DE60308876T2 - AUDIO CHANNEL IMPLEMENTATION - Google Patents

Abstract

Using an M:N variable matrix, M audio input signals, each associated with a direction, are translated to N audio output signals, each associated with a direction, wherein N is larger than M, M is two or more and N is a positive integer equal to three or more. The variable matrix is controlled in response to measures of: (1) the relative levels of the input signals, and (2) the cross-correlation of the input signals so that a soundfield generated by the output signals has a compact sound image in the direction of the spatial center of gravity of the input signals when the input signals are highly correlated, the image spreading from compact to broad as the correlation decreases and progressively splitting into multiple compact sound images, each in a direction associated with an input signal, as the correlation continues to decrease to highly uncorrelated.

Description

TECHNICAL TERRITORY

The The invention relates to audio signal processing. In particular, it concerns the invention involves converting M audio input channels, which represent a sound field, in N audio output channels which represent the same sound field, each channel being a single sound field Audio signal stream is which one-way audio signals M and N are positive integers and M are at least 2 and M N is at least 3 and N is greater than Damn. Typically, a spatial Converter in which N is greater than M is, usually described as a "decoder".

STATE OF TECHNOLOGY

Even though the human has only two ears, hear we sound as a three-dimensional phenomenon, taking on a number Location tools such as head-related transfer functions (HRTFs - Head-Related Transfer Functions) and head movements are dependent. Full-fidelity sound reproduction therefore requires the storage and playback of the full 3D sound field or at least its perceptual features. Unfortunately, the sound recording technique not on detection of the 3D sound field, not even on detection a 2D sound plane and not even detecting a 1D soundline aligned. The current sound recording technology is based exclusively on Acquisition, storage and presentation of zero-dimensional, discrete audio channels aligned.

The most efforts to improve the sound fidelity since Edison's original Invention of sound recording focused on the shutdown the shortcomings his original analog cylinder / disc media with modulated grooves. To this shortcomings included limited, uneven frequency response, Noise, distortions, whine, whimper, wow and flutter, Wear, Pollution and copying losses. Although it is countless pieces Try isolated enhancements including electronic amplification, tape recording, Noise reduction and recording / playback equipment that cost more than Some cars did not have the conventional problems of single channel quality reasonably final solved, until the unique development of the general digital recording and especially the introduction the audio compact disc. Since then, they focused the essential research efforts for audio playback, once apart from some efforts to further Quality improvement digital recordings up to 24-bit / 96kHz sampling on the Reduce the amount required to maintain single channel quality Amount of data, mostly using perceptual coders, and on improving room-fidelity. The latter problem is that Topic of this document.

Efforts to Room fidelity improvement was achieved on two fronts: in attempts to to convey the perceptual features of a full sound field, and in attempts, an approximation of the actual original Sound field to convey. Examples of systems which The first approach is binaural recording and two-speaker virtual surround systems. Such Systems have a number of unfortunate shortcomings on, especially what the reliable Location of sounds in some directions and the requirement of using headphones or one firm, single listener position As.

To provide surround sound to multiple listeners, be it in a living room or in a commercial location such as a movie theater, the only viable alternative was to try to approximate the actual original sound field. Given the discrete channel nature of the sound record, it is not surprising that most efforts to date have been about something that could be described as a modest increase in the number of performance channels. Typical systems include the mono-tracking / three-speaker movie soundtracks of the early '50s, conventional stereo sound,' 60s-era quadraphonic systems, five-channel discrete magnetic soundtracks on 70mm films, Dolby Surround using a matrix in the '70s, AC-3 5.1-channel sound of the' 90s and more recently surround EX 6.1-channel sound. "Dolby", "Pro Logic" and "Surround EX" are trademarks of Dolby Laboratories Licensing Corporation. To a certain extent, these systems offer improved spatial reproduction over monophonic performance. However, mixing a larger number of channels entails more time and penalties for content producers, and the resulting perception is typically that of a few isolated, discrete channels rather than a continuous sound field. Aspects of "Dolby Pro Logic" decoding are described in U.S. Patent 4,799,260. Details of AC-3 are in "Digital Audio Compression Standard (AC-3) ", Advanced Television Systems Committee (ATSC), document A / 52, 20 December 1995 (downloadable from the World Wide Web on the Internet at www.atsc.org/Standards/A52/a_52.doc) also the error list of 22 July 1999 (downloadable from the World Wide Web on the Internet at www.dolby.comltech / ATSC_err.pdf).

As soon as the sound field is described, it is for a decoder in principle possible, the optimal signal feed for derive any output speakers. The one to such Decoder supplied channels are hereby transmitted differently as "main channels", "transmitted Channels "and" input channels ", and any output channel with a location which is not the location one of the input channels corresponds, is called "intermediate channel". One Output channel can also be one with the position of an input channel coincident situation.

In EP 1 054 575 A2 are variants of a wideband audio signal decoder including variants in which eight, six (when using two phantom channels) or five (down-mixing from eight to five) output channels in 360-degree space are derived from a pair of input signals. The manner in which the input signals are distributed to the output signals is determined by a plurality of "normalization" coefficients applied to multipliers in a matrix. The coefficients are derived individually in response to each of the input signals, their sum, their difference, their correlation and their relative phase. For uncorrelated input signals, the normalization coefficients are equal to each other, which contributes to producing a broad, diffused sound image.

EPIPHANY THE INVENTION

According to one The first aspect of the invention comprises a method for implementing M audio input signals, which are each associated with one direction, in N audio output signals, each associated with a direction, where N is greater than M is M is two or more and N is a positive integer is three or more, creating a variable M: N matrix, the Apply the M audio input signals to the variable matrix, the Derive the N audio output signals from the variable matrix and controlling the variable matrix in Reaction to the input signals such that a through the Output signals generated sound field in the direction of the nominal current basic direction of the input signals a compact sound has, if the input signals are strongly correlated, the sound image with decreasing correlation from the compact to a broad state diverges and even further, up to the strongly uncorrelated State, decreasing correlation gradually into several compact Sound images each in a direction associated with an input signal splits.

According to this In the first aspect of the invention, the variable matrix can be controlled in response to measurements: (1) the relative levels of the input signals and (2) the cross-correlation of the input signals. For one Measured value of the cross correlation of the input signals with values in a first range bounded by a maximum value and a Reference value, the sound field in this case, a compact sound have, if the measured value of the cross-correlation is the maximum value, and can it have a wide-ranging sound image, though the measured value of the cross correlation is the reference value, and for one Measured value of the cross correlation of the input signals with values in a second range bounded by the reference value and a Minimum value, the sound field can be the widely divergent sound image when the measured value of the cross-correlation is the reference value, and it can be a variety of compact sound images each in have a direction associated with an input signal when the measured value of the cross-correlation is the minimum value.

According to another aspect of the present invention, a method of converting M audio input signals, each associated with a direction, into N audio output signals each associated with one direction, where N is greater than M, and M is three or More, creating a plurality of variable m: n matrices, where m is a subset of M and n is a subset of N, applying a respective subset of the M audio input signals to each of the variable arrays, deriving a respective one Subset of the N audio output signals from each of the variable arrays, controlling each of the variable arrays in response to the subset of input signals applied thereto, such that a sound field generated by the respective subset of output signals derived therefrom is in the direction of the nominal current base direction the subset of input signals applied to them has a compact sound image when such input signals are highly correlated, with the sound image divergence from the compact to a wide state as the correlation decreases, and even further, to the highly uncorrelated state, decreasing correlation gradually into several compact sound images each in an input signal applied to them linked direction splitting and deriving the N audio output signals from the subsets of N audio output channels.

According to this Another aspect of the present invention may be the variable matrices also be controlled in response to information that the Effect of one or more other variable matrices that are the same Receive input signal, compensate. Furthermore, the derivation The N audio output signals from the subsets of N audio output channels also compensating for multiple variable matrices which are the same Generate output signal include. According to such further aspects According to the present invention, each of the variable arrays can be controlled in response to readings: (a) the relative level of they applied input signals and (b) the cross correlation of the Input signals.

According to one Another aspect of the present invention includes a method for converting M audio input signals, each with a Linked direction are, in N audio output signals, which each linked to one direction are, where N is greater than M is and M is three or more, creating a variable M: N matrix, which responds to scaling that controls matrix coefficients or the matrix outputs control applying the M audio input signals to the variable matrix, creating a variety of scaling generators [scaling factors] for variable m: n matrices, where m is a subset of M and n is a subset of N, applying a respective subset of the M audio input signals each of the generators for Scales for variable matrices, deriving a set of scalings for variable Matrices for respective subsets of the N audio output signals from each of the generators for Scales for variable matrices, controlling each one of the scaling generators for variable Matrices in response to the subset of input signals applied to it in such a way that, if the scales it generates on the variable M: N matrix are applied, one by the respective subset Sound field generated by generated output signals in the nominal current basic direction of the subset of input signals, which the applied scalings produced a compact sound has, if such input signals are highly correlated, the Sound image with decreasing correlation from compact to wide State dissipates and even further, to the point of strongly uncorrelated State, decreasing correlation gradually into several compact Sound pictures in each case with an input signal, which the splits applied scalings generated, linked direction, and deriving the N audio output signals from the variable matrix.

According to this In another aspect of the present invention, the scaling generators may be variable Matrices are also controlled in response to information the effect of one or more other generators for scaling for variable Matrices that receive the same input signal compensate. Further, deriving the N audio output signals from the variable matrix Compensating multiple scaling generators for variable Matrices, which scalings for generate the same output signal include. According to such In still further aspects of the present invention, each of the Generators for Scales for variable matrices in response to readings: (a) the relative level of the input signals applied thereto; and (b) cross-correlation the input signals are controlled.

According to the present Invention, M audio input channels that represent a sound field, in N audio output channels, which represent the same sound field, implemented, with each channel a single audio signal stream is from one direction represents incoming audio signals, M and N are positive integers and M is at least 2 and N is at least 3 and N is greater than Damn. Each input and output channel has a direction associated with it (e.g., azimuth, altitude and optionally removal to a closer or farther away virtual or projected channel). One or more Sets of output channels are generated, each set having one or more output channels. Every sentence is ordinary with two or more spatially adjacent input channels connected, and each output channel in a sentence is determined by determining a Measured value of the cross-correlation of the two or more input channels and a Measured value of the correlations of the levels of the two or more input channels. The measured value of the cross-correlation is preferably a measured value the cross-correlation with zero-time shift, which is the relationship the common energy level with respect to the geometric mean is the input signal power level. The common energy level is preferably the smoothed or Averaged common energy levels, and the input signal energy levels are the smoothed ones or Average input signal energy level.

In one aspect of the present invention, multiple sets of output channels may be associated with more than two input channels, and a method may determine the correlation of the input channels to which each set of output channels is associated according to a hierarchical order that each set or sets is classified according to the number of input channels to which its output channel or channels are linked, the largest number of input channels corresponding to the highest classification, and the processing sets in an order according to their processed hierarchical order. In accordance with one aspect of the present invention, processing also takes into account the results of processing higher order sentences.

The Playback or decoding aspects of the present invention assume that each of the M audio input channels, which represent audio signals coming from one direction, by a Passive-"nearest neighbor" -Amplitudenschwenk coding each source direction has been generated (i.e., it is assumed that a source direction especially on the nearest entrance channel or the nearest input channels mapped) without additional Side chain information needed were (the use of side chain or auxiliary information is optional), which makes it compatible with existing mixing processes, consoles and formats. Although such source signals through express use A passive encoding matrix can generate most conventional recording method such source signals by themselves (thus forming an "effective coding matrix"). The playback or decoding aspects of the present invention are also disclosed in high degree compatible with natural Recording source signals, as she did with five real directional microphones could be made there, taking into account a certain possible Time Delay, sounds coming from intermediate directions tend to be mainly on the nearest one Microphones (in a horizontal array especially on the nearest Pair of microphones).

One A decoder or decoding method according to aspects of the present invention The invention may be described as a grid of coupled processing modules or modular functions (hereinafter "modules" or "decoding modules"), each of which is implemented or each for generating one or more output channels (or alternatively, control signals used to generate one or more Outputs can serve), typically from the two or more closest spatially adjacent, associated with the decoding module Input channels, is used. The output channels typically provide relative proportions of the audio signals in the nearest spatial adjacent, associated with the particular decoding module input channels. As more fully explained below explains The decoding modules are in the sense that modules share inputs use, loosely coupled and there is a hierarchy of Decoding modules. Modules are according to the number of input channels, with which it links are classified in the hierarchy (the module or modules, which or which with the highest Number of input channels connected is / are, is / are ranked highest). A monitor or a monitoring function commands over the modules, allowing common input signals between or below Modules are fairly distributed and higher-order decoder modules can affect the output of lower order modules.

each Decoder module can actually contain a matrix so that it directly generates output signals, or each decoder module can generate control signals that come together with the control signals generated by other decoder modules used to calculate the coefficients of a variable matrix or the scalings of inputs or outputs a fixed matrix to vary all output signals produce.

Decoder modules mimic the functioning of the human ear to try to provide a transparent representation in the perception. signal conversion according to the present Invention of which decoder modules and module functions Aspect can be either on broadband signals or on any frequency band a multiband processor and can, depending on the implementation, once per sample or once per block of samples. A multiband embodiment can either be a filter bank like a discrete filter bank for critical bands or a filter bank with a band structure, which with a with you linked Decoder is compatible, or a transformation configuration like a linear filter bank for Fast Fourier Transform (FFT) or Fast Fourier Transform Modified Discrete Cosine Transform (MDCT - Modified Discrete Cosine Transform).

Another aspect of this invention is that the number of speakers receiving the N output channels is reduced to a practicable number by judicious use of virtual imaging consisting of generating perceived sound images at different positions in the room than where a speaker is located can be reduced. Although the most common use of virtual imaging in the stereo rendering of an image part is on a position between two speakers, with a monophonic signal being panned between the channels, virtual imaging can be considered As an aspect of the present invention, they include processing projected phantom images that produce the auditory impression of being beyond the walls of a room or within the walls of a room. Virtual imaging is not considered a viable method of group performance with a sparse number of channels because it requires the listener to be the same or almost the same distance from the two speakers. For example, in cinemas, the left front and right front speakers are too far apart to provide a viable phantom image of a central image for a large portion of the audience, and therefore instead, given the importance of the central channel as the source of the largest Part of the dialogue, a physical central speaker used.

With As the speakers become denser, a point is reached the virtual mapping between any pair of speakers for one huge Part of the audience is feasible, at least as long as the pans done gently; with enough Speakers are the gaps no longer perceived as such between the speakers.

signal distribution

As mentioned above, A measure of cross-correlation determines the ratio of Ground energy [dominant energy] (common signal components) to non-ground energy [non-dominant energy] (non-common energy) Signal components) in a module and the degree of propagation of the Non-fundamental signal components on the output channels of the module. This may be easier to understand when looking at the signal distribution on the output channels a module under different signal conditions in the case of a Two-input module considered. Unless otherwise stated, The principles outlined extend directly to higher-order modules.

The Problem with the signal distribution is that it is often too few Information gives to the original Distribution of signal amplitudes, much less the signals themselves, regain. The basic information available is the Signal level at each module input and the average cross product the input signals, the common energy level. The cross correlation zero-time shift is the ratio of the common energy level in terms of the geometric mean of the input signal energy levels.

The Meaning of the cross-correlation is that it as a metric the net amplitude of all inputs common signal components serves. If there is a single, somewhere between the inputs of the Module has panned signal (an "inner" or "intermediate" signal) all inputs the same waveform, albeit with possibly different ones Amplitudes, and under these conditions the correlation is 1.0. in the other extreme case, when all input signals are independent, which means that there is no common signal component is the correlation is zero. For correlation values between 0 and 1.0 is assumed to have some intermediate balance levels single common signal component and independent signal components the entrances correspond. Consequently, any input signal condition into a common signal, the "fundamental signal", and after subtracting common signal contributions left over Input signal components containing a "total residual" signal component (the non-fundamental signal component); or residual signal energy). As above mentioned, is the amplitude of a common signal or "fundamental" is not necessarily louder than the residual signal or the non-fundamental signal level.

