), the left central channel (TpSiL), the front central channel (TpFC) and the rear central channel (TpBC) of the audio multi-channel signal is corrected by a common residual signal of these channels. 11. The method according to embodiment 10, wherein the common residual signal is determined by cascaded pair-wise correlation comparisons of channels of the audio multi-channel signal that are not derived from the same correlation comparisons. 12. The method according to embodiment 10, wherein the common residual signal is obtained with the audio downmix signal. 13. The method according to one of the exemplary embodiments 6 to 8, the correction being carried out by the upper central channel by subtracting the upper central channel. 14. The method according to one of the exemplary embodiments 1 to 13, wherein a first correlation comparison of the front left channel (TpFL) of the audio downmix signal and the front right channel (TpFR) of the audio downmix signal is carried out, a second correlation comparison of the front right channel (TpFR) of the audio downmix signal and the rear right channel (TpBR) of the audio downmix signal is carried out, a third correlation comparison of the rear right channel (TpBR) of the audio downmix signal and the rear left channel (TpBL) of the audio downmix signal is carried out, a fourth correlation comparison of the front left channel (TpFL) of the audio downmix signal and the rear left channel (TpBL) of the audio downmix signal is carried out, wherein the results of the first and third correlation comparisons are summed and the results of the second and fourth correlation comparisons are summed. 15. The method according to one of the exemplary embodiments 1 to 14, wherein a correlation comparison outputs the correlated component of the compared signal pair. 16. The method according to one of the exemplary embodiments 1 to 15, wherein the at least three input signals are present in a frequency space representation and / or the common residual signal is output in a frequency space representation. 17. The method according to one of the exemplary embodiments 1 to 16, wherein a correlation comparison comprises the comparison of the signs of the frequency components of the two compared signals. 18. A computer program designed when executed on a processor to execute the method according to one of the exemplary embodiments 1 to 17. 19. Decoding device for decoding an audio multi-channel signal from an audio downmix signal, the audio downmix signal having a front right channel (TpFR), a front left channel (TpFL), a rear left channel (TpBL) and a rear right channel (TpBR), having; a first correlation comparison device for performing correlation comparison between the front left channel (TpFL) and the rear right channel (TpBR); a second correlation comparison device for performing a correlation comparison between the front right channel (TpFR) and the rear left channel (TpBL); Output device for determining an upper central channel (TpC) based on the first correlation comparison and the second correlation comparison.

