BRPI0305746B1 - SPACE TRANSLATION OF AUDIO CHANNEL - 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

(54) Title: SPACE TRANSLATION OF AUDIO CHANNEL (51) Int.CI .: H04S 3/02 (52) CPC: H04S 3/02 (30) Unionist Priority: 07/08/2002 US 60 / 401,983 (73) Title (s): DOLBY LABORATORIES LICENSING CORPORATION (72) Inventor (s): MARK FRANKLIN DAVIS

• · · · · · ·

THE

Invention Patent Descriptive Report for SPACE TRANSLATION OF AUDIO CHANNEL.

TECHNICAL FIELD í

, I r ^-

The present invention relates to audio signal processing. More particularly, the invention relates to the translation of M audio input channels representing a sound field into N audio output channels representing the same sound field, where each channel is a single audio stream representing incoming audio. of a direction, M and N are positive integers, and M is at least 2 and N is at least 3, and N is greater than M. Typically, a spatial translator in which N is greater than M is usually characterized as a decoder.

TECHNICAL FUNDAMENTALS

Although humans have only two ears, sounds are heard as a three-dimensional entity, relying on various location clues, such as head-related transfer functions (HRTFs) and head movement. Full fidelity sound reproduction requires the retention and reproduction of a full 3D sound field, or at least its perceptual clues. Unfortunately, a sound recording technology is not geared towards capturing the 3D sound field, nor capturing a 2D sound plane, nor capturing a 1D sound line. Current sound recording technology is strictly geared towards capturing, preserving and presenting discrete, zero-dimension audio channels.

Most of the effort to improve fidelity since Edison's original sound recording invention has focused on reducing the imperfections of his original analog modulated slot / cylinder media. These imperfections included limited, non-uniform temperature response, distortion, tuning, rapid variation in speed, accuracy of speed, wear, dirt, and loss of copy generation. While there have been a number of gradual attempts at isolated improvements, including electronic amplification, tape recording, noise reduction, and performance devices that cost more than a few cars, the problems still exist.

Individual channel quality additions have been proven, but not finally resolved until the unique development of digital recording in general and, specifically, the introduction of the Audio Compact Disc. Since then, apart from some effort to further extend the quality of digital recording to 24-bit / 96 kHz sampling, primary efforts in audio playback research have focused on reducing the amount of data needed to maintain individual channel quality, mainly using perceptual encoders, and in increasing spatial fidelity. The last problem is the subject of this document.

Efforts to improve spatial fidelity continued on two fronts: trying to carry the perceptual cues of a full sound field and trying to carry one approximately from the original real sound field. Examples of systems that employ the first approach include binaural recording systems and virtual surround based on two speakers. These systems have a number of unfortunate imperfections, especially in reliably locating sounds in some directions and requiring headphones or a single fixed listener position.

For the presentation of a spatial sound to multiple listeners, whether in a living room or in a commercial environment, such as a cinema, the only viable alternative has been trying to get closer to the real original sound field. Given the nature of a discrete sound recording channel, it is not surprising that most efforts so far have involved what could be termed conservative increases in the number of presentation channels. Representative systems include film soundtracks from three mono balanced speakers from the early 1950s, conventional stereo sound, quadrophonic systems from the 1960s, discrete five-channel magnetic soundtracks in 70 mm films, Dolby surround using a matrix in the 70's, AC-3 5.1 channel sound from the 90's and, recently, the Surround-EX 6.1 channel sound. Dolby, Pro Logic and Surround EX are trademarks of Dolby Laboratories Licensing Corporation. To some degree or another, these systems provide reproduction

improved spatial performance compared to a monophonic presentation. However, mixing a larger number of channels incurs more time and cost penalties for content producers, and the resulting perception is typically one of a few discrete channels spread out, rather than a continuous sound field. The aspects of Dolby Pro Logic decoding are described in U.S. Patent No. 4,799,260, the patent of which is incorporated by reference here in its entirety. The details of AC-3 are set out in Digital Audio Compression Standard (AC-3), Advanced Television Systems Committee (ATSC), Document A / 52, December 20, 1995 (available on the World Wide Web at www.atsc. org / Standards / A52 / a_52.document). See also the July 22, 1999 Folha de Errata (available on the World Wide Web at www.dolby.com/tech/ATSC_err.pdf.

Once the sound field is characterized, it is possible, in principle, for a decoder to derive the optimum signal feed for any output speaker. The channels supplied for such a decoder will be referred to here as cardinal, transmitted and input channels, and any output channel with a location that does not match the position of one of the input channels will be referred to as an intermediate channel. An output channel can also have a location coincident with the position of an input channel.

DESCRIPTION OF THE INVENTION

According to a first aspect of the invention, a process for translating M audio input signals, each associated with a direction, into N audio output signals, each associated with a direction, where N is greater than Μ, M is two or more and N is a positive integer equal to three or more, comprises the provision of a matrix of M: N variables, the application of the M audio input signals to the variable matrix, the derivation of the N audio output signals from the variable matrix, and controlling the variable matrix in response to input signals so that a sound field generated by the output signals has a compact sound image in the direction of the primary direction in progress input signals, when the input signals are highly correlated, the image diffuses from compact to wide, as the correlation decreases and progressively divides into multiple compact sound images, each in a direction associated with an input signal, as the correlation continues to decrease until highly uncorrelated5.

According to this first aspect of the invention, the variable matrix can be controlled in response to measurements of: (1) the relative levels of the input signals, and (2) the cross-correlation of the input signals. In that case, for a cross correlation measure of the input signals 10 having values in a first range, limited by a maximum value and a reference value, the sound field can have a compact sound image, when the cross correlation measure is the maximum value, and can have a widely spread image when the cross correlation measure is the reference value, and for a cross correlation measure of the input signals having values in a second range, limited by the reference value and a minimum value, the sound field can have the image widely spread when the cross correlation measure is the reference value, and can have a plurality of compact sound images, each in a direction associated with an input signal, when the cross-correlation measure is the minimum value.