Consider for example, the case of a bow of five channels (L (left), MidL (center-left), C (center), MidR (center-right), R (right)), which refers to a single Lt / Rt pair (Left total / Right total - Left total / Right total) in which the original five channels are to be recovered. If all five channels amplitude equal independent Have signals, Lt and Rt are equal in amplitude at an intermediate value the common energy, which is an intermediate value of the cross-correlation between zero and one (because Lt and Rt are not independent signals are). The same levels can with suitably chosen Levels of L, C and R, without signals from MidL and MidR, can be achieved. Thus one could Two-input / five-output module only that corresponding to the fundamental direction (dominant direction) (in this case C) Output channel and the input signal residues (L, R) after removal of C energy the entrances Lt and Rt feed corresponding output channels, with no signals to the output channels MidL and MidR are given. Such a result is not desirable - unnecessary turn off a channel is almost always a bad choice because of small interference in cause the signal conditions the "switched off" channel to toggle between on and off, causing an annoying rattling noise ("Rattle" corresponds to one fast on / off becoming channel), especially if you hear the "disconnected" channel isolated.

If there are several possible Output distributions for exists a given set of module input signal values hence the reserved Approach, seen from the standpoint of single channel quality, therein, the non-fundamental signal components in accordance with the signal conditions as evenly as possible over the output channels of the module spread. One aspect of the present invention is uniform spreading the available Signal energy, subject to the signal conditions, according to a three-way splitting instead according to a "reason" / "total remainder" two way splitting. Preferably, the three-way splitting comprises basic (common) Signal components, filling (evenly spread) Signal components and an input signal component remainder. Unfortunately there are only enough Information to a two-way splitting (basic signal components and all other signal components). A suitable approach to realize a three-way splitting, where the two-way splitting at correlation values above one certain value, the fundamental signal and propagated non-fundamental signal components used is described herein; at correlation values below this value uses the two-way split the propagated non-fundamental signal components and the rest. The common signal energy is split into "ground" and "evenly spread" energy. The "evenly spread" component contains both "common" and "residual" signal components. Therefore, "spread" means a mixture of common (correlated) and residual (uncorrelated) Signal components.

In front the processing is for a given input / output channel configuration of a given Module calculates a correlation value, which all output channels, which receive the same signal amplitude, corresponds. This correlation value may be referred to as the random_xcor value. For one single center-derived intermediate output channel and two input channels calculate the value of Random_xcor 0.333. For three in equal distances arranged intermediate channels and two input channels For example, calculating the value Random_xcor can yield 0.483. Even though it was found that such time values give satisfactory results deliver, they are not critical. For example, values of about 0.3 or 0.5 usable. In other words, for a module with M inputs and N outputs is there a certain degree of correlation of the M inputs, which considered as representing equal energies in all N outputs can be. This can be achieved by looking at the M entrances, as if they were using a passive N-to-M matrix, which N independent signals receive the same energy, would have been derived although the actual inputs Naturally can be derived in another way. This threshold correlation value is "random_xcor" and he can do one Represent dividing line between two work areas.

During processing, if the cross-correlation value of a module is greater than or equal to the value Random_xcor, then it is scaled to a range of 1.0 to 0: Scaled_xcor = (Correlation - Random_xcor) / (1 - Random_xcor)

Of the Scaled_xcor value represents the amount of the basic signal over the evenly spread level What could possibly be left is, can be evenly on the other output channels of the module.

however is there an extra Factor that takes into account should be, namely that at gradual Increase in eccentricity the nominal current direction of the input signals either the Amount of energy spread should be gradually reduced, if equal distribution is maintained on all output channels, or alternatively, the amount of propagated energy can be maintained should, but those on output channels distributed energy according to the eccentricity of the ground energy should be reduced - in in other words, a reduction of the energy along the output channels. In the latter Case can be extra processing complexity be required to maintain the "output power equal to input power" state.

If on the other hand, the current correlation value is smaller than the value "random_xcor" becomes the basic energy regarded as zero, the evenly spread energy diminishes and after decreases and gets left over to the rest signal, whatever is allowed at the entrances to accumulate. When correlation = zero, there is no internal signal only independent Input signals which are mapped directly to output channels.

• a) Wenn die tatsächliche Korrelation größer als "Zufällige_xcor" ist, gibt es genügend gemeinsame Energie, um in Betracht zu ziehen, dass es ein Grundsignal gibt, das zwischen zwei benachbarte Ausgänge zu lenken (schwenken) ist (oder natürlich in einen bestimmten Ausgang zu speisen ist, wenn seine Richtung zufällig mit diesem Ausgang zusammenfällt); die diesem zugeteilte Energie wird von den Eingängen subtrahiert, um Reste zu erhalten, welche (vorzugsweise gleichmäßig) auf alle Ausgänge verteilt werden.
• b) Wenn die tatsächliche Korrelation genau gleich "Zufälliger_xcor" ist, wird die Eingangsenergie (welche als Gesamt-Rest angesehen werden könnte) gleichmäßig auf alle Ausgänge verteilt (dies ist die Definition von "Zufälliger_xcor").
• c) Wenn die tatsächliche Korrelation kleiner als "Zufällige_xcor" ist, gibt es nicht genügend gemeinsame Energie für ein Grundsignal, so dass die Energie der Eingänge proportional dazu, wieviel kleiner, auf die Ausgänge verteilt wird. Dies ist, als wenn man den korrelierten Teil als den gleichmäßig auf alle Ausgänge zu verteilenden Rest und den unkorrelierten Teil eher wie eine Anzahl von Grundsignalen, welche entsprechend den Richtungen der Eingänge an Ausgänge zu senden sind, behandelte. Wenn die Korrelation im Extremfall gleich Null ist, wird jeder Eingang nur in eine Ausgangsposition gespeist (gewöhnlich in einen der Ausgänge, aber es könnte eine Schwenkposition zwischen zweien davon sein).

The operation of this aspect of the invention can be further explained as follows:

• a) If the actual correlation is greater than "random_xcor", there is enough common energy to consider that there is a fundamental signal to pass between two adjacent outputs ken (panning) is (or of course to feed into a particular output if its direction happens to coincide with this output); the energy allocated to it is subtracted from the inputs to obtain residues which are distributed (preferably equally) to all outputs.
• b) If the actual correlation is exactly "random_xcor", the input energy (which could be considered the total remainder) will be evenly distributed across all outputs (this is the definition of "random_xcor").
• c) If the actual correlation is less than "Random_xcor", there is not enough common energy for a fundamental signal so that the energy of the inputs is distributed proportionally to how much smaller to the outputs. This is like treating the correlated part as the remainder to be distributed equally to all outputs and the uncorrelated part more like a number of primitive signals to be sent to outputs according to the directions of the inputs. If the correlation is zero in the extreme case, each input will only be fed to one home position (usually one of the outputs, but it could be a pan position between two of them).

therefore there is a continuum of full correlation in which a single one Signal according to the relative energies of the inputs between two outputs is panned, via "Random_xcor", in which the inputs evenly all outputs to zero correlation, where M inputs are independent in M home positions are fed.

compensation of interactions

As mentioned above, can be assumed that channel implementation according to a Aspect of the present invention requires a grid of "modules". Because several modules can share a given input channel are Interactions between modules possible which the performance can worsen if not any compensation is applied. Although it is in usually not possible is signals to an input depending on which Module they "belong", can separate the treasure the amount of input signal used by each connected module resulting correlation and direction estimates improve what a improved overall performance.

As mentioned above, There are two types of module interactions: those which Modules at a common or lower hierarchical level (i.e. Modules with an equal number of inputs or fewer inputs), called "neighbors", and those who have modules at a higher hierarchical level (with more inputs) as a given module, but which have one or more common inputs share, as "neighbors higher Order " affect.

First, consider neighbor compensation on a common hierarchical level. To understand the problems caused by neighbor interactions, consider an isolated two-input module with identical L / R (left and right) input signals, A. This corresponds to a single fundamental (common) signal midway between them inputs. The common energy is A ² , and the correlation is 1.0. Suppose a second two-input module with a common signal, B, at its L / R inputs, a common energy B ^2, and also a correlation of 1.0. When the two modules are connected at a common input, the signal at that input is A + B. Assuming that the signals A and B are independent, then the average product of AB equals zero, so that the common energy of the first module A (A + B) = A ² + AB = A ² and the common energy of the second modulus B is (A + B) = B ² + AB = B ² . So the shared energy is not affected by neighboring modules as long as they process independent signals. This is generally a valid assumption. If the signals are not independent, if they are the same, or if, at least, they share substantially common signal components, the system responds to a manner consistent with human ear response - namely, the common input becomes larger, thereby increasing the resulting audio image pulls in the common entrance. In this case, the L / R input amplitude ratios of each module are shifted because the common input has more signal amplitude (A + B) than each of the outer inputs, biasing the direction estimate toward the common input. In this case, the correlation value of both modules is now slightly smaller than 1.0 because the waveforms on both input pairs are different. Because the correlation value determines the degree of propagation of the non-common signal components and the ratio of fundamental energy (common signal component) to non-fundamental energy (non-common signal component), an uncompensated common input signal causes propagation of the non-common signal distribution of each module.

To compensate, a reading of the "attributable to each input of each module" is added common input level energy, and then each module is informed of the total amount of such common input level energy of all adjacent levels of the same hierarchical level at each module input Two ways of calculating the measurement value of the common input level attributable to each input of a module are described herein: which is based on the common energy of the module's inputs (generally described in the next paragraph), and another which is more accurate but requires larger computational resources based on the total energy of the module's internal outputs (discussed below in connection with the arrangement in FIG 6A described).

According to the first Type of calculation of the measured value of each input of a module attributable common input level allows the analysis the input signals of a module no direct solution for the common input level at each entrance, only a share of the total joint energy, which is the geometric mean of the common input energy levels is. Because the common input power level at each input is the Total energy level at this input, which is measured and known is, do not exceed can, becomes the common total energy, subject to the condition below, in esteemed common input levels which are proportional to the observed input levels are, disassembled. After the whole of the common input level for all Module is calculated in the grid (regardless of whether the reading the common input level on the first or the second type of calculation is based), each module is over the total amount of common input levels of all neighboring Module at each input informs, a quantity which indicates as the "neighbor level" of a module each of his inputs referred to as. The module then subtracts at each of its inputs Neighbor level from the input level to derive compensated input levels which are used to calculate the correlation and direction (nominal current basic direction of the input signals).

In the above example, the neighbor levels are initially zero, and because the common input has more signal than each of the end inputs, the first module occupies a common input power level at this input that is above A ² , and the second module occupies a common input level at the same Entrance that is above B ² . Since the total claims are more than the available energy level thereon, the claims are limited to approximately A ² and B ² , respectively. Because there are no other modules connected to the common input, each common input level corresponds to the neighboring level of the other module. Consequently, the compensated input power level seen by the first module is (A 2 + B 2 ) - B 2 = A 2 and is the compensated input power level seen by the second module (A 2 + B 2 ) - A 2 = B 2 ,

however these are just the levels observed with isolated modules would have been. Consequently, the resulting correlation values are 1.0 and the basic directions as desired centered at the correct amplitudes. Nevertheless, the recovered Signals themselves not completely isolated - the Output of the first module has a certain B signal component and vice versa, but this is a limitation of a matrix system, and if the processing is on a multi-band basis, the mixed signal components are at a similar level Frequency, which makes the distinction between them a little questionable. In more complex situations, compensation is usually not so accurate, but the experience with the system says that in practice the compensation most of the effects of neighbor module interactions weakens.

With the introduction of the principles and signals used in neighbor level compensation, the expansion to the compensation of higher order neighbor levels will be quite simple. This applies to situations where two or more modules at different hierarchical levels share more than one input channel. For example, there could be a three-input module sharing two inputs with a two-input module. A signal component common to all three inputs is also common to the two inputs of the two-input module and is processed without compensation by each module at different positions. More generally, there may be a signal component common to all three inputs and a second component common only to the inputs of the two-input module, which requires that their effects for correct conditioning of the output sound field be separated as much as possible. Consequently, the effects of the three inputs common signal, as contained in the common input levels described above, should be subtracted from the inputs before the two-input calculation can be performed correctly. Indeed, the higher order common signal elements should be subtracted not only from the lower level module's input levels but also from its observed common energy level reading before calculating lower levels. This is unlike the effects of common input levels of modules on the same hierarchical level that do not affect the common energy level reading of a neighboring module. Thus, the higher level neighbor levels should be considered and used separately from the neighbor levels of the same order. At the same time, as higher order neighbor levels are transmitted down to lower modules in the hierarchy, remaining common levels of lower level modules should also be propagated up the hierarchy because, as noted above, lower level modules behave like higher level modules than normal neighbors. Some sizes are interdependent and difficult to solve at the same time. To avoid the implementation of resource-intensive complex simultaneous solution calculations, previous calculated values can be transmitted to the appropriate modules. A possible interdependence of modules common input levels at different hierarchical levels can be solved either by using the previous value as above or by performing calculations in a repeating sequence (ie a loop) from the highest to the lowest hierarchical level. Alternatively, a solution of simultaneous equations may be possible, although it may entail a significant computational effort.

Even though the described methods for compensation of interactions only approximate correct values for Providing complex signal distributions, one considers them capable of one Improvement over a raster arrangement that does not take into account module interactions, to accomplish.

SUMMARY THE DRAWINGS

1 Fig. 12 is a plan view schematically illustrating an idealized decoding arrangement such as a 16-channel horizontal array along the walls of a room, a 6-channel array arranged in a circle above the horizontal array, and a single one Additional channel used, shows.

2 FIG. 12 is a block diagram showing an overview of a multi-band transform embodiment of a plurality of modules operating with a central monitor and the example of FIG 1 implements, offers.

3 Figure 4 is a block diagram illustrating the manner in which a supervisor such as a supervisor 201 in 2 facilitate end point scaling.

4A - 4C show a block diagram of a module according to one aspect of the present invention.

5 Fig. 12 is a schematic view showing a hypothetical arrangement of a three-input module fed by a triangle of input channels, three inner output channels, and one basic direction. The view facilitates understanding the distribution of fundamental signal components.

6A and 6B are block diagrams each adapted to (1) generate the estimated total energy for each input of a module in response to the total energy at each input; and (2) in response to a measure of cross-correlation of the input signals, generate an excess end point energy scaling component for each of the endpoints of the module.

7 Figure 12 is a block diagram illustrating a preferred function of the "sum and / or larger of" block 367 in 4C shows.

8th Figure 11 is an idealized illustration of the manner in which one aspect of the present invention generates scaling components in response to a measure of cross-correlation.

9A and 9B to 16A and 16B are series of idealized representations illustrating the output scaling of a module resulting from various examples of input signal conditions.

EMBODIMENTS THE INVENTION

Around Aspects of the present invention to examine was an arrangement with a horizontal array of 5 speakers on each wall of a room with four walls (one Speakers in each corner and three speakers at even intervals between all corners), 16 loudspeakers in total, including common corner loudspeakers, and a ring of 6 speakers above a centrally placed Listener in a vertical angle of about 45 degrees and a single speaker directly above him, a total of 23 speakers, and a subwoofer / LFE (Low Frequency Effects - Low Frequency Effects) Channel, a total of 24 speakers, all from one for 24-channel playback furnished personal computers are built. Even though this system is referred to as a 23.1-channel system according to current usage could be For simplicity, it will be referred to herein as a 24-channel system.