In the following, various exemplary embodiments for determining an autonomous residual will be presented: 1. Method for determining a common residual signal from at least three input signals: Performing a first correlation comparison between a first input signal and a second input signal; Performing a second correlation comparison between a signal based on a signal output from the first correlation comparison and a signal based on the third input signal; Determining the common residual signal on the basis of the signal output from the second correlation comparison. 2. The method according to the previous embodiment, wherein the determination of the first and / or the second correlation comparison outputs the correlated component, the first individual component or the second individual component of the compared signal pair. 3. The method according to one of the previous exemplary embodiments, wherein the at least three input signals are present in a frequency space representation and / or the common residual signal is output in a frequency space representation. 4. The method according to one of the previous exemplary embodiments, wherein the determination of the first and / or the second correlation comparison includes the comparison of the signs of the frequency components of the two compared signals. 5. Method for decoding a multi-channel signal from a downmix signal: Performing a first correlation comparison between a first channel and a second channel of the downmix signal; Performing a second correlation comparison between a signal based on a signal output from the first correlation comparison and a signal based on a third channel of the downmix signal; Determining a correction signal based on the signal output from the second correlation comparison; Determining a first channel, a second channel and a third channel of the multi-channel signal on the basis of the first channel, the second channel and the third channel of the downmix signal; Determining a first additional channel of the multichannel signal on the basis of the first correlation comparison or on the basis of a further correlation comparison between a channel pair of the downmix signal; Correcting at least one of the first channel, the second channel, the third channel or the further channel of the multi-channel signal on the basis of the correction signal. 6. The method according to embodiment 5, having Performing a third correlation comparison between the second channel and the third channel of the downmix signal and determining a second additional channel of the multi-channel signal on the basis of the third correlation comparison; Performing a fourth correlation comparison between the first channel and the third channel of the downmix signal and determining a third additional channel of the multi-channel signal on the basis of the fourth correlation comparison; Correcting the second additional channel and the third additional channel of the multi-channel signal on the basis of the correction signal. 7. The method according to embodiment 5 or 6, wherein the second correlation comparison is carried out between the signal based on a signal output from the first correlation comparison and a signal based on a signal output from the third or fourth correlation comparison. 8. The method according to embodiment 5, having Performing a third correlation comparison between the second channel and the third channel of the downmix signal and determining a second additional channel of the multi-channel signal on the basis of the third correlation comparison; Performing a fourth correlation comparison between the third channel and a fourth channel of the downmix signal and determining a third additional channel of the multi-channel signal on the basis of the fourth correlation comparison; Performing a fifth correlation comparison between the first channel and the fourth channel of the downmix signal and determining a fourth additional channel of the multi-channel signal on the basis of the fifth correlation comparison; Performing a sixth correlation comparison between a signal based on a signal output from the second correlation comparison and a signal based on the third channel of the downmix signal; Performing a seventh correlation comparison between a signal based on a signal output from the sixth correlation comparison and a signal based on the fourth channel of the downmix signal; Determining a correction signal on the basis of the signal output from the seventh correlation comparison; Correcting the additional channels of the multi-channel signal on the basis of the correction signal. 9. The method according to embodiment 5, having Performing a third correlation comparison between the second channel and the third channel of the downmix signal and determining a second additional channel of the multi-channel signal on the basis of the third correlation comparison; Performing a fourth correlation comparison between the third channel and a fourth channel of the downmix signal and determining a third additional channel of the multi-channel signal on the basis of the fourth correlation comparison; Performing a fifth correlation comparison between the first channel and the fourth channel of the downmix signal and determining a fourth additional channel of the multi-channel signal on the basis of the fifth correlation comparison; Performing a sixth correlation comparison between a signal based on a signal output from the fourth correlation comparison and a signal output from the fifth correlation comparison; Performing a seventh correlation comparison between a signal based on a signal output from the second correlation comparison and a signal based on a signal output from the sixth correlation comparison; Determining a correction signal on the basis of the signal output from the seventh correlation comparison; Correcting the additional channels of the multi-channel signal on the basis of the correction signal. 10. The method according to embodiment 5, wherein two channels of the multi-channel signal from the first Correlation comparison are determined, the method further having Performing a third correlation comparison between the second channel and the third channel of the downmix signal and determining two channels of the multi-channel signal from the third correlation comparison; Performing a fourth correlation comparison between the third channel and a fourth channel of the downmix signal and determining two channels of the multi-channel signal from the fourth correlation comparison; Carrying out a fifth correlation comparison between the first channel and the fourth channel of the downmix signal and determining two channels of the multi-channel signal from the fifth correlation comparison; Performing a sixth, seventh, eighth and ninth correlation comparison between two channels of the multichannel signal that were not determined from the same correlation comparison; Perform cascaded correlation comparisons between pairs of signals based on the four signals resulting from the sixth, seventh, eighth, and ninth Correlation comparison output, and outputting a signal from the cascaded correlation comparison; Determining a correction signal based on the signal output from the cascaded correlation comparison; Correcting the additional channels of the multi-channel signal on the basis of the correction signal. 11. The method according to one of the exemplary embodiments 5 to 10, wherein a number of channels of the downmix signal are compared by cascaded correlation comparisons, the output signal of the cascaded correlation comparisons being used to determine the correction signal. 12. The method according to embodiment 5, having Carrying out a third correlation comparison between the third channel and a fourth channel of the downmix signal, wherein the first additional channel of the multi-channel signal is determined on the basis of the first correlation comparison and the third correlation comparison, and wherein the second correlation comparison is between a signal output from the first correlation comparison based signal and the signal based on the signal output from the third correlation comparison is performed; Correcting the first additional channel of the multi-channel signal on the basis of the correction signal. 13. The method according to one of the exemplary embodiments 5 to 12, wherein the one additional channel is carried out by subtracting the correction signal twice. 14. The method according to one of the exemplary embodiments 5 to 13, comprising transforming the channels of the downmix signal into the frequency space before the correlation comparisons are carried out. 15. The method according to embodiment 14, wherein the decoding method has a channel-dependently adjustable first profile for high-frequency channels and a second profile for low-frequency channels, with the first profile increasing the sampling rate in the frequency space by zeropadding compared to the second profile or with the second Profile the sampling rate for the transformation into the frequency domain is selected to be lower than the sampling rate of the corresponding channel of the downmix signal. 16. A method for determining a common residual signal of at least three input signals, characterized by the determination of the common residual signal by determining the common signal components of the at least three input signals. 17. The method according to embodiment 16, wherein the determination of the first and / or the second correlation comparison comprises the comparison of the signs of the frequency components of the two compared signals. 18. The method according to embodiment 16 or 17, the common signal components of the at least three input signals having the comparison of the signs of the frequency components of the input signals. 19. The method according to one of the exemplary embodiments 16 to 18, wherein a real part and / or an imaginary part of a frequency component of the signal of the common signal components of the at least three input signals is determined as the real part and / or the imaginary part of the corresponding frequency component of one of the at least three input signals, which is the minimum absolute value of the real part and / or the imaginary part if all signs of the real parts and / or the imaginary parts of the corresponding frequency component are the same as those of the at least three input signals. 20. The method according to any one of the preceding claims, wherein at least one of the input signals or a signal formed therefrom is corrected by the common residual. 21. Computer program designed to execute the method according to exemplary embodiments 1 to 15 when executed on a processor.