According to another aspect of the present invention, a process for translating M audio input signals, each associated with a direction, into N audio output signals, each associated with a direction, where N is greater than that M, and M is three or more, comprises the provision of a plurality of matrices of m: n variables, where m is a subset of M and n is a subset of N, the application of a respective subset of the M input signals of audio to each of the variable arrays, the derivation of a respective subset of the N audio output signals from each of the variable arrays, the control of each of the variable arrays in response to the subset of input signals applied to them, so that a sound field generated by the respective subset of output signals derived from it has an image of

compact sound in the direction of the primary direction in nominal travel of the subset of input signals applied to it, when such input signals are highly correlated, the image spreading from compact to wide as the correlation decreases and progressively dividing into multiple sound images compact, each in a direction associated with an input signal applied to it, as the correlation continues to decrease until highly uncorrelated, and the derivation of the N audio output signals from the subsets of N audio output channels.

According to this other aspect of the present invention, the variable arrays can also be controlled in response to information that compensates for the effect of one or more other variable arrays receiving the same input signal. Furthermore, derivation of the N audio output signals from the subsets of N audio output channels can also include compensation for multiple variable arrays producing the same output signal. According to such additional aspects of the present invention, each of the variable arrays can be controlled in response to measurements of: (a) the relative levels of the input signals applied to it, and (b) the cross correlation of the input signals .

According to yet another aspect of the present invention, a process for translating M audio input signals, each associated with a direction, into N audio output signals, each associated with a direction, where N is greater than that M, and M is three or more, comprises the provision of a matrix of variable M: N in response to scale factors that control the matrix coefficients or control the matrix outputs, the application of the M audio input signals to variable matrix, the provision of a plurality of variable matrix scale factor generators m: n, where m is a subset of M and n is a subset of N, the application of a respective subset of the M audio input signals for each of the variable matrix scale factor generators, the derivation of a set of variable matrix scale factors for the respective subsets of the N audio output signals from each of the variable scale factor generators matrix variable z, in response to the control system of

»· · · ···· I» · <»· · · (• · input signals applied to them, so that when the scaling factors generated by them are applied to the variable matrix M: N, a field of sound generated by the respective subset of output signals produced has a compact sound image in the nominal primary direction in progress of the subset of input signals that produced the applied scale factors, when such input signals are highly correlated, the image spreading from compact to wide as the correlation decreases and progressively dividing into multiple compact sound images, each in a direction associated with an input signal that produced the applied scale factors, as the correlation continues to decrease until highly uncorrelated, and the derivation of the N audio output signals from the variable matrix.

In accordance with yet another aspect of the present invention, variable matrix scale factor generators can also be controlled in response to information that compensates for the effect of one or more other variable matrix scale factor generators receiving the same input signal. Furthermore, the derivation of the N audio output signals from the variable matrix can include the compensation of multiple variable matrix scale factor generators producing scale factors for the same output signal. According to yet other aspects of the present invention, each of the variable matrix scale factor generators can be controlled in response to means of: (a) the relative levels of the input signals applied to it and (b) the correlation cross-entry signals.

In accordance with the present invention, M channels of audio input representing a sound field are translated into N channels of audio output representing the same sound field, where each channel is a unique audio stream representing audio arriving from one direction, M and N are positive integers, and M is at least 2 and N is at least 3, and N is greater than M. Each input and output channel has an associated direction (for example, azimuth, elevation and, optionally, distance, to enable a virtual or projected channel closer or further away). a

or more sets of output channels are generated, each channel having one or more output channels. Each set is usually associated with two or more spatially adjacent input channels, and each output channel in a set is generated by determining a measure of the cross correlation of the two or more input channels and a measure of the interrelationships level of the two or more input channels. The cross correlation measure, preferably, is a measure of the zero time displacement cross correlation, which is the ratio of the common energy level with respect to the geometric mean of the input signal energy levels. The common energy level is preferably the common attenuated or weighted energy level, and the input signal energy levels are the attenuated or weighted input signal energy levels.

In one aspect of the present invention, multiple sets of output channels can be associated with more than two input channels, and a process can determine the correlation of input channels, to which each set of output channels is associated, according to a hierarchical order, so that each set or each of the output windows is classified according to the number of input channels with which their channel or their output channels is (are) associated, the largest number of input channels having the highest rating, and processing processes sets in order according to their hierarchical order. In addition, according to one aspect of the present invention, processing takes into account the results of processing higher order sets.

The execution or decoding aspects of the present invention assume that each of the M audio input channels representing an audio arriving from one direction was generated by a balanced amplitude encoding of the nearest neighbor to the passive matrix of each direction of source (ie, a source direction is assumed to primarily map the nearest input channel (s)), without the need for additional side chain information (the use of side or auxiliary chain information is optional), making it compatible with existing mixing techniques, consoles and formats. Although such signs

·· ···· · »····· · ··· sources can be generated by the explicit use of a passive encoding matrix, the more conventional recording techniques inherently generate such source signals (thus constituting a matrix of effective coding). The execution or decoding aspects of the present invention are also largely compatible with the natural recording source signals, as could be done with five real directional microphones, since, allowing for some possible time delay, sounds arriving from intermediate directions tend to be mapped mainly to the nearest microphones (in a horizontal arrangement, specifically to the closest pair of microphones).

A decoder or decoding process according to the aspects of the present invention can be implemented as a lattice of coupled processing modules or modular functions (from this point on, decoding modules or modules), each of which is used for generation of one or more output channels (or, alternatively, control signals that can be used to generate one or more channels), typically from two or more of the closest spatially adjacent input channels associated with the decoding module . The output channels typically represent relative proportions of the audio signals in the closest spatially adjacent input channels associated with the particular decoding module. As explained in more detail below, the decoding modules are coupled together so that the modules share inputs and there is a hierarchy of decoding modules. The modules are ordered in the hierarchy according to the number of input channels they are associated with (the module or modules with the highest number of associated input channels is (are) ranked higher). A supervisor or a supervisory function governs the modules so that common input signals are evenly shared between or among the modules, and the higher-order decoder modules can affect the output of the lower-order modules.