1 Fig. 12 is a plan view schematically showing an idealized decoding device in the manner of the test device just described. Five far-reaching horizontal input channels are called squares 1' . 3 ' . 5 ' . 9 ' and 13 ' shown on the outer circle. A vertical channel, which can be derived from the five far - reaching inputs by correlation or generated reverberation or fed separately (as in 2 ), is as the dashed square 23 ' shown in the center. The 23 far-reaching output channels are as numbered filled circles 1 - 23 shown. The outer circle of 16 output channels lies on a horizontal plane, the inner circle of 6 output channels is 45 degrees above the horizontal plane. output channel 23 is directly above one or more listeners. Five two-input decoding modules are by parentheses 24 - 28 around the outer circle connecting each pair of horizontal input channels. Five additional vertical two-input decoding modules are brackets 29-33 representing the vertical channel to each of the horizontal inputs. output channel 21 , the raised channel in the back center, is made up of a three-input decoding module 34 which is indicated by arrows between exit channel 21 and input channels 9 . 13 and 23 is shown derived. Thus lies three-input module 34 a hierarchical level higher than its lower two-input neighbor modules in the hierarchy 27 . 32 and 33 , In this example, each module is linked to a respective pair or trio of nearest spatially adjacent input channels. Each module in this example has at least three neighbors at the same hierarchical level. For example, the modules 25 . 28 and 29 Neighbors of the module 24 ,

Although the in 1 If the decoder modules shown have different three, four or five output channels, a decoder module may have any reasonable number of output channels. An output channel can be between two or more input channels or at the same position as an input channel. Thus, in the example in 1 each of the input channel positions also has an output channel. Two or three decoding modules share each input channel.

Although the arrangement in 1 five modules ( 24 - 28 ) (each with two inputs) and five inputs ( 1' . 3 ' . 5 ' . 9 ' and 13 ' ) used to 16 horizontal outputs ( 1 - 16 ), which represent positions along the four walls of a room, can achieve similar results with a minimum of three inputs and three modules (each with two inputs, each module sharing an input with another module).

By using multiple modules, each module has output channels in a curve or line (as in the example in 1 and 2 ), uncertainties in decoding, such as occur in prior art decoders in which correlations less than zero are indicated as indicating backward directions, can be avoided.

Even though Input and output channels by their physical position or at least by their direction can be described does it make sense to describe them with a matrix because of these provides a well-defined signal relationship. Each matrix element (Line i, column j) is a transfer function which Input channel i sets in relation to output channel j. matrix elements are usually signed multiplicative coefficients, can but also phase or delay terms contain (in principle, each filter) and can functions of the frequency (with respect to discrete frequencies at each frequency other matrix). This is easy in the case of on the outputs one fixed dynamic matrix scaling applied but also for Variable matrixing either by means of a separate scaling for each Matrix element or, in the case of more complicated matrix elements, as simple scalar scaling, where matrix elements themselves are variable, e.g. a variable delay.

When mapping physical positions to matrix elements, there is some flexibility; basically For example, embodiments of aspects of the present invention may handle mapping an input channel to any number of output channels, and vice versa, but the most common situation is to accept signals that result in simple sum factors of 1.0 in the sum of squares to save power , are mapped to the nearest output channels only. Such mappings often take place via a sine / cosine panning function.

For example, with two input channels and three inner output channels lying in-line between them, and the two end-point output channels coincident with the input positions (ie, an M: N modulus where M equals 2 and N equals 5) one can assume that the span represents 90 degrees of the arc (the area in which sine and cosine change from 0 to 1 or vice versa) so that the channels are 90 degrees / 4 intervals = 22.5 degrees apart so that the channels have the following matrix coefficients (cos (angle), sin (angle)): Lout coefficients = cos (0), sin (0) = (1, 0) MidLout coefficients = cos (22.5), sin (22.5) = (0.92, 0.38) Cout coefficients = cos (45), sin (45) = (0.71, 0.71) MidRout coefficients = cos (67.5), sin (67.5) = (0.38, 0.92) Rout coefficients = cos (90), sin (90) = (0, 1)

Thus, for the case of a fixed coefficient matrix and scaled variable gain at each matrix output, the signal output is in each of the five output channels (where "SF" is a scaling for a particular output identified by the subscript): Lout = Lt (SF L ) MidLout = ((0.92) Lt + (0.38) Rt)) (SF MidL ) Cout = ((0.45) Lt + (0.45) Rt)) (SF C ) MidRout = ((0.38) Lt + (0.92) Lt)) (SF MIDR ) Rout = Rt (SF R )

in the In general, assuming an array of input channels, you can choose the nearest one Inputs with straight lines, what possible Represent decoder modules, connect conceptually. ("Possible" because, if there is no output channel that needs to be derived from a module that does not need the module). For typical Arrangements can be any output channel on a line between two input channels from a two-input module be derived (if sources and transmission channels in one common level, any source appears in at most two input channels, in which case there is no advantage in using more than two inputs). An output channel in the same position as an input channel is an endpoint channel, maybe more than one module. One Output channel that is not on a line or in the same position as an input (e.g., inside or outside of one formed by three input channels) Triangle), requires a module with more than two inputs.

Decoding modules with more than two inputs are useful when a common signal occupies more than two input channels. This can happen, for example, when source channels and input channels are not in one plane: a source channel can be mapped to more than two input channels. This happens in the example in 1 when mapping 24 channels (16 horizontal ring channels, 6 raised ring channels, 1 vertical channel and LFE) to 6.1 channels (including one composite vertical channel). In this case, the channel at the back of the center of the raised ring is not on a direct line between two of the source channels, it lies in the middle of one through the Ls ( 13 ), the Rs ( 9 ) and the upper channel ( 23 ), so that a three-input module is required to extract it. One way to map elevated channels to a horizontal array is to map each of them onto more than two input channels. This allows the 24 channels of the example in 1 on a conventional 5.1-channel array. In this alternative, a plurality of three-input modules can accommodate the increased Ka extract the remaining signal components, and the remaining signal components can be processed by two-input modules to extract the main horizontal ring of channels.

in the Generally, it is not necessary at all possible To test combinations of signal commonality among the input channels. at Planar channel arrays (e.g., in horizontally arranged array directions performing channels) enough it usually, Pairwise similarity comparisons spatial adjacent channels perform. In baldachin or ball surface-like arranged channels For example, the signal sharing may extend to three or more channels. Use and detection of signal sharing can also be serve additional Convey signal information. For example, a vertical signal component by mapping to all five full channels a horizontal five-channel array being represented.

decisions about that, which input channel combinations besides a default input / output mapping matrix to be analyzed on commonality, need while configuring of the translator or translator function only once per input / output channel translator or translator function arrangement to be met. The "initial Figure "(vor Processing) derives a passive "master" matrix which maps the input / output channel configurations into Relationship to spatial Alignment of the channels puts. As an alternative, the processor or processing part generate time-varying scalings, one per output channel, which either the output signal level of the matrix, the otherwise one simple, passive matrix, or the matrix coefficients change yourself. The scalings are derived from a combination of (a) basic, (b) evenly spread (Fill), and (c) residual (end point) signal components as described below.

A master matrix is useful when configuring an array of modules as in the example in 1 shown and related below 2 described. For example, by examining the master matrix, one can derive how many decoder modules are needed, how to connect them, how many input and output channels each module has, and the relationship of the matrix coefficients, inputs, and outputs of each module to each other, deduce. These coefficients can be taken from the master matrix; only the non-zero values are needed unless an input channel is also an output channel (ie an endpoint).

Each module preferably has a "local" matrix, which is the part of the master matrix applicable to the particular module. In the case of a multi-module arrangement such as the example in FIG 1 and 2 For example, the module may use the local matrix for purposes of generating scaling (or matrix coefficients) for controlling the master matrix as discussed below 2 and 4A - 4C or for the purpose of generating a subset of the output signals, which output signals are composed by a central method such as a monitor as described in connection with FIG 2 described, use. In the latter case, such a monitor compensates for multiple versions of the same output signal generated by modules having a common output signal in a manner analogous to the way in which monitors 201 in 2 a final scaling is determined to replace the provisional scalings generated by modules that generate tentative scaling for the same output channel.

in the Case of multiple modules, which are more scalable than output signals can generate such modules continually the self-concerned matrix information about a supervisor from a master matrix instead of having a local matrix. However, less computational effort is required if the module has its own matrix has. In the case of a single, independent module the module has a local matrix, which is the only one required Matrix is (actually the local matrix is the master matrix), and becomes that local one Matrix used to produce output signals.

If Unless otherwise indicated, descriptions of embodiments are provided the invention with several modules to the alternative, in which Modules generate scalings.

Any decoder module output channel having only a non-zero coefficient in the local matrix of the module (this coefficient is 1.0, since the coefficients in the sum of squares are 1.0) is an endpoint channel. Output channels with more than one nonzero coefficient are inner output channels. Consider a simple example. When output channels O1 and O2 are both derived from input channels I1 and I2 (but with different coefficient values), a two-input module connected between I1 and I2 is needed which generates outputs O1 and O2, possibly among others. In a more complex case, if there are 5 inputs and 16 outputs and one of the decoder modules is On I1 and I2 and feed outputs O1 and O2 so that: O1 = AI1 + BI2 + 0I3 + 0I4 + 0I5 (no input from input channels I3, I4 or I5) and O2 = CI1 + DI2 + 0I3 + 0I4 + 0I5 (no contribution from input channels I3, I4 or I5),
then the decoder can have two inputs (I1 and I2) and two outputs, and the scaling these are in relation to each other: O1 = AI1 + BI2 and O2 = CI1 + DI2.

Either the master matrix or the local matrix can, in the case of a single, independent Module have matrix elements that do more than multiplication can. For example, you can Matrix elements, as mentioned above, a filter function such as a phase or delay term and / or a Filter, which is a function of frequency included. An example a filtering that can be applied is a matrix out of sheer delays, which reprocessed phantom images can process. In practice For example, such a master or local matrix can be divided into two Functions are split: one that uses coefficients, around the output channels derive, and a second, which applies a filter function.

2 FIG. 4 is a block diagram showing an overview of a multi-band transform embodiment which is the example in FIG 1 implements, offers. For example, a PCM audio input with multiple interleaved audio signal channels is placed on a supervisor or a monitoring function 201 (in the following "supervisor 201 "), which includes means for canceling the interleaving, the separate streams of each one of six audio signal channels ( 1' . 3 ' . 5 ' . 9 ' . 13 ' and 23 ' ), which the interleaved input carries, and applies each to a time domain to frequency domain transform or transformation function (hereinafter "forward transform"). Alternatively, the audio channels may be received in separate streams, in which case no de-interleaving means is required.

As mentioned above, signal conversion in accordance with the present invention can be performed either on wideband signals or on each frequency band of a multiband processor which is either a filter bank such as a critical band discrete filter bank or a filter bank with a band structure compatible with a decoder associated therewith. or a transformation configuration such as a Fast Fourier Transform (FFT) or Modified Discrete Cosine Transform (MDCT) linear filter bank. 2 . 4A - 4C and other figures will be described in the context of a multi-band transform configuration.

In 1 . 2 and other figures are an optional LFE input channel (a possible seventh input channel in FIG 1 and 2 ) and output channel (a possible 24th output channel in 1 and 2 ) not shown for simplicity. The LFE channel can usually be handled in the same way as the other input and output channels, but with its own scaling fixed at "1" and its own matrix coefficient also fixed at "1". In cases where the source channels have no, but the output channels have an LFE (for example, 2: 5.1 upmixing), an LFE channel may become fifth by using a low pass filter applied to the sum of the channels (for example, a Butterworth filter Order with a cut-off frequency of 120 Hz) or, to avoid cancellation when adding the channels, a phase-corrected sum of the channels can be used. In cases where the input has an LFE channel but not the output, the LFE channel may be added to one or more of the output channels.

Continue with the description of 2 : Modules 24 - 34 receive corresponding of the six inputs 1' . 3 ' . 5 ' . 9 ' . 13 ' and 23 ' on the in 1 shown way. Each module generates a preliminary scaling ("PSF") output for each of its associated audio output channels, as in 1 shown. Thus, for example, Modul receives 24 inputs 1' and 3 ' and generates preliminary scaling outputs PSF1, PSF2 and PSF3. Alternatively, as mentioned above, each module may be for each of them linked audio output channels produce a preliminary set of audio outputs. Each module can also work with a supervisor 201 communicate as explained below. From the supervisor 201 Information sent to different modules may contain information about neighbor levels and higher order neighbor level information, if any. Information sent by each module to the supervisor may include the estimated total energy of the internal outputs attributable to each of the inputs of the module. The modules may be part of a control signal generating part of the entire system 2 be considered.

A supervisor like the supervisor 201 in 2 can perform a number of different functions. For example, a supervisor may determine if more than one module is in use, and if not, the supervisor need not perform any neighbor level related functions. During initialization, the supervisor may inform the or each module of the number of inputs and outputs it has, the matrix coefficients concerned, and the sample rate of the signal. As mentioned earlier, he can read the blocks of interleaved PCM samples and "de-box" them into separate channels. He may apply unlimited time domain measures, for example, in response to additional information indicating that the source signal was amplitude limited and indicating the degree of that limitation. When the system is operating in a multiband mode, it can apply fenestration and a filterbank (eg, FFT, MDCT, etc.) to each channel (so that multiple modules do not perform redundant transformations that significantly increase processing overhead) and streams of transform values to each module for processing. Each module returns a two-dimensional array of scales to the supervisor: a scale for all the transformers in each subband of each output channel (if in a multiband transform configuration, otherwise one scaling per output channel) or alternatively a two-dimensional array of output signals: a set of complex transform bins for each subband of each output channel (if in a multiband transform configuration, otherwise one output per output channel). The supervisor can smooth the scales and apply them to the signal path matrixing (matrix 203 , described below) to give complex output channel spectra (in a multiband transform configuration). Alternatively, when the module generates output signals, the supervisor may derive the output channels (in a multiband transform configuration: complex output channel spectra), thereby compensating for local matrices which produce the same output signal. It can then do an inverse transform and, in the case of MDCT, perform windowing and overlap addition for each output channel, interleaving the output samples to form a composite multichannel output stream (or, optionally, can do without interleaving to multiple Output streams) and sends it to an output file, sound card or other final destination.

Even though various functions by a supervisor, as described herein, or by several monitors accomplished can be A person of ordinary skill in the art will recognize that different or all of these functions rather in the modules themselves than by one all or some of the modules common monitors are running can. For example, if there is only a single, stand-alone module, needs no distinction between module functions and monitor functions to be made. Although in the case of several modules a common monitor eliminate the required overall processing power by eliminating or reducing redundant processing tasks, can be the elimination of a common supervisor or its simplification enable, that added modules easily can be for example, to upgrade to more output channels.

Back to the description of 2 - the six entrances 1' . 3 ' . 5 ' . 9 ' . 13 ' and 23 ' are also applied to a variable matrix or a variable matrixing function 203 (hereafter "matrix 203 ") applied matrix 203 can be considered part of the signal path of the system in 2 be considered. matrix 203 also receives as inputs from the supervisor 201 a set of final scalings SF1 to SF23 for each of the 23 output channels of the example in FIG 1 , The final scalings may be used as the output of the control signal portion of the system 2 be considered. As explained below, the monitor gives 201 preferably, as final scalings for the matrix, the preliminary scalings continue for each "inner" output channel, but in response to information received from modules, the supervisor determines final scalings for each endpoint output channel. An "inner" output channel is located between the two or more "endpoint" output channels of each module. Alternatively, if the modules produce output signals rather than scaling, then no matrix 203 required; the monitor itself generates the output signals.