Closing remarks

Overall, the principles shown can be algorithmically expanded as required and are not limited to the efficient compression of multi-signals of any, even the highest, order, for the purpose of efficient storage or transmission, for example between an encoder and a decoder.

The present systems for determining at least one common signal and / or a first individual signal and / or a second individual signal from two input signals and in particular the autonomous residual determination can in principle be used in connection with applications of any nature with Fourier transforms or inverse Fourier transforms can be applied. These applications are basically understood to be included in the present invention.

Alternatively, any other method for determining a common signal, a first individual signal and a second individual signal from two input signals could be used.

Furthermore, the storage and / or transmission of data (eg a file or other storage means or transmission means) with a downmix signal / sum signal and / or with an averaged or autonomously determined residual or with a residual averaged from several residuals is also intended here be.

A multichannel signal with n channels can in turn contain a further multichannel signal with n − 1> 2 channels, a further multichannel signal with n − 2> 2 channels, etc.

Conversely, a further multi-channel signal of a higher order can in turn be derived from a multi-channel signal with n> 2 or n − 1> 2 or n − 2> 2 etc. channels.

If the invention protected by the claims has exemplary embodiments / areas of protection of the unpublished CH 00 743/14, it is hereby explicitly disclosed by reference that all the exemplary embodiments disclosed in CH 00 743/14 that fall under the scope of the claims, both are disclosed positively as an embodiment of the invention and negatively as a disclaimer of the invention. This means that the scope of protection covered by the patent claims can be divided into the exemplary embodiments disclosed in CH 00 743/14 and the exemplary embodiments disclosed by the granted scope minus the exemplary embodiments disclosed in CH 00 743/14 (individually, all together or in any combination) remaining exemplary embodiments / remaining scope of protection.

If the invention protected by the claims has exemplary embodiments / areas of protection of the unpublished CH 0 369/14, it is hereby explicitly disclosed by reference that all the exemplary embodiments disclosed in CH 0 369/14 that fall under the scope of protection of the claims, both are disclosed positively as an embodiment of the invention and negatively as a disclaimer of the invention. This means that the scope of protection covered by the patent claims can be divided into the exemplary embodiments disclosed in CH 0 369/14 and the exemplary embodiments disclosed by the granted scope of protection minus the exemplary embodiments disclosed in CH 0 369/14 (individually, all together or in any combination) remaining exemplary embodiments / remaining scope of protection.