Each decoder module, in effect, can include a

so that it generates output signals directly, or each decoder module can generate control signals that are used, together with the control signals generated by other decoder modules, to vary the coefficients of a variable matrix or the scaling factors of inputs to or from a fixed matrix, in order to generate all output signals.

The decoder modules emulate the operation of the human ear to try to provide transparent reproduction in perceptual terms. A signal translation in accordance with the present invention, whose decoder modules and function modules are one aspect, can be applied to broadband signals or to each frequency band of a multi-band processor, and depending on the implementation, can be performed once per sample or once per sample block. A multiple band modality may employ a filter bank, such as a discrete critical band filter bank or a filter bank having a band structure compatible with an associated decoder, or a transform configuration, such as a filter bank linear FFT (Fast Fourier Transform) or MDCT (Modified Discrete Cosine Transform).

Another aspect of this invention is that the number of speakers receiving the N output channels can be reduced to a practical number by a judicious confidence in the formation of virtual image, which is the creation of sound images perceived in other positions in the space beyond where the speaker is located. Although the most common use of virtual image formation is in the stereo reproduction of an image divided between two speakers, by the balanced formation of a monophonic signal between the channels, a virtual image formation, as contemplated as an aspect of the present invention, it may include rendering ghostly projected images that provide the auditory impression of being beyond the walls of a room or within the walls of a room. The formation of virtual image is not considered to be a viable technique for group presentation with a sparse number of channels,

»·······« · • · · · • ··· · · that it requires the listener to be equidistant from the two speakers, or almost so. In cinemas, for example, the front left and right speakers are too far away to obtain a useful phantom virtual image formation from a central image for most audiences; so, given the importance of the central channel as the source of most of the dialogue, a physical center speaker is used instead.

As the density of the loudspeakers is increased, a point will be reached where virtual image formation is feasible between any pair of loudspeakers for most of the audience, at least until the tension in which the balanced ones are attenuated; with enough speakers, the spaces between the speakers are no longer perceived that way.

Signal Distribution

As mentioned above, a cross-correlation measure determines the ratio of dominant energy (common signal components) to non-dominant (unusual signal components) in a module and the degree of diffusion of non-dominant signal components across the output channels. of the module. This can be better understood by considering the signal distribution of the output channels of a module under different signal conditions, in the case of a two-input module. Unless stated otherwise, the established principles extend directly to higher-order modules.

The problem with signal distribution is that there is often very little information for retrieving the original signal amplitude distribution, let alone the signals themselves. The basic information available is the signal levels at each module input and the weighted cross product of the input signals, the common energy level. The cross correlation of zero time displacement is the ratio of the common energy level with respect to the geometric mean of the input signal energy levels.

The significance of cross-correlation is that it functions as a measure of the net amplitude of signal components common to all

• ·· ······· «· · · · ···· · • · · · · · · ♦ · · ·· ·· · the inputs. If there is a single panorama signal anywhere between the module's signals (an internal or intermediate signal), all inputs will have the same waveform, although possibly with different amplitudes, and under these conditions, the correlation will be 1.0 . At the other extreme, if all input signals are independent, meaning that there is no common signal component, the correlation will be zero. Intermediate correlation values between 0 and 1.0 can be considered to correspond to intermediate equilibrium levels of some single common signal component and independent signal components at the inputs.

Consequently, any input signal condition can be divided into a common signal, the dominant signal, and into input signal components left after subtracting common signal contributions, comprising all the rest of the signal component (the non-dominant energy or signal residue). As mentioned above, the common or dominant signal amplitude is not necessarily higher than the residual or non-dominant signal levels.

For example, consider the case of a five channel arc (L (Left), MidL (Middle Left), C (Central), MidR (Middle Right), R (Right)) mapped to a single Lt / Rt pair (total right and left total) in which it is desired to recover the original five channels. If all five channels have independent signals of equal amplitude, then Lt and Rt will be of equal amplitude, with a common intermediate energy value corresponding to an intermediate cross-correlation value between zero and one (because Lt and Rt are not independent signals ). The same levels can be obtained with appropriately chosen levels of L, C and R, without signs of MidL and MidR. Thus, a module with two inputs and five outputs could supply only the output channel corresponding to the dominant direction (C, in this case), and the output channels corresponding to the input signal residues (L, R), after removing the C power of Lt and Rt inputs, giving no signals to the MidL and MidR output channels. Such a result is undesirable - deactivating a channel unnecessarily is almost always a bad choice because small disturbances in signal conditions will cause the switched off channel to switch between on and off, causing an irritating vibration (vibration is a channel quickly on and off), especially when the disconnected channel is heard isolated.

Consequently, when there are multiple possible output signal distributions for a given set of module input signal values, the conservative approach from an individual channel quality point of view is to spread the non-dominant signal components as evenly as possible among the module output channels, consistent with signal conditions. One aspect of the present invention is a uniform diffusion of the available signal energy, subject to signal conditions, according to a three-way division, rather than a two-way division of dominant versus everything else. Preferably, the three-way split comprises dominant (common) signal components, fill signal components (uniform diffusion) and residue from input signal components. Unfortunately, there is only enough information to make a two-way split (dominant signal components and all other signal components). A suitable approach for performing a three-way split is described here, in which for correlation values above a particular value, the two-way split employs the dominant and non-dominant diffusion signal components; for correlation values below that value, the two-way division employs the non-dominant diffusion signal components and the residue. The common signal energy is divided between dominant and uniformly diffused. The uniform broadcast component includes both common and waste signal components. Therefore, a diffusion involves a mixture of common (correlated) and residue (uncorrelated) signal components.