In the example in 1 It is assumed that the end point output channels are connected to the input channel although they are not required to coincide, as further discussed elsewhere. Thus are output channels 2 . 4 . 6 - 8th . 10 - 12 . 14 - 16 . 17 . 18 . 19 . 20 . 21 and 22 inner output channels. The inner output channel 21 is located between or is bracketed by three input channels (input channels 9 ' . 13 ' and 23 ' ), while the other inner channels are each between two input channels (or bracketed by them). Because there are multiple tentative scalings for these endpoint output channels, which are shared between or under modules (ie, output channels 1 . 3 . 5 . 9 . 13 and 23 ), determines the supervisor 201 the final end point scalings (SF1, SF3, etc.) among the scalings SF1 to SF23. The final inner output scales (SF2, SF4, SF6, etc.) are the same as the preliminary scalings.

3 Figure 4 is a block diagram illustrating the manner in which a supervisor such as the supervisor 201 in 2 facilitate end point scaling. The supervisor does not sum all the outputs of the modules which share an input to obtain endpoint scaling. Instead, he combines additive, as in a combinator 301 , the estimated total internal energy for an input from each module sharing the input with others, such as input 9 ' which the modules are 26 and 27 in 2 share. This sum represents the total energy level at the input, which is consumed by the internal outputs of all connected modules. Then it subtracts this sum from the smoothed input energy level at that input (eg the output of the smoother 325 or 327 in 4B as described below) of any of the modules that share the input (module in this example) 26 or module 27 ), as in combinator 303 , It is sufficient to choose any of the smoothed inputs of the modules at the common input, although the levels may differ slightly from module to module because the modules adjust their time constants independently of each other. The difference at the output of combinator 303 is the desired output energy level at this input, which energy level must not go below zero. By dividing this desired output signal level by the smoothed input level at that input, as in dividers 305 and performing a square root operation as in block 307 , you get the final scaling (SF9 in this example) for this output. It should be noted that the supervisor derives a single final scaling for each such shared input no matter how many modules share the input. An arrangement for determining the estimated total energy of the inner outputs attributable to each of the inputs of the module is discussed below 6A described.

Because the levels are energy levels (a second order magnitude) as opposed to amplitudes (a first order magnitude), a square root operation is applied after the division operation to obtain the final scaling (scalings are associated with first order quantities). The addition of the internal levels and the subtraction from the total input level are all done in a pure sense of energy because it is assumed that internal outputs of different modules are independent (uncorrelated). If this assumption is not true even in an unusual situation, the computation may yield more left over signal at the input than should be there, which may result in a slight spatial distortion in the reproduced sound field (eg, a slight drag of other nearby inner images towards the input ), but in the same situation the human ear reacts similarly. The scaling for the inner output channels, such as PSF6 to PSF8 in module 26 , are passed by the supervisor as final scales (they are not changed). For simplicity shows 3 just the generation of one of the final endpoint scales. Other final endpoint scaling can be derived accordingly.

Back to the description of 2 As mentioned above, the variability in the variable matrix 203 complicated (all coefficients variable) or simple (groupwise varying coefficients as if applied to the inputs or outputs of a fixed matrix). Although each of the approaches can be used to produce substantially the same results, it has been found that one of the simpler approaches, that is, a fixed matrix with variable gain downstream for each output (with the gain of each output being scaled), is satisfactory Provides results, and this approach is used in the embodiments described herein. Although a variable matrix in which each matrix coefficient is variable is useful, it has the disadvantage of having more variables and requiring more processing power.

The supervisor 201 also performs an optional time-domain smoothing of the final scaling before going to the variable matrix 203 be applied. In a variable matrix system, output channels are never "turned off", the coefficients are arranged to amplify some signals and cancel out others. A system of fixed matrices with variable gains as in the Be however, embodiments of the present invention turn channels on and off and are more prone to unwanted "rattle" artifacts. These can occur despite the two-step smoothing described below (eg smoothing elements 319 / 325 etc.). For example, if scaling is close to zero because only a small change is needed to go from 'little' to 'nothing' and back, transitions to and from zero can cause audible clatter.

The by the supervisor 201 optional smoothing smoothes out the output scales with variable time constants, which depend on the size of the absolute difference ("abs-diff") between newly derived current scaling values and a running value of the smoothed scaling. For example, if the abs-diff is greater than 0.4 (and, of course, ≤ 1.0), little or no smoothing is applied; a small additional amount of smoothing is applied to abs-diff values between 0.2 and 0.4; and below a value of 0.2, the time constant is a continuous inverse function of the abs-diff. Although these values are not critical, it has been noted. that they reduce audible rattle artifacts. Optionally, in a multiband version of a module, the time constants of the scaling smoothing elements may also be in the manner of the frequency smoothing elements described below 413 . 415 and 417 in 4A scale with both frequency and time.

As stated above, the variable matrix is 203 preferably a fixed decoding matrix with variable scaling (gains) at the matrix outputs. Each matrix output channel may have (fixed) matrix coefficients, which would have been the encode-downmix coefficients for that channel, if there had been a discrete input coder (rather than mixing source channels directly to the downmixed array, disabling a discrete coder). Preferably, the coefficients in the sum of squares are 1.0 for each output channel. The matrix coefficients are fixed once it is known where the output channels are (as discussed above with respect to the "master" matrix), while the scaling controls the output gain of each channel are dynamic.

On the modules 24 - 34 in 2 Inputs containing frequency domain transformers may be grouped into frequency subbands by each module after initial energy and common energy levels at the bin level have been calculated, as discussed below. Thus, there is a preliminary scaling (PSF in 2 ) and a final scaling (SF in 2 ) for each frequency subband. The by matrix 203 generated frequency domain output channels 1 - 23 each contain a set of transformation compartments (subband-sized groups of transformation compartments are treated with the same scale). The sets of frequency domain transformers are implemented by a frequency domain to domain transform or transformation function 205 (hereinafter "inverse transformation"), which is a function of the supervisor 201 but shown separately for clarity, each in a set of PCM output channels 1 - 23 transformed. The supervisor 201 can the resulting PCM channels 1 - 23 to provide a single interleaved PCM output stream, or leave the PCM output channels as separate streams.

4A - 4C show a block diagram of a module according to one aspect of the present invention. The module receives two or more input signal streams from a supervisor, such as the supervisor 201 in 2 , Each input contains a set of complex frequency domain transformers. Each input, 1 to m, is assigned to a function or device (such as function or device 401 for input 1 and function or facility 403 for input m), which calculates the energy of each bin, which is the sum of the squares of the real and imaginary values of each transformation bin (only the paths for two inputs, 1 and m, are shown to simplify the drawing). Each of the inputs is also assigned to a function or device 405 which calculates the common energy of each bin over the input channels of the module. In the case of an FFT embodiment, this can be calculated by taking the cross product of the input samples (in the case of two inputs L and R, for example, the real part of the complex product of the complex L-value and the complex conjugate to the complex R-value ). Real-value embodiments only need to cross-multiply the real value for each input. For more than two inputs, the special cross multiplication method described below may be used, namely, if all signs are equal, the product will be given a positive sign, otherwise it will be given a negative sign and will be (always two: they are) according to the ratio of the number of possible positive results either all positive or all negative) to the number of possible negative results.

Pairwise Calculation of the common energy

For example, assume that an input channel pair A / B has a common signal X ne individual, uncorrelated signals Y and Z contains: A = 0.707X + Y B = 0.707X + Z where the scalings are 0.707 = √ 0.5 Provide for a power-efficient mapping to the nearest input channels.

Because X and Y are uncorrelated,

That because X and Y are uncorrelated, the total energy is in the input channel A is the sum of the energies of the signals X and Y.

Corresponding:

Since X, Y and Z are uncorrelated, the average cross product of A and B is:

Therefore is in the case of an output signal which is two adjacent ones input channels in half sharing, which is also independent, uncorrelated signals may contain the averaged cross product the signals equal to the energy of the common signal component in every channel. If the common signal is not divided in half is, i. when it is pivoted to one of the inputs is the mean cross product is the geometric mean between the Energy of the common components in A and B, resulting from normalizing with the square root of the ratio the channel amplitudes estimates the common energy of individual channels can be derived. actual Time averages are calculated after smoothing stages, such as described below.

calculation common energy higher order

Around derive the combined energy of decoding modules with three or more inputs, must be averaged Cross products of all input signals are formed. Easy to do one paanrweisen processing of inputs does not distinguish between separate output signals between each single pair of inputs and an all common signal.

For example, consider three input channels A, B, and C, which consist of uncorrelated signals W, Y, Z, and a common signal X: A = X + W B = X + Y C = X + Z

When the average cross-product is calculated, all terms be deleted, the combinations of W, Y, and Z contain, as in the second order calculation, whereby the average value of X ³ remains:

Unfortunately, if X is a zero-mean-time signal, as expected, the mean of its cube is zero. Unlike averaging X ² , which is positive for any nonzero value of X, X ^{3 has} the same sign as X, and therefore positive and negative contributions tend to cancel out. Obviously, the same holds true for any odd power of X, corresponding to an odd number of module inputs, but even exponents greater than two may lead to false results; For example, four inputs with components (X, X, -X, -X) have the same product / mean as (X, X, X, X).

This Problem leaves by using a variant of the method of averaged products to solve. Before is averaged, the sign of each product is removed by the absolute value of the product is taken. The omens of each individual terms of the product are examined. If they are all the same are, the absolute value of the product is on the average determining means applied. If one or more of the signs differ from the different, is the negative value of the absolute value of the product averaged. As the number of possible Combinations with the same sign are not necessarily the same is like the number of possible Combinations with different signs will become a weighting factor out of proportion the number of combinations with the same sign in number consists of combinations with different signs on the applied to compensated negated absolute value products. To the For example, in a three-input module, there are two ways that the signs are the same; of eight, there are six options that the signs are different, resulting in a scaling of 2/6 = 1/3 results. This compensation causes the integrated or summed product grows in a positive direction, though and only if there is an all entrances a common decoding module signal component.

In order to the mean values of modules of different order are comparable, have to however, they all have the same dimensions. In a conventional Second order correlation is about averages of two-input multiplications and therefore with sizes the dimensions of energy or power. Consequently, the terms to be averaged are changed into higher order correlations, so they too have the dimensions of performance. For a correlation k. Order the individual product absolute values are therefore raised to the power of 2 / k before being averaged.

Of course, regardless order, the individual input energies of a module, if necessary, as the average of the square of the corresponding input signal is calculated become; they do not need to be in the k first. Potency raised and then to a size second Order to be reduced.

Back to the description of 4A The transform box outputs of each one of the blocks may be provided by a function or facility 407 . 409 respectively 411 be grouped into subbands. For example, the subbands can approximate the critical bands of the human ear. The remainder of the module embodiment in FIG 4A - 4C works separately and independently on each subband. To simplify the drawing, only the operation is shown on a sub-band.

Each subband of blocks 407 . 409 and 411 is applied to a frequency smoother or a frequency smoothing function 413 . 415 respectively 417 (hereinafter referred to as "frequency smoothing member"). The purpose of the frequency smoothing elements is explained below. Each frequency-smoothed subband from a frequency smoothing element is considered to be optional "fast" smoothing or smoothing functions 419 . 421 respectively 423 (hereafter referred to as "fast smoothing terms") which provide time-domain smoothing. Although preferred, the fast smoothing terms can be eliminated if the time constant of the fast smoothing terms is close to the block length time of the forward transform that generated the input bins (e.g., a forward transform in the supervisor 201 in 1 ). The fast smoothing terms are "fast" relative to the "slow" variable time constant smoothing or smoothing functions 425 . 427 and 429 (hereafter "slow smoothers") which receive the respective outputs of the fast smoothers. Examples of time constant values of fast and slow smoothing terms are given below.

therefore no matter if fast smoothing through the inherent Function of a forward transformation or by a fast smoothing link a two-step smoothing function is preferred, in which is the second, slower level variable. However, one can only smoothing stage deliver acceptable results.

The time constants of the slow smoothing elements within a module are preferably synchronous with each other. This can be done, for example, by applying the same control information to each slow smoother and by configuring each slow smoother so that they are in the same Way to respond to applied control information. The derivation of the information for controlling the slow smoothing members is described below.

Preferably, each pair of smoothing members is the same as the pairs 419 / 425 . 421 / 427 and 423 / 429 connected in series as in 4A and 4B shown in which a fast smoothing member feeds a slow smoothing member. A series arrangement has the advantage that the second stage resists short, fast signal peaks at the input of the pair. However, by configuring the pairs of smoothing elements in parallel, similar results can be achieved. For example, in a parallel arrangement, the second stage resistance can be implemented in a series arrangement against short, fast signal peaks in the logic of a time constant controller.

each Stage of the two-stage smoothing elements can through a single-pole low-pass filter (a "leaky integrator") as an RC low-pass filter (in an analog embodiment) or equivalent a first order low pass filter (in a digital embodiment) be implemented. For example, in a digital embodiment the first order filters each as a "biquad" filter, a common IIR filter second order, in which some of the coefficients are set to zero are, be realized, so that the filter like a filter first Order works. Alternatively you can the two smoothing members combined into a single second-order biquad stage, although it is simpler, coefficient values for the second (variable) level when it is separated from the first (fixed) stage.

It should be noted that in the embodiment in FIG 4A . 4B and 4C all signal levels are expressed as energy levels (squared levels) except when an amplitude is required by pulling a square root. Smoothing is applied to the energy levels of applied signals, whereby the smoothers detect rms values rather than averages (mean value smoothers are fed by linear amplitudes). Because the signals applied to the smoothing members are squared levels, the smoothing members respond to sudden increases in signal level faster than average smoothing members, as increases are amplified by the squaring function.

The two-stage smoothing elements thus provide a time average for each subband of the energy of each input channel (that of the 1st channel is provided by a slow smoother 425 and that of the m. Channel through a slow smoothing link 427 supplied) and the average for each subband of the common energy of the input channels (is provided by a slow smoother 429 provided).

The energy average outputs of the slow smoothing elements ( 425 . 427 . 429 ) are on combinators 431 . 433 respectively 435 in which (1) the neighboring energy levels (if any) (for example, supervisor 201 in 2 ) are subtracted from the smoothed energy level of each one of the input channels, and (2) the higher order neighbor energy levels (if any) (for example, supervisor 201 in 2 ) are subtracted from each one of the energy average outputs of the slow smoothing elements. For example, every entrance has 3 ' receiving module ( 1 and 2 ) two adjacent modules and receives neighbor energy level information which compensates for the effect of these two adjacent modules. However, none of these modules is a "higher order" module (ie, all modules, which input channel 3 ' share are two-input modules). In contrast, module is 28 ( 1 and 2 ) an example of a module in which a higher order module shares one of its inputs. Thus, for example, in module receives 28 the energy average output of a slow smoother for input 13 ' a compensation of neighboring levels of higher order.

The resulting "neighbor-compensated" energy levels for each subband of each individual input of the module are assigned to a function or device 437 which calculates a nominal current fundamental direction of these energy levels. The direction indication can be calculated as the vector sum of the energy-weighted inputs. For a two-input module, this is simplified to the L / R ratio of the smoothed and neighbor-compensated input signal energy levels.

For example, consider a planar surround array in which the positions of the channels for the case of two inputs are given as 2-tuples representing x, y coordinates. It is assumed that the listener in the center is at about (0, 0). The left front channel in normalized spatial coordinates lies on (1, 1). The right front channel lies on (-1, 1). When the left input amplitude (Lt) is 4 and the right input amplitude (Rt) is 3, using these amplitudes as weighting factors, the nominal current basic direction is: (4 * (1, 1) + 3 * (-1, 1)) / (4 + 3) = (0.143, 1) or a little left of center on a left and right connecting horizontal line.

Alternatively, the spatial direction, after defining a master matrix, may be expressed in matrix coordinates rather than in physical coordinates. In this case, the input amplitudes normalized to square sum one are the effective matrix coordinates of the direction. In the above example, the levels on the left and right are 4 and 3, normalized to 0.8 and 0.6. Consequently, the "direction" is (0.8, 0.6). In other words, the nominal current basic direction is a square-rooted one of the square root of the neighbor-compensated smoothed input energy level. block 337 generates the same number of outputs indicating a spatial direction as the number of inputs of the module (two in this example).