Claims (15)

A method for decoding or upmixing an upmix signal having at least five channels from a downmix signal having at least three channels: Performing at least two correlation comparisons between at least two channel pairs of the at least three channels of the downmix signal, each one Correlation comparison of one of the at least two channel pairs results in at least one correlation comparison signal, wherein the at least one correlation comparison signal of one of the at least two channel pairs has a common signal of one of the channel pairs and / or a first individual signal of one of the channel pairs and / or a second individual signal of one of the channel pairs the common signal of one of the at least two channel pairs contains the common signal components of this channel pair and / or the first individual signal of one of the at least two channel pairs contains the signal components specific to the first channel of the channel pair and / or the second individual signal of one of the at least two channel pairs represents the second individual signal Channel of the channel pair contains specific signal components; Determining a common residual signal based on the common signal portions of the correlation comparison signals of the at least two channel pairs; Correcting at least one of the correlation comparison signals or a signal formed therefrom based on the common residual signal. 2. The method of claim 1, wherein for the determination of the common residual signal, the common signal components of an amount of the correlation comparison signals are used which has a correlation comparison signal of each correlation comparison of the at least two correlation comparisons. 3. The method according to claim 1 or 2, wherein the correlation comparison signals used for the determination of the common residual signal only first and / or second individual signals of the channel pairs of the same weight or only common signals of the channel pairs of the same weight or both contain at least one individual signal and at least one common signal, wherein in the case where the correlation comparison signals both contain at least one individual signal and at least one common signal, each of the at least one common signal relative to each at least one individual signal having a negative factor , preferably minus one half, is weighted. 4. The method of claim 1, wherein determining the common signal portions of the correlation comparison signals comprises comparing the signs of the frequency components of the correlation comparison signals. 5. The method according to claim 4, wherein a real part and / or an imaginary part of a frequency component of the signal of the common signal components of the at least three input signals is determined as the real part and / or the imaginary part of the corresponding frequency component of one of the at least three input signals which has the minimum absolute value of the Real part and / or the imaginary part is when all the signs of the real parts and / or the imaginary parts of the corresponding frequency component of the at least three input signals are equal. 6. The method according to any one of claims 1 to 5, wherein the determination of the common signal components of the correlation comparison signals has at least one correlation comparison or a series connection of at least two correlation comparisons. 7. The method according to any one of claims 1 to 6, wherein the common signals are corrected by the common residual signal. A method according to claim 7, wherein each common signal is corrected by the addition of the common residual signal or by the addition of the common residual signal multiplied by a factor, preferably two. 9. The method according to any one of claims 1 to 8, wherein at least one of the individual signals is corrected by the common signal of the correlation comparison, the channel pair can also be used to determine the corresponding individual signal. 10. The method of claim 9, wherein either the common signals used for the correction are already corrected by the common residual signal, or the individual signals are additionally corrected by the common residual. 11. The method according to any one of claims 1 to 10, wherein at least two channels of the upmix signal are determined on the basis of the common signals and the common residual signal, and / or at least three channels of the upmix signal are determined on the basis of the individual signals and the common residual signal. 12. The method according to any one of claims 1 to 11, wherein the upmix signal has eight channels and the downmix signal four channels wherein four correlation comparisons are made between the four respective adjacent channel pairs, which provide the eight correlation comparison signals having four common signals and four individual signals, the common residual signal being determined on the basis of the common signal components of four of the eight correlation comparison signals output from the four correlation comparisons, wherein the eight channels of the upmix signal are determined on the basis of the eight correlation comparison signals and the common residual signal. 13. The method of claim 12, wherein four center channels of the upmix signal are determined on the basis of the four common signals and the common residual signal, and / or four side channels of the upmix signal are determined on the basis of the four individual signals, the four center channels and the common residual signal. 14. Computer program formed when executed on a processor to perform the method according to claims 1 to 13 from. 15. Decoding device for decoding or upmixing an upmix signal from a downmix signal: a correlation comparison device for performing at least two correlation comparisons between at least two channel pairs of the at least three channels of the downmix signal, each one Correlation comparison of one of the at least two channel pairs results in at least one correlation comparison signal, wherein the at least one correlation comparison signal of one of the at least two channel pairs has a common signal of one of the channel pairs and / or a first individual signal of one of the channel pairs and / or a second individual signal of one of the channel pairs the common signal of one of the at least two channel pairs contains the common signal components of this channel pair and / or the first individual signal of one of the at least two channel pairs contains the signal components specific to the first channel of the channel pair and / or the second individual signal of one of the at least two channel pairs represents the second individual signal Channel of the channel pair contains specific signal components; A residual determination device for determining a common residual signal on the basis of the common signal components of the correlation comparison signals of the at least two channel pairs; and Correction signal means for correcting at least one of the correlation comparison signals or a signal formed therefrom on the basis of the common signal.