Before processing, for a given input / output channel configuration for a given module, a correlation value is calculated corresponding to all output channels receiving the same signal amplitude. This correlation value can be referred to as the random_xcor value. For a single intermediate output channel derived from the center and two input channels, the random_xcor value can be calculated as 0.333.

For three equally spaced intermediate channels and two input channels, the random_xcor value can be calculated as 0.483. Although these time values have been shown to provide satisfactory results, they are not critical. For example, values of about 0.3 and 0.5, respectively, can be used. In other words, for a module with M inputs and N outputs, there is a particular degree of correlation of the M inputs that can be considered to represent equal energies in all N outputs. This can be achieved by considering the M inputs as if they had been derived using a passive matrix N by M receiving

No independent signals of equal energy, although, of course, the actual inputs can be derived by other means. This threshold correlation value is random_xcor, and can represent a dividing line between two operating regimes.

Then, during processing, if a module's raw correlation value is greater than or equal to the random_xcor value, it is scaled to a range of 1.0 to 0:

scaled_xcor = (correlation - random_xcor) / (1 - random_xcor)

The scaled_xcor value represents the amount of dominant signal above the uniform broadcast level. Anything left can be distributed equally to the other output channels of the module.

However, there is an additional factor that must be considered, specifically, that, as the primary direction in nominal progress of the input signals becomes progressively more off-center, the amount of diffusion energy must be progressively reduced if an equal distribution for all output channels it is maintained or, alternatively, the amount of diffusion energy must be maintained, but the energy distributed to the output channels must be reduced in relation to the center output of the dominant energy - in other words, a slope of energy along the output channels. In the latter case, additional processing complexity may be required to keep the output power equal to the input power.

If, on the other hand, the current correlation value is less than

the random_xcor value, the dominant energy is considered to be zero, the uniformly diffused energy is progressively reduced and the residual signal, whatever is left, is allowed to accumulate at the inputs. In correlation = zero, there is no internal signal, only the independent input signals that are mapped directly to the output channels.

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

a) when the actual correlation is greater than random_xcor, there is enough common energy to consider that there is a dominant signal to be directed (balanced) between two adjacent exits (or, obviously, fed to one exit, if its direction is useful coincide with that of an exit); the energy attributed to it is subtracted from the inputs to give waste which is distributed (preferably evenly) among all the outputs.

b) when the real correlation is precisely, random_xcor, the input energy (which could be considered as all waste) is evenly distributed among all the outputs (this is the definition of random_xcor).

c) when the real correlation is less than random_xcor, there is not enough common energy for a dominant signal, so that the energy of the inputs is distributed between the outputs with proportions dependent on the smaller one. This is as if someone else treated the correlated part as waste, to be uniformly distributed among all the outputs, and the non-correlated part as several dominant signals to be sent corresponding to the directions of the inputs. At the extreme of the correlation being zero, each input is fed to an output position only (usually one of the outputs, but a balance position between two of them could be possible).

So, there is a continuum between full correlation, with a sign

single balance between two outputs, according to the relative energies of the inputs, through random_xcor with the inputs evenly distributed among all outputs, until a zero correlation with M inputs fed independently for M output positions.

Interaction Compensation

As mentioned above, a channel translation in accordance with an aspect of the present invention can be considered to involve a lattice of modules. Due to the fact that multiple modules can share a given input channel, interactions are possible between modules and can degrade performance, unless some compensation is applied. Although it is generally not possible to separate signals at an input according to which module they go to, an estimate of the amount of an input signal used by each connected module can improve the resulting correlation and direction estimates, resulting in improved performance.

As mentioned above, there are two types of demo interactions:

those that involve modules at a common or lower hierarchy level (that is, modules with an equal number of entries or fewer entries), referred to as neighbors, and modules at a higher hierarchy level (having more entries) than one given module, but sharing one or more common inputs, referred to as higher-order neighbors.

Consider, first, a neighbor offset at a common hierarchy level. To understand the problems caused by a neighbor interaction, consider an isolated two input module with identical L / R input signals (left and right). This corresponds to a single (common) dominant signal halfway between the inputs. The common energy is A ² and the correlation is 1.0. Assume a second module of two inputs with a common signal, B, at its L / R inputs, a common energy B ² , and also a correlation of 1.0. If the two modules are connected to a common input, the signal at that input will be A + B. Assuming that signals A and B are independent, then the product

weighted AB will be zero, so the common energy of the first module will be A (A + B) = A ² + AB = A ² , and the common energy of the second module will be B (A + B) = B ² + AB = B ² . Thus, common energy is not affected by neighboring modules, as long as they process independent signals. This is usually a valid hypothesis. If the signals are not independent, are the same, or at least substantially share common signal components, the system will react in a manner consistent with the response of the human ear - specifically, the common input will be larger, making the audio image be attracted to the common entrance. In this case, the L / R input amplitude ratios of each module are shifted, because the common input has more of a signal amplitude (A + B) than an external input, which makes the direction estimate to be oriented towards the common entrance. In this case, the correlation value of both modules is now something less than 1.0, because the waveforms of both input pairs are different. Because the correlation value determines the degree of diffusion of the non-common signal components and the ratio of the dominant (common signal component) to non-dominant (non-common signal component) energy, an uncompensated common input signal does with which the non-common signal distribution of each module is broadcast.

To compensate, a measure of the common input level attributable to each input in each module is estimated and then each module is reported with reference to the total amount of that common input level energy from all neighboring levels at the same hierarchy level in each module entry. Two ways of calculating the common input level measure attributable to each input of a module are described here: one which is based on the common energy of the inputs to the module (generally described in the next paragraph), and another, which is more accurate, but requires greater computational resources, which is based on the total energy of the module's interior outputs (described below in relation to the arrangement in figure 6A).

According to the first method of calculating the level measure

of common input attributable to each input of a module, the analysis of input signals from a module does not allow to directly solve the common input level at each input, only a proportion of the total common energy, which is the geometric average of the levels of common input power.