The on the direction-determining function or device 337 Applied neighbor-compensated smoothed energy levels for each subband of each of the module's inputs are also limited to a function or device 339 which computes the neighbor-compensated cross-correlation ("neighbor-compensated_xcor"). block 339 Also receives as an input the average shared energy of the inputs of the module for each subband from the slow variable smoothing element 329 which in combinator 335 with neighboring energy levels of higher order, if any, has been compensated. The neighbor-compensated cross-correlation is in block 339 as the compensated higher order common ground energy divided by the root M. where M is the number of inputs calculates the product of the neighbor compensated smoothed energy levels for each of the input channels of the module to give a true mathematical correlation value in the range 1.0 to derive -1.0. Preferably, values from 0 to -1.0 are considered zero. Neighbor-compensated_xcor provides an estimate of the cross-correlation that exists in the absence of other modules.

The neighbor compensated_xcor from block 339 is then placed on a weighting device or function 341 which weights the neighbor compensated_xcor with the neighbor compensated direction information to produce a directionally weighted neighbor compensated cross correlation ("direction weighted_xcor"). The weighting increases with increasing deviation of the nominal current base direction from a centered state. In other words, unequal input amplitudes (and therefore energies) cause a proportional increase in direction weighted_xcor. Direction-weighted_xcor provides an estimate of the compactness of the sound image. Thus, in the case of a two-input module with, for example, a left input L and a right input R, the weighting increases, while the direction from the center either deviates to the left or to the right (ie, at the same degree of deviation from the center) Weighting in each direction the same). For example, in the case of a two input module, the value "Neighbor Compensated_xcor" is weighted by an L / R or R / L ratio so that uneven signal distribution drives the direction weighted_xcor to 1.0. For such a two-input module,
if R ≥ L, Direction Weighted_xcor = (1 - ((1 - Neighbor Compensated_xcor) * (L / R)), and if R <L, Direction Weighted_xcor = (1 - ((1 - Neighbor Compensated_xcor) · (R / L))

at Modules with more than two inputs requires the calculation of the direction-weighted_xcor from the neighbor-weighted_xcor for example, replacing the ratio L / R or R / L in the above by a "uniformity" reading, which varies between 1.0 and 0. For example, the uniformity measurement for one any number of inputs to calculate are the input signal levels with the total input power to normalize, which in normalized input levels, which is in one Sum of energy (squared) to 1.0 results. Then Each normalized input level is normalized by the corresponding normalized input level Input level of a signal centered in the array divided. The smallest ratio the uniformity Reading. That's why, for example, with a three-input module with a zero level input, the uniformity reading is zero and the direction weighted_xcor equals one. (In this case lies the signal on the edge of the three-input module on a line between two of his entrances, and a (lower in the hierarchy) two-input module determines where on the Line is the nominal current basic direction and how wide along this line the output should be spread.)

Back to the description of 4B The direction-weighted_xcor is further weighted by its application to a function or device 443 which applies a "random_xcor" weight, to create an "Effective_xcor". Effective_xcor provides an estimate of the distribution of input signals.

random_xcor is the mean cross product of the input quantities divided by the square root of the medium input energies. The value of Random_xcor can be determined by assuming that the output channels originally Module input channels and calculate the value of xcor, which comes from all these channels independent, but equal-level signals, which are passively downmixed, results are calculated. According to this approach, in case of a three-output module with two inputs, Random_xcor calculated as 0.333 and will, for the case of a five-output module (three inner outputs) with two entrances, random_xcor calculated as 0.483. The value of Random_xcor needs for each Module to be calculated only once. Although it was determined that such values of random xcor provide satisfactory results, the values are not critical and can at your discretion Other values are used by the system designer. A change the value of random_xcor affects the dividing line between the two workspaces of the signal distribution system as described below. The exact location This dividing line is not critical.

The by function or device 343 Running random_xcor weighting can be viewed as a re-normalization of the value "direction weighted_xcor" to get an effective_xcor: Effective_xcor = (Direction Weighted_xcor - Random_xcor) / (1 - Random_xcor), if direction-weighted_xcor ≥ random_xcor, otherwise Effective_xcor = 0

Random_xcor weighting speeds up the decrease of direction-weighted_xcor during a decrease of direction-weighted_xcor below 1.0, so if direction-weighted_xcor equals random_xcor is, the value "Effective_xcor" is zero. Because the outputs of a module represent directions along an arc or line, Values of Effective_xcor that are less than zero are considered equal Treated zero.

Information for controlling the slow smoothing elements 325 . 327 and 329 are derived from the energies of the non-neighbor compensated, slow and fast smoothed input channels and from the common energy of the slow and fast smoothed input channels. In particular, a function or device calculates 345 in response to the energies of the rapidly-smoothed input channels and the common energy of the rapidly-smoothed input channels, a fast, non-neighbor-compensated cross-correlation. A function or device 347 calculates, in response to the fast-smoothed input channel energies, a fast non-neighbor compensated direction (ratio or vector, as discussed above in connection with the description of Block 337 discussed). A function or device 349 calculates a slow non-neighbor-compensated cross-correlation in response to the energies of the slowly-smoothed input channels and the common energy of the slowly-smoothed input channels. A function or device 351 calculates a slow non-neighbor compensated direction (ratio or vector, as discussed above) in response to the slowly smoothed input channel energies. The fast non-neighbor-compensated cross-correlation, the fast non-neighbor-compensated direction, the slow non-neighbor-compensated cross-correlation and the slow non-neighbor-compensated cross-correlation are combined with direction-weighted_xcor from block 341 to a facility or function 353 applied the information to control the variable slow smoothing elements 325 . 327 and 329 to adjust its time constants (hereafter "adjustment time constants"). Preferably, the same control information is applied to each variable slow smoother. Unlike the other variables fed into the time constant selection box which compare a fast to a slow measurement, the directional weighted xcor is preferably used without reference to any fast value so that when the absolute value of the directional weighted xxcor is greater than a threshold, adjustment time constants 353 to select a faster time constant. Rules for the operation of the "adaptation time constant" 353 are set out below.

In general, in a dynamic audio system, it is desirable to use, as far as possible, slow time constants which remain at a motionless value to minimize audible breaks in the reproduced sound field, unless a "new event" occurs in the audio signal, in which case it is desirable that a control signal change rapidly to a new, motionless value and then stay at that value until another "new event" occurs. Typically, audio processing systems have balanced amplitude changes in a "new event". However, when working with cross-products or cross-correlation, the new and the amplitude are not always the same: a new event can cause a decrease in the cross-correlation. By sensing changes in module-related parameters, namely, cross-correlation and direction measurements, the time constants of a module can become faster and quickly assume a new control state, as desired.

To The consequences of improper dynamic behavior include walking the image, rattling (a fast turn on / off channel), pumps (unnatural level changes) and, in a multi-band embodiment, Twittering (rattling and pumping on a band-by-band basis). Some of these effects are especially critical regarding the quality of isolated channels.

An embodiment like that of 1 and 2 uses a grid of decoding modules. Such a configuration brings two classes of dynamics problems: dynamics between modules and dynamics in modules. In addition, the various ways of implementing audio processing (e.g. broadband, multiband using FFT or MDCT with linear filter bank or discrete filter bank for critical bands or others) each require their own optimization of dynamic behavior.

The elementary decoding process in each module depends on a measurement of the energy ratios of the input signals and a measurement of the cross-correlation of the input signals (in particular, the above-described direction-weighted correlation (direction-weighted_xcor); the output of Block 341 in 4B ), which together control the signal distribution to the outputs of a module. The derivation of such elementary quantities requires smoothing, which in the time domain requires the calculation of a time-weighted average of the instantaneous values of these quantities. The range of time constants required is quite large: very short (for example 1 ms) for fast transient changes in signal conditions to very long (eg 150 ms) for low correlation values, where the instantaneous variation is likely to be much larger than the true averaged value.

A common one Method for implementing the behavior of variable time constants is, in analogous terms, the use of an "acceleration" diode. If the Instantaneous level exceeds the averaged level by a threshold amount, conducts the diode, causing a shorter one effective time constant. A disadvantage of this method is that a momentary peak in an otherwise stationary entrance a big change of the smoothed Level, which then decreases very slowly, what an unnatural Emphasis on isolated tips, which otherwise has little audible consequences would have entails.

In connection with the embodiment of the 4A - 4C The correlation calculation described above makes the use of acceleration diodes (or their DSP equivalent) problematic. For example, all the smoothing elements in a particular module preferably have synchronized time constants so that their smoothed levels are comparable. Therefore, a comprehensive (equalized) time constant switching mechanism is preferred. In addition, a rapid change in signal conditions is not necessarily linked to an increase in the common energy level. The use of an accelerating diode for this level is likely to produce biased, inaccurate estimates of the correlation. Therefore, embodiments of aspects of the present invention preferably use two-stage smoothing without a diode-equivalent acceleration. Estimates of correlation and direction may be derived at least from both the first and second smoothing stage stages to adjust the second stage time constant.

For each Pair of smoothing links (e.g., 319/325), the time constant of the first stage, the solid fast level, set to a fixed value such as 1 ms. The time constants of the second stage, the variable slow stage, can for example, from 10 ms (fast), 30 ms (medium) and 150 ms (slow) selectable. Although it has been found that such time constants are satisfactory Deliver results, their values are not critical and can after discretion Other values are used by the system designer. Furthermore can the time constant values of the second stage tend to be continuously variable to be discreet. The choice of time constants does not just have to on the signal conditions described above, but also can based on a hysteresis mechanism that works with a "quick flag" that works serves to ensure that if a serious fast transient Tip occurs, the system stays in fast mode and use the mean time constant avoids until the signal conditions again enable the slow time constant. This can help ensure rapid adaptation to new signal conditions.

The decision as to which of the three possible time constants of the second stage should be used can be made by means of "adaptation time constants". 353 according to the following rules for the case of two entrances:
If the absolute value of direction weighted_xcor is less than a first reference value (for example, 0.5) and the absolute difference between fast non-neighbor compensated_xcor and slow non-neighbor compensated_xcor is less than the same first reference value and the absolute difference between the fast and the slow Direction ratio (each of which has a range +1 to -1) is smaller than the same first reference value, the slow time constant of the second stage is used and the "quick flag" is set to "true", thereby allowing the subsequent selection of the mean time constant becomes.

Otherwise, if the "quick flag" is "true", the absolute Difference between the fast and the slow non-neighbor compensated_xcor greater than the first reference value and less than a second reference value (for example, 0.75) is the absolute difference between the fast and the slow transient L / N ratio greater than is the first reference value and less than the second reference value and the absolute value of the direction-weighted_xcor greater than the first reference value and smaller than the second reference value, the mean time constant becomes the second stage selected.

Otherwise the fast time constant of the second stage is used and the "quick flag" is set to "false", causing the subsequent one Use of the mean time constant is suppressed until the slow time constant is selected again.

In In other words, the slow time constant is chosen when all three states less than a first reference value, the mean time constant is selected, if all states between a first reference value and a second reference value lie and the previous state was the slow time constant, and the fast time constant is chosen when one or more the conditions are greater than the second reference value.

Even though was found that the rules and reference values just mentioned Provide satisfactory results, they are not critical and can at your discretion of the system developer modifications of the rules or other rules, which fast and slow cross-correlation and fast and slow Take direction be used. About that can out other changes be made. For example, it may be easier, but the same be effective, equivalent to one of the diode acceleration Processing, but with the same mode of operation, so if any smoothing member in a module in fast mode, also all other smoothing elements be switched to fast mode. It may also be desirable be, separate smoothing members for time constant determination and having signal distribution, the smoothing terms for time constant determination have fixed time constants and only the signal distribution time constants be varied.

Because even in fast mode, the smoothed signal levels more Need milliseconds, to adapt, a time delay can be built into the system so that control signals can adapt before moving on to a signal path be applied. In a broadband embodiment can this delay as a discrete delay (for example 5 ms) in the signal path. For multiband (transformation) versions is the delay a natural one Sequence of block processing, and when the analysis of a block ago signal path matrices performed this block must, no express delay to be required.

Multiband embodiments of aspects of the invention use the same time constants and rules as broadband versions, except that the sampling rate of the smoothing terms is at the signal sampling rate divided by the block size (e.g. the block rate) can be adjusted so that in the smoothing elements used coefficients are adjusted correctly.

For frequencies below 400 Hz, in multiband embodiments, the time constants are preferably scaled inversely to the frequency. In the broadband version, this is not possible because there are no separate smoothing terms for different frequencies; therefore, as a partial compensation, a moderate bandpass / pre-emphasis filter can be applied to the input signal of the control path to emphasize medium and mid high frequencies. This filter may have, for example, a two-pole high-pass characteristic with a cut-off frequency of 200 Hz and a two-pole low-pass characteristic with a cut-off frequency of 8000 Hz and a predistortion network, 6 dB boost between 400 Hz and 800 Hz and a further 6 dB boost between 1600 Hz and 3200 Hz. Although it was found that such a filter is suitable, the filter properties are not critical and other parameters can be used at the discretion of the system designer.

In addition to time-domain smoothing, multiband versions of aspects of the invention also preferably use frequency-domain smoothing, as discussed above 4A described (frequency smoothing 413 . 415 and 417 ). For each block, the non-neighbor compensated energy levels may be averaged with a sliding frequency window and adjusted to approximate a 1/3 octave bandwidth (critical band bandwidth) before being applied to the subsequent time domain processing described above. Because the transform-based filter banks inherently have linear frequency resolution, the width of this window (in number of transform coefficients) increases with increasing frequency, and usually is only one transform coefficient at low frequencies (below about 400 Hz). Therefore, all smoothing applied to multiband processing relies more on low-frequency, time-domain smoothing and higher-frequency frequency-domain smoothing, where fast timing is sometimes more likely to be required.

Continue to the description of 4C - preliminary scaling (in 2 represented as "PSFs") which ultimately affect the ground / fill / end point signal distribution may be through a combination of devices or functions 455 . 457 and 459 which calculate "basic" scaling components, "fill" scaling components, "excess endpoint energy" scaling components, respective normalizers or normalization functions 361 . 363 and 365 and a device or function 367 which produces either the largest of the fundamental and fill scaling components and / or the additive combination of the fill scaling components and the excess endpoint energy scaling components. The preliminary scalings can be sent to a supervisor such as a supervisor 201 in 2 be sent if the module is one of a plurality of modules. Preliminary scaling can each have a range from zero to one.

Dominant scale factor components

In addition to Effective_xcor receives setup or function 355 ("Calculate basic scaling components") the neighbor-compensated direction information from block 337 and information regarding the local matrix coefficients from a local matrix 369 such that it can determine the N nearest output channels (where N = number of inputs) that can be applied to a weighted sum to give the coordinates of the nominal current basic direction, and apply the "base" scaling components to them, to give the basic coordinates. The output of block 355 is either a scaling component (per subband) when the nominal current direction of origin coincides randomly with an output direction, or else consists of several scaling components (one per number of inputs per subband) that bracket the nominal current direction and are applied in appropriate proportions to pan or map the base signal to the correct virtual position in a power efficient sense (ie, for N = 2, the two associated fundamental channel scaling components in the sum of squares should yield Effective_xcor).

For a two-input module are all output channels in a line or arc, making it a natural sequence (from "left" to "right"), and it lights up without further ado, which channels each other the next lie. For those discussed above hypothetical case with two input channels and five output channels with sine / cos coefficients as shown, it can be assumed that the nominal current basic direction (0.8, 0.6) is between the middle left channel ML (0.92, 0.38) and the central channel C (0.71, 0.71). This can be done by finding two consecutive channels be achieved, in which the L coefficient greater than the L coordinate of the nominal current basic direction is and the Channel to the right has an L coefficient, the small one as the base L coordinate is.