Due to the fact that the common input energy level at each input cannot exceed the total energy level at that input, which is measured and known, the total common energy is factored into common input levels proportional to the observed input levels, subject to the qualification below. Once the set of common input levels is calculated for all modules in the lattice (regardless of whether the measurement of common input levels is based on the first or second calculation method), each module is informed of the total input levels common to all neighboring modules at each input, an amount referred to as a module's neighbor level at each of its inputs. The module then subtracts the neighbor level from the input level at each of its inputs, for the derivation of compensated input levels, which are used for the calculation of the correlation and direction (primary direction in nominal progress of the signals input).

For the example cited above, the neighbor levels are initially zero, because the common input has more signal than a final input, the first module claims a common power level on that input in addition to A ² and the second module claims a level common input in the same input in addition to B ² . Since the total claims are more than the level of energy available there, the claims are limited to around A ² and B ² , respectively. Due to the fact that there are no other modules connected to the common input, each common input level corresponds to the neighbor level of the other module. Consequently, the compensated input power level seen for the first module is:

(A ² + B ² ) - B ² = A ² and the input power level compensated by the second module is:

(A ² + B ² ) - A ² = B ² .

However, these are just the levels that would have been observed

with isolated modules. Consequently, the resulting correlation values will be 1.0, and the dominant directions will be centered, at the appropriate amplitudes, as desired. Nevertheless, the recovered signals themselves will not be completely isolated - the output of the first module will have some component of signal B and vice versa, but this is a limitation of a matrix system, and if the processing is carried out on a multiple band, the mixed signal components will be on a similar frequency, making the distinction between them somewhat debatable. In more complex situations, compensation will usually not be as accurate, but experience with the system indicates that compensation in practice mitigates most of the neighboring module's interaction effects.

Having established the principles and signals used in neighboring level compensation, an extension to higher order neighbor level compensation is reasonably straightforward. This applies to situations in which two or more modules at different levels of hierarchy share more than one common 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 will also be common to both inputs of the two-input module and, without compensation, will be rendered in different positions by each module. More generally, there can be a signal component common to all three inputs and a second component common to only two input module inputs, requiring their effects to be separated as far as possible, for an appropriate rendering of the output sound field. . Consequently, the common signal effects of three inputs, as performed at the common input levels described above, must be subtracted from the inputs, before the calculation of two inputs can be performed properly. In fact, common signal elements of the highest order must be subtracted not only from the lowest level module input levels, but from their observed measure of common energy level as well, before proceeding with the higher level calculation. low. This is different from the effects of common module entry levels

at the same hierarchy level, which do not affect the common energy level measurement of a neighboring module. Thus, neighboring levels of the highest order should be considered and used separately from neighboring levels of the same order. While the higher or5 dem neighbor levels are passed down to the lowest modules in the hierarchy, the remaining common levels of lower level modules must also be passed up in the hierarchy, because, as mentioned above , the lower level modules act as common neighbors for the higher level modules. Some amounts are interdependent and difficult to resolve simultaneously. In order to avoid performing resource-intensive computations of complex simultaneous solutions, previous calculated values can be passed to the relevant modules. A potential interdependence of common module input levels at different levels of hierarchy can be resolved by using the previous value, as above, or by performing calculations in a repetitive sequence (ie, a loop), from the highest hierarchy level to the lowest. Alternatively, a simultaneous equation solution may also be possible, although it may involve a non-trivial computational effort.

Although the interaction compensation techniques described only send approximately correct values for complex signal distributions, they are believed to provide an improvement over a lattice arrangement that fails to take module interactions into account.

BRIEF DESCRIPTION OF THE DRAWINGS

Figure 1 is a top plan view schematically showing an ideal decoding arrangement in the manner of a test arrangement employing a sixteen-channel horizontal arrangement around the walls of a room, a six-channel arrangement arranged in a circle above the horizontal arrangement and a single air channel.

Figure 2 is a functional block diagram providing an overview of a multiple band transform modality of a plurality of modules operating with a central supervisor implementing

»· * · ♦ ·· the example in figure 1.

Figure 3 is a functional block diagram useful in understanding the way in which a supervisor, such as a supervisor 201 in figure 2, can determine an end point scale factor.

Figures 4A to 4C show a functional block diagram of a module in accordance with an aspect of the present invention.

Figure 5 is a schematic view showing a hypothetical arrangement of a three input module powered by a triangle of input channels, three internal output channels and a dominant direction. The view is useful in understanding the distribution of dominant signal components.

Figures 6A and 6B are functional block diagrams that show, respectively, a suitable arrangement for (1) the generation of the total estimated energy for each input of a module in response to the total energy of each input, and (2) in response to a cross-correlation measure of the input signals, generating an end point energy scale factor component for each of the module end points.

Figure 7 is a functional block diagram showing a preferred function of the sum block and / or greater than 367 of figure 4C.

Figure 8 is an idealized representation of the way in which an aspect of the present invention generates scale factor components in response to a cross-correlation measure.

Figures 9A and 9B to Figures 16A and 16B are series of idealized representations that illustrate the output scale factors of a module resulting from several examples of input signal conditions.

MODES FOR CARRYING OUT THE INVENTION

In order to test aspects of the present invention, an arrangement was employed, which has a horizontal arrangement of 5 speakers on each wall of a room having four walls (one speaker on each corner and three evenly spaced between each corner) ), 16 speakers in total, allowing for common corner speakers, plus a 6-speaker ring above a listener located centrally at a vertical angle

about 45 degrees, plus a single speaker directly above, a total of 23 speakers, plus a subwoofer / LFE channel (low frequency effects), a total of 24 speakers, all powered from a personal computer configured to run on 24 channels. Although in current jargon this system can be referred to as a 23.1 channel system, for simplicity it will be referred to here as a 24 channel system.

Figure 1 is a top plan view that schematically shows a decoding arrangement idealized in the manner of the test arrangement just described. Five horizontal broadband input channels are shown as squares 1 ', 3', 5 ', 9' and 13 'in the outer circle. A vertical channel, which can be derived from the five broadband inputs through generated correlation or reverberation, or separately supplied (as in figure 2), is shown as the dotted square 23 'in the center.

The twenty-three broadband output channels are shown as solid circles numbered 1 through 23. The outer circle of sixteen output channels is in a horizontal plane, the inner circle of six output channels is forty-five degrees above the horizontal plane. Output channel 23 is directly above one or more listeners. Five two-input decoding modules are outlined by keys 24 to 28 around the outer circle, connected between each pair of horizontal input channels. Five vertical decoding modules with two additional inputs are outlined by switches 29 to 33 connecting the vertical channel to each of the horizontal inputs. An output channel 21, the raised central rear channel, is derived from a three input decoding module 34 illustrated as arrows between output channel 21 and input channels 9, 13 and 23. Thus, the output module three inputs 34 is a higher level in the hierarchy than its neighboring modules of lower hierarchy of two inputs 27, 32 and 33. In this example, each module is associated with a respective pair or trio of spatially adjacent input channels closest together. Every module in this example has at least three neighbors at the same level. For example, modules 25, 28 and 29 are neighbors of module 24.

• · · · · ♦ · »w · · * · · · · · · · · · · · · · · * · ·«

Although the decoding modules represented in figure 1 have, in varying degrees, three, four or five output channels, a decoding module can have any reasonable number of output channels. An output channel can be located intermediate to two or more input channels or in the same position as an input channel. Thus, in the example in figure 1, each of the input channel locations is also an output channel. Two or three decoding modules share each input channel.

Although the arrangement in figure 1 employs five modules (24 to 28) (each having two inputs) and five inputs (Γ, 3 ', 5', 9 'and 13') for the derivation of sixteen horizontal outputs (1 to 16 ) representing locations around four walls of a room, similar results can be obtained with a minimum of three entries and three modules (each having two entries, each module sharing an entry with another module).

By using multiple modules in which each module has output channels in an arc or line (such as the example in figures 1 and 2), the decoding ambiguities found in the decoders of the prior art, in which correlations less than zero are decoded as indicating backward directions, can be avoided.

Although input and output channels can be characterized by their physical position or at least by their direction, their characterization with a matrix is useful, because it provides a well-defined signal relationship. Each matrix element (row i, column j) is a transfer function relating an input channel i to an output channel j. The matrix elements are usually signed multiplicative coefficients, but they can also include phase or delay terms (in principle, any filter), and they can be frequency functions (in terms of discrete frequency, a different matrix at each frequency) . This is straightforward in the case of dynamic scale factors applied to the outputs of a fixed matrix, but this lends itself to a variable matrix formation, by having a separate scale factor for each matrix element, or for more elaborate matrix elements than simple scalar scaling factors, in which the matrix elements themselves are variable, for example, a variable delay.

There is some flexibility in mapping physical positions to matrix elements; at first, the aspect modalities of the present invention can deal with mapping an input channel to any number of output channels and vice versa, but the most common situation is to assume signals mapped to the nearest output channels via simple scalar factors, which, for power preservation, add up to square 1.0. Such mapping is often done using a sine / cosine panorama function.

For example, with two input channels and three internal output channels in a line between them plus the two endpoint output channels coinciding with the input positions (ie, an M: N module, where M is 2 and N is 5), it can be assumed that the span represents 90 degrees of arc (the range in which the sine or cosine changes from 0 to 1 or vice versa), so that each channel is spaced 90 degrees / 4 intervals = 22.5 degrees, giving the channel matrix coefficients (cos (angle), sen (angle)):

Lef Coefs = cos (0), without (0) = (1.0)

MidLout coefs = cos (22.5), without (22.5) = (0.92, 0.38)

Coefs Cout = cos (45), without (45) = 0.71.0.71)

MidRout coefs = cos (67.5) without (67.5) = (0.38, 0.92)

Rout coefs = cos (90), without (90) = (0.1)

Thus, for a matrix with fixed coefficients and a variable gain controlled by a scale factor at each matrix output, the output signal at each of the five output channels is (where SF is a scale factor for particular exit identified by the subscriber):

Lout = Lt (SF _L )

MidLout + ((0.92) Lt + (0.38) Rt)) (SF _Mid L)

Cout = ((0.45) Lt + (0.45) Rt)) (SF _C )

MidRout = ((0.38) Lt + (0.92) Lt)) (SFmícir)

Rout = Rt (SFr)

Generally, given an arrangement of input channels, one can, conceptually, join the nearest entrances by straight lines, representing

··· ·· ···· ·· ··· • · · · · · · · · ·· ···· · · * · · · • · · ·· · »··· ·· · · ·· ·· • · ··· • · · · · · · · · · · ··ring potential decoder modules. (They are potential because if there is no output channel that needs to be derived from a module, the module is not needed.) For typical arrangements, any output channel on a line between two input channels can be derived from one two input module (if sources and transmission channels are in a common plane, then any other source appears in a maximum of 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, perhaps more than one module. An output channel not in a row or in the same position as an input (for example, inside or outside a triangle formed by three input channels) requires a module having 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 occur, for example, when the source channels and the input channels are not on a plane: a source channel can map more than two input channels. This occurs, for example, in figure 1, when mapping 24 channels (16 horizontal ring channels, 6 high ring channels, 1 vertical channel, plus LFE) to 6.1 channels (including a vertical channel with 20 pockets). In this case, the central rear channel in the raised rings is not in a direct line between two of the source channels, it is in the middle of a triangle formed by the channels Ls (13), Rs (9) and the top (23), of so that a three input module is required to extract it. One way to map elevated channels for a horizontal arrangement is to map each channel to more than two input channels. This allows the 24 channels in the example in Figure 1 to be mapped to a conventional 5.1 channel arrangement. In that alternative, a plurality of three input modules can extract the elevated channels, and the signal components left can be processed by two input modules for the extraction of the main horizontal ring ho30.