The basic scaling components are allocated to the two nearest channels in the sense of constant power. To do this, a system of two equations and two unknowns is solved, where the unknowns are the fundamental scaling component of the channel to the left of the basic direction (SFL) and the corresponding scaling component to the right of the nominal current fundamental direction (SFR). (These equations are for SFL and SFR solved). Erste_Grundkoord = SFL · Matrix value 1 of the left channel + SFR · Matrix value 1 of the right channel Second_round_coord = SFL · matrix value 2 of the left channel + SFR · matrix value 2 of the right channel

It Note that with the left and right channels here the channels are meant, which enclose the nominal current basic direction, not the input channels L and R of the module.

The solution is the calculations of the anti-ground level of each channel, normalized to the sum of squares 1.0, which are used as the basic distribution scaling components (SFL, SFR) for the other channel, respectively. In other words, the anti-ground value of an output channel having coefficients A, B for a signal having coordinates C, D is the absolute value of AD-BC. For the considered numerical example: Anti-ground (channel ML) = abs (0.92 x 0.6 - 0.38 x 0.8) = 0.248 Anti-ground (channel C) = abs (0.71 x 0.6 - 0.71 x 0.8) = 0.142 (where "abs" means taking the absolute value).

Normalizing the latter two numbers to the sum of squares 1.0 gives values of 0.8678 and 0.4969, respectively. Thus, when these values are switched to the opposite channels, they are the basic scaling components (note that the value of the base scale, before the directional weighting, is the square root of Effective_xcor): ML reason sf = 0.4969 · sqrt (Effective_xcor) C reason sf = 0.8678 · sqrt (Effective_xcor) (The fundamental signal is closer to Cout than to MidLout).

The use of the anti-ground component of one channel, normalized, as the basic scaling component of the other channel becomes easier to understand, considering what happens when the nominal current basic direction randomly points precisely to one of the two selected channels. Suppose that the coefficients of one channel are [A, B] and the coefficients of the other channel are [C, D] and that the coordinates of the nominal current basic direction [A, B] are (pointing to the first channel), then : Anti-ground (first channel) = abs (AB - BA) Anti-ground (second channel) = abs (CB - DA)

It Note that the first anti-ground value is zero. If the two anti-ground signals on Square sum 1.0, the second anti-ground value is equal 1.0. After switching receives the first channel has a basic scaling component of 1.0 (times square root from Effective_xcor) and receives the second channel 0.0, as desired.

If this approach is extended to modules with more than two inputs, there is no longer any natural order that occurs when the channels are in a line or arc. Again, for example, block calculates 337 in 4B the coordinates of the nominal current fundamental direction by taking the input amplitudes after adjacent compensation and normalizing them to square sum one. block 455 in 4B for example, then the N identifies nearest channels (where N = number of inputs) that can be applied to a weighted sum to give the base coordinates. (Note: Distance or proximity can be calculated as the sum of the squared coordinate differences, as if they were (x, y, z) space coordinates). Thus, one does not always select the N nearest channels because they must be weighted to give the nominal current basic direction.

For example, assume a three-input module fed by a triangle of channels: Ls, Rs, and Up like in 5 , Suppose there are three closely spaced inner output channels near the bottom of the triangle with the following local matrix coefficients of the module: [0.71, 0.69, 0.01], [0.70, 0.70, 0 , 01] or [0.69, 0.71, 0.01]. Suppose the nominal current fundamental direction is slightly below the center of the triangle and has the coordinates [0.6, 0.6, 0.53]. (Note: the center of the triangle has the coordinates [0.5, 0.5, 0.707].) The three nearest channels to the nominal current current direction are these three inner channels at the bottom, but they add up to 0 when scaling is used and 1 not to the base coordinates, so instead choose two from the bottom and the top end channel to distribute the fundamental signal and solve the three equations for the three weighting factors to complete the reason computation and with continue the fill and end point calculations.

In the examples in 1 and 2 There is only a single three-input module and this is used to derive only a single inner channel, which simplifies the calculations.

Fill scale factor components

In addition to Effective_xcor receives setup or function 357 ("Calculate Filling Scaling Components") Random_xcor, direction-weighted_xcor from block 341 , "EQUIAMPL"("EQUIAMPL" is defined and explained below) and information regarding the local matrix coefficients from the local matrix (in case the same fill scaling component is not applied to all outputs, as discussed below 14B illustrated). The output of block 457 is a scaling component for each module output (per subband).

As explained above, Effective_xcor is equal to zero if the direction weighted_xcor is less than or equal to Random_xcor. If direction-weighted_xcor ≥ Random_xcor, the fill scaling component is for all output channels Fill Scaling Component = sqrt (1 - Effective_xcor) · EQUIAMPL

therefore is when direction-weighted_xcor = Random_xcor, the Effective_xcor equals 0, so (1 - Effective_xcor) is 1.0, hence the filling amplitude scaling component equal to EQUIAMPL (which ensures output power = Input power in this state). This point is the maximum value the fill scaling components to reach.

If Weighted_xcor is less than Random_xcor, the base scaling component (s) is zero and the fill scaling components as the direction-weighted xcor approaches zero are reduced to zero: Fill Scaling Component = sqrt (Direction Weighted_xcor / Random_xcor) · EQUIAMPL

therefore is at the border, where direction-weighted xcor = random xcor, the provisional Fill scale factor component again equal EQUIAMPL, whereby consistency with the results of above equation for the Case that direction-weighted_xcor is greater than Random_xcor becomes.

Not only is a value of Random_xcor associated with each decoder module, but also a value of "EQUIAMPL", which is a scaling value that all scalings should have, if the signals are bleach-distributed, thus saving power, namely: EQUIAMPL = square root of (number of decoder module input channels / number of decoder module output channels)

For example, for a two-input module with three outputs: EQUIAMPL = sqrt (2/3) = 0.8165 where "sqrt ()" means "square root of ()"

For a two-input module with four outputs: EQUIAMPL = sqrt (2/4) = 0.7071

For a two-input module with five outputs: EQUIAMPL = sqrt (2/5) = 0.6325

Although such EQUIAMPL values have been found to provide satisfactory results, the values are not critical and other values may be used at the discretion of the system designer. Changes in the value of EQUIAMPL affect the levels of the output channels for the "stuffing" condition (intermediate correlation of the input signals) with respect to the levels of the output channels for the condition "basic" (maximum correlation of input signals) and the "pure endpoints" condition (minimum correlation of input signals).

Endpoint scale factor components

In addition to neighbor-compensated_xcor (from block 439 . 4B ) receives device or function 359 ("Calculate Scaling Components for Excess End-Point Energy") the respective smoothed non-neighbor-compensated energy of the 1st to m. Input (from blocks 325 and 327 ) and, optionally, information on the local matrix coefficients from the local matrix (in case one or both of the endpoint outputs of the module do not coincide with an input and the module applies excess endpoint energy to the two outputs with the direction of the input of nearest directions as discussed below). The output of block 359 is a scaling component for each endpoint output when the directions coincide with input directions, otherwise it consists of two scaling components, one for each of the outputs closest to the end, as discussed below.

However, those of Block 359 Excess endpoint energy scaling components did not generate the only "endpoint" scaling components. There are three other sources of endpoint scaling components (in the case of a single, stand-alone module, two):
First, within the calculations of the tentative scaling of a particular module, the endpoints are potential candidates for basic signal scaling components of Block 355 (and normalizer 361 ).

Second, in the "fill" calculation of block 357 (and normalizer 363 ) in 4C , which treats endpoints along with all inner channels as potential fill candidates. Any non-zero fill scaling component can be applied to all outputs, even the endpoints and the selected base outputs.

Third, if there is a grid of multiple modules, a supervisor (such as supervisor 201 in the example in FIG 2 ) a final fourth assignment of the "endpoint" channels, as discussed above 2 and 3 described.

With it block 459 calculate the "excess endpoint energy" scaling components, the total energy at all internal outputs based on neighbor-compensated_xcor is thrown back to the inputs of the module to estimate how much of the energy of the inner outputs is contributed by each input ("inner end") Energy at input n "), and this energy is used to calculate the excess endpoint energy scaling component at each module output that coincides with an input (ie endpoint).

Throwing back the internal energy to the entrances is also required by a supervisor such as a supervisor 201 in 2 to provide needed information to compute neighbor levels and neighbor levels of higher order. One way to calculate the internal energy contribution at each of the inputs to a module and determine the excess endpoint scaling component for each endpoint output is in FIG 6A and 6B shown.

6A and 6B are block diagrams, each in a module such as any of the modules 24 - 34 in 2 a suitable arrangement for (1) generating the estimated total internal energy for each input of a module, 1 to m, in response to the total energy at each input, 1 to m, and (2) in response to the neighbor compensated_xcor (see 4B , the output of block 439 ), Generating a scale component for excess endpoint energy for each of the endpoints of the module. The estimated total internal energy for each input of a module ( 6A ), in the case of a multi-module arrangement, is needed by the supervisor, and in each case the module itself, to generate the excess end point energy scaling components.

Using the in blocks 455 and 457 in 4C derived scaling components, among other information, calculates the order in 6A the estimated total energy at each inner output (but not at the endpoint outputs). Using the calculated inner output energy levels, it multiplies each output level by the matrix coefficient which relates that output to each input ["m" inputs, "m" multipliers], which provides the energy contribution of that input to that output. For each input, it sums all the energy contributions of all the inner output channels, to gain the entire internal energy contribution of this input. The total internal energy contribution of each input is reported to the supervisor and used by the module to calculate the excess endpoint energy scaling component for each endpoint output.

Out 6A In particular, it can be seen that the smoothed total energy level for each (preferably non-neighbor-compensated) module input is applied to a set of multipliers, each to a multiplier for each of the module's inner outputs. To simplify the illustration shows 6A two inputs "1" and "m" and two internal outputs "X" and "Z". The smoothed total energy level for each module input is multiplied by a matrix coefficient (the local matrix of the module) which relates the respective input to one of the module's internal outputs (note that the matrix coefficients are their own inverses because) Matrix coefficients in the square sum one). This happens for every combination of input and internal output. Consequently, as in 6A shown, the smoothed total energy level at input 1 (which, for example, at the output of the slow smoothing element 425 in 4B can be won) on a multiplier 601 which applies this energy level with a matrix coefficient that converts the inner output X into input 1 multiplied, resulting in a scaled output energy level component X ₁ at output X. Accordingly, multipliers deliver 603 . 605 and 607 scaled energy level components X _m , Z ₁ and Z _m .

The energy level components for each inner output (eg X ₁ and X _m ; Z ₁ and Z _m ) are combined 611 and 613 according to neighbor-compensated_xcor amplitude / power summed. When the inputs of a combiner are in phase, as indicated by a neighbor-weighted cross-correlation of 1.0, their linear amplitudes add up. If they are uncorrelated, as indicated by a neighbor-weighted cross-correlation of zero, their energy levels will add up. If the cross-correlation is between one and zero, the sum is partly an amplitude sum and partly a power sum. To properly sum the inputs of each combiner, both the amplitude sum and the power sum are calculated and weighted by neighbor-compensated_xcor and (1-neighbor-weighted_xcor), respectively. To obtain the weighted sum, either the square root of the power sum is taken to obtain an equivalent amplitude, or the linear amplitude sum is squared to obtain its power level before the weighted sum is calculated. For example, taking the latter approach (weighted sum of powers) when the amplitude levels are 3 and 4 and Neighbor-weighted_xcor, the amplitude sum is 3 + 4 = 7, or a power level of 49, and the power power sum is 9 + 16 = 25. Thus, the weighted sum is 0.7 * 49 + (1-0.7) * 25 = 41.8 (power energy level) or, after pulling the square root, 6.47.

The summation products (X ₁ + X _m , Z ₁ + Z _m ) become multipliers for each of the X and Z outputs 613 and 695 multiplied by the scaling components to produce the total energy level at each inner output, which can be identified as X 'and Z'. The scaling component for each of the inner outputs becomes block 467 ( 4C ) won. It should be noted that the "Scaling components for excess endpoint energy" from Block 459 ( 4C ) do not affect internal exits and not those of the arrangement in 6A calculations are involved.

The total energy level at each inner output X 'and Z' is thrown back to the respective inputs of the module by multiplying each by a matrix coefficient (the local matrix of the module) which relates the respective output to each of the inputs of the module becomes. This happens for every combination of inner output and input. Consequently, as in 6A shown, the total energy level X 'at the inner output X to a multiplier 617 which uses the energy level with a matrix coefficient that relates the inner output X in relation to the input 1 sets (which, as mentioned above, is identical to its inverse), which multiplies a scaled energy level component X ₁ 'at input 1 supplies.

It It should be noted that if a value of second order as the total energy level X 'by a value first order is weighted as a matrix coefficient Second order weight is required. This is equivalent to pull the square root of the energy to get an amplitude Multiply this amplitude by the matrix coefficient and Squaring the result to return to an energy value.

Accordingly, multipliers deliver 699 . 621 and 623 scaled energy levels X _m ', Z ₁ ' and Z _m '. The energy components to each output (eg X ₁ 'and Z ₁ ', X _m 'and Z _m ') are in combiners 625 and 627 according to neighbor-compensated_xcor amplitude / power summed, as above in the context with combinators 611 and 613 described. The outputs of the combiners 625 and 627 set the estimated total internal energy for the inputs 1 In the case of a grid with multiple modules, this information is sent to the supervisor as a supervisor 201 in 2 so that the supervisor can calculate neighbor levels. The supervisor collects all internal total energy contributions from each input from all modules connected to that input and then informs each module for each of its inputs, which was the sum of all other total internal energy contributions from all other modules connected to that input. The result is the neighbor level for this input of this module. Generation of neighbor level information is described below.

The through each of the inputs 1 and estimated total internal energy contributed is also required by the module to calculate the excess endpoint energy scaling component for each endpoint output. 6B shows how such scaling component information can be calculated. For ease of illustration, only the calculation of scaling component information for an endpoint is shown, it being understood that a corresponding calculation is performed for each endpoint output. The through an entrance like entrance 1 Contributed estimated total internal energy is in a combinator or a combine function 629 from the smoothed total input energy for this same input, in this example input 1 subtracted (the same smoothed total energy level at input 1 , obtained, for example, at the output of the slow smoothing element 425 in 4B which is based on a multiplier 601 is applied). The result of the subtraction will be in divider or divide function 631 by the smoothed total energy level for the same input 1 divided. The square root of the result of the division is in a square root puller or a square root pull function 633 drawn. It should be noted that the divider or the divide function 631 (as well as other dividers described herein) should contain a denominator zero test. In this case, the quotient can be set to zero.

If it is only one, independent Module gives, are the provisional Endpoint scale factor components therefore due to the already made determination of the basic scaling, Füllskalierungen and Scales for excess endpoint energy certainly.

Thus, scalings are assigned to all output channels, including endpoints, so that they can now be used to perform signal path matrixing. However, if there is a grid of multiple modules, each input is assigned the endpoint scaling of each module connected to it, so that each input to which more than one module is connected has multiple scaling assignments from each connected module. In this case, the supervisor (like supervisor 201 in the example in 2 ) a final fourth assignment of the "endpoint" channels, as discussed above 2 and 3 described. In doing so, the supervisor determines final endpoint scalings that nullify any scaling assignments made by individual modules as endpoint scaling.