In general, it is not necessary to check all possible combinations of signal community among the input channels. With arrangements

flat channels (for example, channels representing horizontally arranged directions), it is usually appropriate to carry out a similarity comparison for spatially adjacent channels. For channels arranged on the roof or on the surface of a sphere, the signal community can extend to three or more channels. The use and detection of the signal community can also be used to carry additional signal information. For example, a vertical signal component can be represented by mapping to all five full-range channels in a horizontal five-channel arrangement.

Decisions as to which input channel combinations to analyze for the community, along with a standard input / output mapping matrix, need to be made only once per translator or input / output channel translator role arrangement. The initial mapping (before processing) derives from a passive master matrix that relates the input / output channel settings to the spatial orientation of the channels. As an alternative, the processor or processing portion of the invention can generate time-varying scale factors, one per output channel, which modify the output signal levels of what would otherwise have been a simple passive matrix or the coefficients of the matrix itself. The scale factors, in turn, derive from a combination of signal components (a) dominant, (b) uniform spread, and (c) residue (full stop), as described below.

A master matrix is useful in configuring a module array, as shown in the example in figure 1 and further described below in relation to figure 2. By examining the master matrix, one can deduce, for example, how many decoder modules are required, how they are connected, how many input and output channels each has and the matrix coefficients relating each of the module inputs and outputs. These coefficients can be taken from the master matrix; only non-null values are required, unless an input channel is also an output channel (that is, a period).

Each module, preferably, has a local matrix, which

is that portion of the master matrix applicable to the particular module. In the case of a multiple module arrangement, such as, for example, figures 1 and 2, the module can use the local matrix in order to produce scale factors (or matrix coefficients) for the control of the master matrix, such as is described below in relation to figures 2 and 4A to 4C, or for the purpose of producing a subset of the output signals, the output signals of which are assembled by a central process, such as a supervisor, as described in relation to figure 2 A supervisor like this, in the latter case, compensates for multiple versions of the same output signal produced by modules having a common output signal in a manner analogous to the way in which supervisor 201 in figure 2 determines a final scale factor to replace the preliminary scale factors produced by the modules that produce preliminary scale factors for the same output channel.

In the case of multiple modes that produce scaling factors instead of output signals, modules can continuously obtain matrix information relevant to themselves from a master matrix through a supervisor, rather than having a local matrix . However, less computational effort is required if the module has its own local matrix. In the case of a single independent module, the module has a local matrix, which is the only required matrix (in effect, the local matrix is the master matrix), and that local matrix is used for the production of output signals.

Unless otherwise indicated, descriptions of modalities of the invention having multiple modules are with reference to the alternative in which the modules produce scale factors.

Any decoding module output channel with only a non-zero coefficient in the local module matrix (that coefficient is 1.0, since the coefficients have the quadratic sum calculated up to 1.0) is an end point channel. Output channels with more than one non-zero coefficient are internal output channels. Consider a simple example. If output channels O1 and 02 are both derived from input channels 11 and I2 (but with different coefficient values), then a module is needed

two inputs connected between 11 and 12 generating outputs O1 and 02, possibly among others. In a more complex case, there are five inputs and 16 outputs, and one of the decoder modules has inputs 11 and I2 and feeds outputs 01 and 02, so that:

01 = AI1 + BI2 + 0I3 + 0I4 + 0I5 (note no contribution from input channels I3, I4 or I5), e = C 11 + D I2 + Ο I3 + Ο I4 + Ο I5 (note no input channel inputs I3,14 or I5), then, the decoder can have two inputs (11 and I2), two outputs, and the related scale factors are:

O1 = A 11 + B I2e 02 = C 11 + D I2.

The master matrix or the local matrix, in the case of a single independent module, can have matrix elements that work to provide more than multiplication. For example, as mentioned above, matrix elements can include a filter function, such as a phase or delay term, and / or a filter that is a frequency function. An example of filtering that can be applied is a matrix of pure delays, which can render projected ghost images. In practice, a master or local matrix like this can be divided, for example, into two functions, one that employs coefficients for the derivation of the output channels, and a second that applies a filter function.

Figure 2 is a functional block diagram that provides an overview of a multiple band transform mode that implements the example in Figure 1. A PCM audio input, for example, having multiple interlaced audio signal channels, is applied to a supervisor or supervisory function 201 (from this point on, supervisor 201), which includes a deinterleaver that retrieves separate streams from each of the six channels of audio signals (1 ', 3', 5 ', 9 ', 13' and 23 ') carried by the interlaced input and apply each one to a transform or transformed function from time domain to frequency domain (from this point onwards, the forward transform). Alternatively, the channels of

audio can be received in separate streams, in which case no deinterleaver is required.

As mentioned above, a signal translation according to the present invention can be applied to broadband signals or to each frequency band of a multiple band processor, which can employ a filter bank, such as a filter bank discrete critical band filter, or a filter bank having a band structure compatible with an associated decoder, or a transform configuration, such as a FFT (Fast Fourier Transform) or MDCT (Co-Transform) linear bank filter Modified Discrete sine). Figures 2, 4A to 4C and other figures are described in the context of a multi-band transform configuration.

An optional LFE input channel (a potential seventh input channel in figures 1 and 2) and an output channel (a potential 24 ° output channel in figures 1 and 2 are not shown in figures 1 and 2) ). The LFE channel can be treated generally in the same way as the other input and output channels, but with its own scale factor set at 1 and its own matrix coefficient, also set at 1. In cases where the source does not have LFE, but the output channels do (for example, a 2: 5.1 upmix), an LFE channel can be derived by using a low-pass filter (for example, a Butterworth filter of fifth order with a corner frequency of 120 Hz) applied to the sum of the channels, or to avoid cancellation by adding the channels, a corrected phase sum of the channels can be used.