In There are no assurances that it is actually a practical arrangement an endpoint position corresponding output channel direction, although this is often the Case is. If there is no physical endpoint channel, but at least gives a physical channel beyond the endpoint, the endpoint energy becomes on the nearest to the end physical channels panned, as if it were a basic signal component. In a horizontal array, these are the two of the endpoint positions nearest Channels, which preferably use a constant energy distribution (the two scalings result in the sum of squares 1.0). In other Words, if a sound direction is not the position of a real one Sound channel corresponds, even if this direction is an end point signal is, it is preferred to take them to the nearest available pair of real channels to swing, because the sound, if it is moved slowly, suddenly from one output channel to another jumps. So it is, if there is no physical endpoint sound channel, inappropriate, an endpoint signal to pan to one of the endpoint positions nearest sound channel, except there is no physical channel beyond the endpoint, in which case There is no choice but to reach the endpoint position nearest To pan sound channel.

Another way to implement such panning is for the supervisor as the supervisor 201 in 2 in that, based on an assumption that each input also has a corresponding output channel (ie, corresponding inputs and outputs coincide so that they each represent the same position) to produce "final" scalings. Then an output matrix like the variable matrix 203 in 2 mimic an output channel to one or more suitable output channels if there is no actual output channel directly corresponding to an input channel.

As mentioned above, the outputs of each one of the devices or functions 455 . 457 and 459 ("Calculate Scaling Component") to respective normalizers or functions 461 . 463 and 465 applied. Such standardization devices are desirable because of the blocks 455 . 457 and 459 calculated scaling components are based on neighbor-compensated levels, while the final signal path-matrixing (in the case of multiple modules in the master matrix, or in the case of a stand-alone module in the local matrix) requires non-adjacent compensated levels (the input signals applied to the matrix are not neighbor-compensated). Typically, scaling components are reduced in value by a normalizer.

A suitable way to implement normalizers is as follows. Each normalizer receives the neighbor compensated smoothed input energy for each of the module's inputs (as from combiners 331 and 333 ), the non-neighbor-compensated smoothed input power for each of the inputs of the module (as from blocks 325 and 327 ), local matrix coefficient information from the local matrix, and the respective outputs of the blocks 355 . 357 and 359 , Each normalizer calculates, assuming a scale of 1, a desired output for each output channel and an actual output level for each output channel. Then, it divides the computed desired output for each output channel by the calculated actual output level for each output channel and subtracts the square root of the quotient to apply a possible preliminary scaling to "Sum and / or Larger of". 367 to win. Consider the following example.

Assuming the smoothed non-neighbor-compensated input energy levels of a two-input module are 6 and 8, and the corresponding neighbor-compensated energy levels are 3 and 4. Further assume a central inner output channel has matrix coefficients = (0.71, 0.71) , or squared: (0.5, 0.5). If the module selects an initial scaling for this channel (based on neighbor compensated levels) of 0.5, or squared 0.25, then the desired output level of that channel (under the simplistic assumption of pure power summation and using neighbor compensated levels ): 0.25 x (3 x 0.5 + 4 x 0.5) = 0.875.

Because the actual input levels are 6 and 8 when the above scaling (squared) of 0.25 is used for final signal path matrialization, the output level is 0.25 x (6 x 0.5 + 8 x 0.5) = 1.75 instead of the desired output level of 0.875. The normalizer adjusts the scaling to obtain the desired output level when non-neighbor compensated levels are used.
Actual output, assuming that SF = 1, = (6 x 0.5 + 8 x 0.5) = 7.
(Desired output level) / (Actual output, assuming SF = 1) = 0.875 / 7.0 = 0.125 = final scaling, squared
Final scaling for this output channel = sqrt (0.125) = 0.354 instead of the initially calculated value of 0.5.

"Sum and / or Greater of" 367 Preferably, it sums the corresponding fill and end point scaling components for each output channel per subband and selects the larger of the fundamental and fill scaling components for each output channel per subband. The function of the block "Sum and / or Greater of" 367 in their preferred form can as in 7 be described. Namely, basic scaling components and fill scaling components become a device or function 701 applied, which selects the larger of the scaling components for each output ("Larger of" 701 ) and apply it to an additive combiner or combiner function 703 applies which or which of the scaling components from "Larger of" 701 is summed with the end point energy scales for each output. Alternatively, acceptable results can be obtained if "Sum and / or Greater of" 467 : (1) sums in region 1 and region 2, (2) takes the larger value in region 1 and region 2, or (3) takes the larger value in region 1 and sums in region 2.

8th Figure 11 is an idealized illustration of the manner in which one aspect of the present invention generates scaling components in response to a measure of cross-correlation. This figure is particularly useful in connection with the examples in FIG 9A and 9B to 16A and 16B , As mentioned above, the generation of scaling components may be considered to comprise two regions or workspaces: a first region, region 1, bounded by "pure ground" and "uniformly filled" in which the available scaling components are a mixture of ground and fill -Skalierungs components, and a second region, Region 2, bounded by "Equally Filled" and "Pure Endpoints" in which the available scaling components are a mixture of fill scaling components and excess endpoint energy scaling components. The boundary condition "pure ground" exists when the direction-weighted_xcor equals one. Region 1 (base and fill) extends from this boundary to the point where the direction-weighted_xcor equals Random_xcor, the "uniformly-filled" state. The limit condition "pure endpoints" is present when the direction-weighted_xcor equals zero. Region 2 (fill and end point) extends from the boundary condition "evenly filled" to the boundary condition "pure endpoints". The "equally filled" boundary may be considered to be either in Region 1 or Region 2. As mentioned below, the exact boundary point is not critical.

8th illustrates that as the fundamental scaling component (s) decreases in value (decreases), the fill scaling components increase in value, reaching a maximum when the fundamental scaling component (s) reaches a null value, at which As the fill scaling components decrease in value, the scaling components for excess endpoint energy increase in value. The result, when applied to an appropriate matrix receiving the input signals of the module, is an output signal distribution which provides a compact sound image when the input signals are highly correlated, which diverge (widen) and decrease as the correlation decreases from compact to wide even further, up to the strongly uncorrelated state, decreasing correlation gradually splits from the broad state into several sound images at one end point or bends outwards.

Although it is desirable that there be a single spatially compact sound image (in the nominal current direction of the input signals) in the case of full correlation and a plurality of spatially compact sound images (each at an end point) in the case of a correlation equal to zero spatially spread sound image between these extremes in other ways than in 8th be achieved shown. For example, it is not critical that the values of the fill scale components for the case Random_xcor = Direction Weighted_xcor reach a maximum, or that the values of the three scale components change linearly, as shown. Changes in relationships in 8th (and the equations herein) and other relationships between a suitable cross-correlation reading and scaling values that are useful, the compact basal to wide spread to compact endpoint signal distribution for a cross correlation measure from highly correlated to highly uncorrelated are also considered by the present invention. For example, rather than gaining a compact ground-to-widespread to compact endpoint signal distribution, rather than using a two-region approach as described above, such results can be achieved through a mathematical approach, such as by using the pseudo-inverse-based equation resolution.

Examples of output scales

A series of idealized representations, 9A and 9B to 16A and 16B , illustrate the output scales of a module for various examples of input signal conditions. For simplicity, a single, standalone module is assumed so that the scalings it produces for a variable matrix are the final scalings. The module and a variable matrix associated with it have two input channels (such as "Left" L and "Right" R) which coincide with two endpoint output channels (which may also be referred to as L and R). In this series of examples there are three inner output channels (such as "left center" Lm, "center" C and "right center" Rm).

The meanings of "Pure Basic", "Basic and Mixed", "Evenly Filled", "Blend and Endpoints Mixed", and "Pure Endpoints" are used in conjunction with the examples in 9A and 9B to 16A and 16B explained in more detail. In each pair of figures (for example, 9A and 9B), the figure "A" shows the energy levels of two inputs, "Left" L and "Right" R, and shows the figure "B" Scaling components for the five outputs, "Left" L, Left Center Lm, Center C, Right Center Rm and Right R. The figures are not to scale.

In 9A For example, the input energy levels shown as two vertical arrows are the same. In addition, both the direction-weighted_xcor (and the effective_xcor) is 1.0 (full correlation). In this example, there is only a single nonzero scaling, in 9B is represented as a single vertical arrow at C, which is applied to the central inner channel C output, resulting in a spatially compact fundamental signal. In this example, the output is centered (L / R = 1) and therefore coincidentally coincides with the central inner output channel C. If there is no matching output channel the fundamental signal is applied in appropriate proportions to the nearest output channels to sweep the fundamental signal to the correct virtual position between them. For example, if there were no central output channel C, the left-center output channels Lm and right-centered Rm would have scalings not equal to zero so that the fundamental signal would be equally applied to the outputs Lm and Rm. In this case of full correlation (pure fundamental), there are no fill and no endpoint signal components. So they are by block 467 ( 4C ) generated the same provisional scalings as those by block 361 generated normalized fundamental scaling components.

In 10A For example, the input power levels are equal but direction weighted_xcor is less than 1.0 and greater than Random_xcor. Thus, the scaling components are those of region 1 - mixed primitive and fill scaling components. The larger of the normalized basic scaling component (from block 361 ) and the normalized fill-scale component (from block 363 ) becomes (by block 367 ) is applied to each output channel so that the basic scaling is at the same central output channel C as in 10B but is smaller, and the fill scales appear on each of the other output channels, L, Lm, Rm, and R (including endpoints L and R).

In 11A the input power levels remain the same, but direction-weighted_xcor random_xcor. Consequently, the scales, 11B , those of the boundary condition between regions 1 and 2 - the uniformly filled state in which there are no base or end point scales, only fill scales with the same value at each output (hence "evenly filled"), as by the same Arrows appear at each exit. The fill scale levels reach their highest value in this example. As discussed below, fill scales may be applied unevenly, as in a decreasing manner, depending on input signal conditions.

In 12A the input energy levels remain the same, but direction-weighted_xcor is less than Random_xcor and greater than zero (Region 2). Consequently, there is, as in 12B shown, fill and endpoint scaling, but no basic scaling.

In 13A the input power levels remain the same, but direction-weighted_xcor is zero. Consequently, the in 13B shown scaling those of the boundary condition "pure endpoints". There are no inner output scales, only endpoint scaling.

In the examples in 9A / 9B to 13A / 13B is the direction-weighted_xcor (as by block 441 in 4B generated) equal to the neighbor-compensated_xcor (as by block 439 in 4B generated) because the energy levels of the two inputs are the same. In 14A however, the input power levels are not equal (L is greater than R). Although in this example the neighbor-weighted_xcor is equal to random_xcor, those in 14B shown resulting scaling not uniformly applied to all channels Füll scalings as in the example in 11A and 11B , Instead, the unequal input energy levels cause a proportional increase in the direction weighted_xcor (proportional to the extent to which the nominal current basic direction deviates from its central position) to become larger than the neighbor compensated_xcor, causing the scaling to increase more Weighted "purely basic" (as in 8th shown). This is a desirable result because heavily L- or R-weighted signals should not have a large width; they should have a compact width near the end point of the L or R channel. The in 14B The resulting output shown consists of a nonzero ground scale closer to the L output than to the R output (the neighbor compensated directional information, in this case, places the fundamental component coincidentally exactly at the left center position Lm). Scaling amplitudes and no endpoint scaling (the directional weight pushes operation in region 1 in 8th (mixed ground and fill)).

The scalings in 14B corresponding five outputs can be expressed as follows: Lout = Lt (SF L ) MidLout = ((0.92) Lt + (0.38) Rt)) (SF MidL ) Cout = ((0.45) Lt + (0.45) Rt)) (SF C ) MidRout = ((0.38) Lt + (0.92) Lt)) (SF MIDR ) Rout = Rt (SF R ).

Thus, in the example in 14B Although the scaling (SF) for each of the four outputs except MidLout are equal (Fill), the corresponding signal outputs are not equal because Lt is greater than Rt (resulting in more signal output to the left) and the baseline output is at "center left "is greater than the scale indicates. Because the nominal current basic direction coincides with the "center left" output channel, the ratio of Lt to Rt is the same as the matrix coefficients for the "left center" output channel, 0.92 to 0.38. Suppose that these are the actual amplitudes for Lt and Rt. To calculate the output levels, multiply these levels by the corresponding matrix coefficients, and then add and scale with the respective scalings: Output amplitude (output channel_sub_i) = sf (i) * (Lt_Koeff (i) * Lt + Rt_Koeff (i) * Rt)

Although one considers preferably the mixture of amplitude and energy addition (as in the with 6A related computations), in this example the cross-correlation is quite high (large basic scaling) and a normal summation can be performed: Lout = 0.1 × (1 × 0.92 + 0 × 0.38) = 0.092 MidLout = 0.9 * (0.92 x 0.92 + 0.38 x 0.38) = 0.900 Cout = 0.1 x (0.71 x 0.92 + 0.71 x 0.38) = 0.092 MidRout = 0.1 x (0.38 x 0.92 + 0.92 x 0.38) = 0.070 Rout = 0.1 * (0 * 0.92 + 1 * 0.38) = 0.038

therefore shows this example that because Lt is greater than Rt, the signal outputs to Lout, Cout, MidRout and Rout are unequal, though the scalings for this outputs are the same.

The fill scales can be equally distributed to the output channels as in the examples in FIG 10B . 11B . 12B and 14B shown. Alternatively, instead of being uniform, the fill scaling components may in some way function as a function of the (correlated) ground and / or (uncorrelated) endpoint input signal components (or, equivalently, as a function of the "directionally weighted_xcor" value) the position can be varied. For example, with moderately high values of direction weighted_xcor, the fill scaling component amplitudes may curve convexly such that output channels near the nominal current direction of the base receive more signal levels than more distant channels. For direction-weighted_xcor = random_xcor, the fill-scale component amplitudes may flatten to a uniform distribution, and for direction-weighted_xcor <random_xcor, the amplitudes may curve concavely, favoring channels near the end-point directions.

Examples of such curved fill scaling amplitudes are in 15B and 16B shown. The exit in 15B results from an input ( 15A ) which is the same as described above 10A is. The exit in 16B results from an input ( 16A ) which is the same as described above 12B is.

Communication between Module and monitor regarding neighbor levels and neighbor levels of higher order

• (a) einen zum Auslesen und Berichten der Informationen, welche der Überwacher benötigt, um Nachbarpegel und Nachbarpegel höherer Ordnung (wenn vorhanden) zu berechnen. Die vom Überwacher benötigten Informationen bestehen aus der jedem der Eingänge des Moduls zuzuschreibenden geschätzten inneren Gesamtenergie wie zum Beispiel durch die Anordnung in 6A erzeugt.
• (b) einen anderen zum Empfangen und Anwenden der Nachbarpegel (wenn vorhanden) und Nachbarpegel höherer Ordnung (wenn vorhanden) vom Überwacher. Im Beispiel in 4B werden die Nachbarpegel in jeweiligen Kombinatoren 431 und 433 von den geglätteten Energiepegeln jedes Eingangs subtrahiert und werden die Nachbarpegel höherer Ordnung (wenn vorhanden) in jeweiligen Kombinatoren 431, 433 und 435 von den geglätteten Energiepegeln jedes Eingangs und der gemeinsamen Energie über die Kanäle subtrahiert.

Each module in an arrangement of several modules as in the example in 1 and 2 requires two mechanisms to communicate between him and a supervisor such as a supervisor 201 in 2 to support:

• (a) one to read out and report the information the supervisor needs to compute neighbor levels and higher order neighbor levels (if any). The information required by the supervisor consists of the estimated total internal energy attributable to each of the inputs of the module, such as the arrangement in FIG 6A generated.
• (b) another to receive and apply the neighbor levels (if any) and higher order neighbor levels (if any) from the supervisor. In the example in 4B become the neighbor levels in respective combiners 431 and 433 subtracted from the smoothed energy levels of each input and become the higher order neighbor levels (if any) in respective combiners 431 . 433 and 435 is subtracted from the smoothed energy levels of each input and the common energy over the channels.

• (1) ermittelt er, ob die geschätzten inneren Gesamtenergiebeiträge jedes Eingangs (summiert aus allen mit diesem Eingang verbundenen Modulen) den verfügbaren Gesamtsignalpegel an diesem Eingang übersteigen. Wenn die Summe den verfügbaren Gesamtpegel übersteigt, skaliert der Überwacher jede von jedem mit diesem Eingang verbundenen Modul berichtete innere Energie zurück, so dass sie sich zum Gesamt-Eingangspegel summieren.
• (2) informiert er jedes Modul über seine Nachbarpegel an jedem Eingang als die Summe aller anderen inneren Energiebeiträge dieses Eingangs (wenn vorhanden).

Once a supervisor knows all the estimated total internal energy contributions of each input of each module:

• (1) it determines whether the estimated total internal energy contributions of each input (summed from all modules connected to that input) exceed the available total signal level at that input. If the sum exceeds the total available level, the supervisor scales each internal energy reported by each module associated with that input to add up to the total input level.
• (2) it informs each module of its neighbor level at each input as the sum of all other internal energy contributions of that input (if any).

Higher order neighbor levels are neighbor levels of one or more higher order modules that share the inputs of a lower level module. The above calculation of neighbor levels only affects modules at a particular input which are on the same hierarchical level: all three input modules (if any), then all two input modules, etc. A higher order neighbor level of a module is the sum of all Neighbor level of all higher order modules at this input (ie, the higher order neighbor level on one input of a two input module is the sum of all third, fourth and higher order modules, if any, sharing the nodes of a two input module) , Once a module knows what its higher order neighbor levels are at a particular one of its inputs, it subtracts it, along with the neighbor levels of the same hierarchical level, from the total input power level of that input to gain the neighbor compensated level at that input node. This is in 4B shown where the neighbor level for input 1 and input m in combiners 431 respectively 433 from the outputs of the variable slow smoothing elements 425 and 427 are subtracted and the higher level neighbor levels for input 1 , Input m and the common energy in combinators 431 . 433 respectively 435 from the outputs of the variable slow smoothing elements 425 . 427 and 429 be subtracted.

One difference between the use of neighbor levels and higher order neighbor levels for compensation is that the higher order neighbor levels are also used to compensate for the common energy across the input channels (eg, by subtracting a higher order neighbor level in combiner 435 ). The explanation for this difference is that the common level of a module is not affected by adjacent modules of the same hierarchical level, but can be affected by a higher order module that shares all the inputs of a module.

For example, let's say input channels Ls (Left Surround), Rs (Right Surround) and Up, with an inner output channel in the middle of the triangle between them (raised ring at the back), and an inner output channel on a line between Ls and Rs (horizontal Main ring at the back), the former output channel requires a three-input module to recover the signal common to all three inputs. Then the latter output channel, being on a line between two inputs (Ls and Rs), requires a two-input module. However, the common overall signal level observed by the two-input module includes common elements of the three-input module which do not belong to the latter output channel; therefore, one subtracts the square root of the paired products of the higher order neighbor levels from the common energy of the two input module to determine how much common energy is due solely to its inner channel (the latter). Consequently, in 4B from the smoothed common energy level (from block 429 ) subtracts the higher-order derived common level to produce a neighbor-compensated common energy level (from combinator 435 ), which is used by the module to generate (in block 439 ) calculate the neighbor compensated_xcor.

The The present invention and its various aspects can be found in Analog circuits or more likely as software functions, which in digital signal processors, programmed digital Universal computers and / or special digital computers are running, be implemented. Interfaces between analog and digital signal streams can in appropriate hardware and / or as functions in software and / or Firmware executed be. Although the present invention and its various aspects analog or digital signals can be used in practical Applications most or all processing functions are likely in the digital field on digital signal streams, in which audio signals are represented by samples executed.

Claims (52)

A method of converting M audio input signals, each associated with a direction, into N audio output signals, each associated with one direction, where N is greater than M and M is two or more, including providing one or more variable matrices ( 203 ; 24 - 34 ) for each variable matrix ( 203 ; 24 - 34 ) applying m of the M audio input signals thereto and deriving n of the N audio output signals therefrom, where if there is only one variable matrix, m equals M and n equals N, and if there are a plurality of arrays , m is a subset of M and n is a subset of N, and the values of m and n for one matrix may differ from those of other matrices, controlling each variable matrix in response to the m input signals applied to it such that one of the m input signals Sound field generated from its derived n output signals has a compact sound image in the nominal current basic direction of the m input signals applied to it, when such input signals are highly correlated, the sound image divergence from compact to wide as the correlation decreases, and even further to the strongly uncorrelated state, decreasing correlation gradually turns into several compact sound images each in one with one attached to them set the input signal associated direction, and deriving the N audio output signals from the output signals of the one or more variable arrays ( 203 ; 24 - 34 ). The method of claim 1, wherein each of the variable matrices ( 203 . 24 - 34 ) is a variable coefficient variable matrix or is a variable matrix with fixed coefficients and variable outputs and each variable matrix ( 203 ; 24 - 34 ) is controlled by varying the variable coefficients or by varying the variable outputs. The method of claim 1, wherein each of the variable matrices ( 203 ; 24 - 34 ) is controlled in response to measurements of: (1) the relative levels of their input signals, and (2) the cross-correlation of their input signals. Method according to Claim 3, in which, for a measure of the cross-correlation, the variable matrix ( 203 ; 24 - 34 ) input values having values in a first range bounded by a maximum value and a reference value, the sound field having a compact sound image when the measure of cross-correlation is the maximum value, and a wide-spread sound image when the measure of cross-correlation is the reference value, and for a measure of cross-correlation to a variable matrix ( 203 ; 24 - 34 ) input values having values in a second range bounded by the reference value and a minimum value, the sound field having the wide-apart sound image when the measure of the cross-correlation is the reference value, and a plurality of compact sound images each in one with the variable matrix ( 203 ; 24 - 34 ) applied input signal when the measure of cross-correlation is the minimum value. The method of claim 4, wherein the reference value is approximately the value of a measure of the cross correlation of each variable matrix ( 203 ; 24 - 34 ) in the case of equal energy in each of the output signals of each such variable matrix. Method according to Claim 3, in which a measure of the relative levels of the input signals to the variable matrix ( 203 ; 24 - 34 ) in response to a smoothed energy level of each such input signal. Method according to claim 3 or claim 6, wherein a measure of the relative levels of the input signals of the variable matrix ( 203 ; 24 - 34 ) is a nominal current direction of such input signals. Method according to Claim 3, in which a measure of the cross-correlation of the input signals of a variable matrix ( 203 ; 24 - 34 ) in response to a smoothed common energy of such input signals divided by the Mth root of the product of the smoothed energy level of each such input signal, where M is the number of input signals of the variable matrix ( 203 ; 24 - 34 ). Method according to one of claims 6, 7 or 8, in which the smoothed energy level of each input signal of a variable matrix ( 203 ; 24 - 34 ) is obtained by time-domain smoothing with variable time constant. Method according to one of claims 6, 7 or 8, in which the smoothed energy level of each input signal of a variable matrix ( 203 ; 24 - 34 ) is obtained by frequency-domain smoothing and time-domain smoothing with variable time constant. Method according to Claim 8, in which the common energy of the input signals of a variable matrix ( 203 ; 24 - 34 ) is obtained by cross-multiplying the amplitude levels of these input signals. Method according to Claim 11, in which the smoothed common energy of the input signals of a variable matrix ( 203 ; 24 - 34 ) is obtained by time-domain smoothing with variable time constant of the common energy of the input signals. The method of claim 12, wherein the smoothed energy level of each input signal of a variable matrix ( 203 ; 24 - 34 ) is obtained by time-domain smoothing with variable time constant. Method according to Claim 11, in which the smoothed common energy of the input signals of a variable matrix ( 203 ; 24 - 34 ) is obtained by frequency-domain smoothing and time-domain smoothing with variable time constant of the common energy of these input signals. The method of claim 14, wherein the smoothed energy level of each input signal of a variable matrix ( 203 ; 24 - 34 ) is obtained by frequency-domain smoothing and time-domain smoothing with variable time constant. Method according to any one of claims 9, 10, 12, 13, 14 and 15, in which the time domain smoothing with variable time constant by smoothing with both a fixed time constant and a variable time constant accomplished becomes. Method according to any one of claims 9, 10, 12, 13, 14 and 15, in which the time domain smoothing with variable time constant by smoothing is executed with only one variable time constant. The method of claim 16 or claim 17, wherein which the variable time constant is variable in steps. The method of claim 16 or claim 17, wherein which the variable time constant is continuously variable. A method according to claim 16 or claim 17, wherein the variable time constant is in response to measures of the relative levels of the input signals of a variable matrix ( 203 ; 24 - 34 ) and their cross-correlation. Method according to Claim 6, in which the smoothed energy level of each input signal of a variable matrix ( 203 ; 24 - 34 ) is obtained by time-domain smoothing with variable time constant, the energy level of each input signal with substantially the same time constant. Method according to Claim 3, in which the measure of the relative levels of the input signals of a variable matrix ( 203 ; 24 - 34 ) and their cross-correlation are each obtained by time-domain smoothing with variable time constant, in which the same time constant is applied to each smoothing. The method of claim 8, wherein the measure of cross-correlation is a first measure of the cross-correlation of the input signals of a variable matrix ( 203 ; 24 - 34 ), and an additional measure of the cross-correlation is obtained by applying a measure of the relative levels of such inputs to the first measure of cross-correlation to produce a weighted measure of the cross-correlation. The method of claim 23, wherein a further additional measure of the cross-correlation of the input signals of a variable matrix ( 203 ; 24 - 34 ) by applying a scaling which is approximately equal to a value of a measure of the cross-correlation of such input signals in the case of equal energy in each of the outputs of the variable matrix ( 203 ; 24 - 34 ) is won. The method of claim 1, wherein when there is more than one variable matrix ( 203 ; 24 - 34 ), the variable matrices ( 203 . 24 - 34 ) are also controlled in response to information showing the effect of a or several other variable matrices ( 203 ; 24 - 34 ), which receive the same input signal, compensate. The method of claim 1 or claim 25, wherein deriving the N audio output signals from the output signals of more than one variable matrix ( 203 ; 24 - 34 ) compensating for multiple variable matrices ( 203 ; 24 - 34 ) which produce the same output signal. A method of converting M audio input signals, each associated with a direction, into N audio output signals, each associated with one direction, where N is greater than M and M is two or more, comprising providing a variable M : N matrix ( 203 ), which responds to control signals, the application of the M audio input signals to the variable matrix ( 203 ), providing one or more generators for variable matrix control signals ( 24 - 34 . 201 ), for each generator, applying m of the M audio input signals thereto and deriving a set of variable matrix control signals for n of the N audio output signals therefrom, where m is a subset of M and n is a subset of N and the values of m and n for a generator may differ from those of other generators, the control of each generator for variable matrix control signals ( 24 - 34 . 201 ) in response to the m input signals applied thereto, such that when the control signals generated by it are applied to the variable M: N matrix ( 203 ), a sound field generated by the n generated output signals in the nominal current basic direction of m input signals which generated the applied control signals has a compact sound image when such input signals are highly correlated, the sound image decreasing in correlation from compact to wide State, and even further, as far as the uncorrelated state, decreasing correlation gradually breaks up into a plurality of compact sound images each in a linked direction with an input signal generating the applied control signals, and deriving the N audio output signals the variable matrix ( 203 ). The method of claim 27, wherein the variable M: N matrix ( 203 ) is a variable coefficient variable matrix or a variable matrix with fixed coefficients and variable outputs, and the variable matrix ( 203 ) is controlled by varying the variable coefficients or by varying the variable outputs. A method according to claim 27, wherein each of the generators for variable matrix control signals ( 24 - 34 . 201 ) is controlled in response to measures of (1) the relative levels of the m input signals applied to the generator, and (2) the cross-correlation of the m input signals applied to the generator. The method of claim 29, wherein for a measure of cross-correlation the m input signals applied to a generator with values in a first range, bounded by a maximum value and a Reference value, the sound field has a compact sound pattern, if the Measure of Cross correlation is the maximum value, and a wide divergent one Sound picture has, if the measure of Cross correlation is the reference value, and for a measure of the cross correlation of the m input signals applied to a generator with values in a second range bounded by the reference value and a Minimum value, the sound field the broadly divergent sound image has, if the measure of Cross-correlation is the reference value, and a variety of compact Sound pictures in each case with one attached to the generator Linked input signal Direction has, if the measure of Cross correlation is the minimum value. The method of claim 30, wherein the reference value about the value of a measure the cross-correlation of the m input signals applied to each generator for the Case of equal energy in each of the output signals of the variable Matrices is. The method of claim 29, wherein a measure of the relative Level of the m input signals of a generator in response to a smoothed Energy level of each such input signal is. The method of claim 29 or claim 32, wherein which is a measure of relative level of the m input signals of the generator a nominal current basic direction of such input signals. The method of claim 29, wherein a measure of the cross-correlation of the m input signals of a generator in response to a smoothed common energy of such input signals divided by the mth root of the product of the smoothed energy level of each such input signal, where m is the number of input signals of the generator. Method according to any of the claims 32, 33 and 34, in which the smoothed Energy level of each input signal of a generator by time-domain smoothing with variable time constant is obtained. Method according to any of the claims 32, 33 and 34, in which the smoothed Energy level of each input signal of a generator by frequency domain smoothing and Time domain smoothing is obtained with variable time constant. The method of claim 34, wherein the common Energy of the input signals of a generator by cross-multiplication such levels of input signal amplitudes is won. The method of claim 37, wherein the smoothed common Energy of the m input signals of a generator by time domain smoothing with variable time constant of the common energy of the input signals is won. The method of claim 38, wherein the smoothed energy level each input signal of a generator by time domain smoothing with variable time constant is obtained. The method of claim 37, wherein the smoothed common Energy of the input signals of a generator by frequency domain smoothing and Time domain smoothing with variable time constant of the common energy of such input signals is won. The method of claim 40, wherein the smoothed energy level each input signal of a generator by frequency domain smoothing and Time domain smoothing is obtained with variable time constant. Method according to any one of claims 35, 36, 38, 39, 40 and 41, wherein the time domain smoothing with variable time constant by smoothing with both a fixed time constant and a variable time constant accomplished becomes. Method according to any one of claims 35, 36, 38, 39, 40 and 41, wherein the time domain smoothing with variable time constant by smoothing is performed with only a variable time constant. A method according to claim 42 or claim 43, wherein which the variable time constant is variable in steps. A method according to claim 42 or claim 43, wherein which the variable time constant is continuously variable. A method according to claim 42 or claim 43, wherein which the variable time constant in response to measures of relative Level of the m input signals of a generator and their cross-correlation is controlled. The method of claim 32, wherein the smoothed energy level each input signal of a generator by time domain smoothing with variable time constant of the energy levels of each input signal with essentially the same time constant is obtained. The method of claim 29, wherein the dimensions of the relative Level of the input signals of a generator and their cross-correlation each by time domain smoothing with variable time constant, in which on each smoothing the same time constant is applied. The method of claim 34, wherein the measure of cross-correlation a first measure of Cross-correlation of the input signals of a generator is and additional Measure of Cross-correlation by applying a measure of the relative levels of such Input signals obtained on the first measure of cross-correlation to produce a direction-weighted measure of cross-correlation. The method of claim 49, wherein a further additional measure of cross-correlation is the one output signals of a generator by applying a scaling which is approximately equal to a value of a measure of the cross-correlation of such input signals, in the case of equal energy in each of the outputs of the variable matrix ( 203 ) is won. The method of claim 27, wherein when more than a generator for Scales for variable matrices, the scaling generators for variable Matrices are also controlled in response to information the effect of one or more other generators for scaling for variable matrices, which receive the same input signal, compensate. The method of claim 51, wherein deriving the N audio outputs from the variable matrix ( 203 ) includes compensating for multiple variable-matrix scaling generators generating scalings for the same output signal.