In cases where the input has an LFE channel, but not the output, the LFE channel can be added to one or more of the output channels.

Continuing with the description in figure 2, modules 24 to 34 receive the appropriate inputs from inputs T, 3 ', 5', 9 ', 13' and 23 'as shown in figure 1. Each module generates a factor output preliminary scale (PSF) for each of the audio output channels associated with it, as shown in figure 1. Thus, for example, module 24 receives inputs Γ and 3 'and generates preliminary scale factor outputs PSF1, PSF2 and

PSF3. Alternatively, as mentioned above, each module can generate a preliminary set of audio outputs for each of the audio output channels associated with it. Each module can also communicate with a 201 supervisor, as further explained below. Information sent from supervisor 201 to various modules can include neighbor level information and higher order neighbor level information, if any. An information sent to the supervisor from each module can include a total estimated energy from inside the outputs attributable to each of the module inputs. The modules can be considered as part of a control signal generation portion of the total system of figure 2.

A supervisor, such as supervisor 201 of figure 2, can perform several different functions. A supervisor can, for example, determine whether more than one module is in use and, if not, the supervisor does not need to perform any functions related to neighbor levels. During an initialization, the supervisor can inform the or each module the number of inputs and outputs it has, the matrix coefficients related to them, and the signal sample rate. As already mentioned, it can read the interlocking PCM sample blocks and deinterlace them in separate channels. It can apply a non-limiting action in the time domain, for example, in response to additional information indicating that the source signal has limited amplitude and the degree of that limitation. If the system is operating in a multi-band mode, it can apply a frame and a filter bank (for example, FFT, MDCT, etc.) to each channel (so that multiple modules do not perform redundant transforms that substantially increase the processing effort) and pass transformed value streams to each module for processing. Each module returns a two-dimensional array of scaling factors to the supervisor: a scaling factor for all transform bins in each subband of each output channel (when in a multiple band transform configuration: otherwise, a factor scale by output channel), or alternatively, a two-dimensional array of output signals: a set of bins

complex transform for each subband of each output channel (when in a multiple band transform configuration; otherwise, one output signal per output channel). The supervisor can mitigate the scale factors and apply them to the signal path matrix (matrix

203, described below) to produce (in a multi-band transform configuration) complex output channel spectra. Alternatively, when the module produces output signals, the supervisor can derive the output channels (complex spectra of the output channel, in a multi-band transform configuration), compensating for local matrices that produce the same output signal. Then, it can perform an inverse transform plus a frame and add by superposition, in the case of MDCT, for each output channel, interlacing the output samples to form a composite multiple channel output stream (or, optionally, it you can omit interlacing to provide multiple output streams), and send it to an output file, a sound card or another final destination.

Although several functions can be performed by a supervisor, as described here, or by multiple supervisors, a person skilled in the art will appreciate that several or all of those functions can be performed in the modules themselves, rather than by a supervisor common to all or to some of the modules. For example, if there is only a single independent module, there is no need to distinguish between module functions and supervisor functions. Although, in the case of multiple modes, a common supervisor can reduce the overall processing power required by eliminating or reducing redundant processing tasks, the elimination of a common supervisor or its simplification can allow modules to be easily added to each other, for example example, to update to more output channels.

Returning to the description of figure 2, the six entries Γ, 3 ', 5', 9 ’,

13 'and 23' are also applied to a variable matrix or to a variable matrix function 203 (from this point on, matrix 203). The 203 matrix can be considered as part of the signal path of the system of the figure ···· · ·· ··· * / ϊ / ~ \ ® · _φ · · · ·· ····· · ί ι / \ . .. · ..: · .. · μ)

2. Matrix 203 also receives as inputs from supervisor 201 a set of final scale factors SF1 to SF23 for each of the 23 output channels in the example in figure 1. The final scale factors can be considered to be the output of the portion of the system control signal in figure 2. As further explained below, supervisor 201 preferably passes, as final scale factors to the matrix, the preliminary scale factors for each interior output channel, but the supervisor determines final scales for each endpoint output channel, in response to information it receives from the modules. An inner output channel is intermediate between the two or more endpoint output channels for each module. Alternatively, if the modules produce output signals instead of scaling factors, no 203 matrix is required; the supervisor itself produces the output signals.

In the example in Figure 1, it is assumed that the endpoint output channels coincide with the input channel locations, although they are not required to match, as is discussed elsewhere here. Thus, output channels 2, 4, 6 to 8, 10 to 12, 14 to 16, 17, 18, 19, 20, 21 and 22 are interior output channels. The inner outlet channel 21 is intermediate or embraced by three input channels (the input channels 9 ', 13' and 23 '), while the other inner channels are each intermediate (between or embraced by) two input channels. Because there are multiple preliminary scaling factors for those endpoint output channels that are shared between or between modules (that is, output channels 1,3, 5, 9, 13 and 23), supervisor 201 determines the end point scale factors (SF1, SF3, etc.) among the scale factors SF1 to SF23. The final indoor output scale factors (SF2, SF4, SF6, etc.) are the same as the preliminary scale factors.

Figure 3 is a functional block diagram useful in understanding the way in which a supervisor, such as the supervisor 201 in figure

2, can determine an end point scale factor. The supervisor does not add all the outputs of the modules sharing an input to obtain an end point scale factor. Instead, it combines

additive, as in a 301 combiner, the total estimated interior energy for an input from each module that shares the input, such as input 9 ', which is shared by modules 26 and 27 in figure 2. This sum represents the total energy level at the input claimed by the interior outputs of all connected modules. It then subtracts that sum of the attenuated input energy level at that input (for example, the attenuator output 325 or 327 in figure 4B, as described below) from any of the modules that share the input (module 26 or module 27 , in this example), as in a 303 combiner. It is sufficient to choose any of the attenuated module inputs at the common input, although the levels may differ slightly from module to module, because each module adjusts its time constants independently of each other. others. The difference, at the output of combiner 303, is the desired output signal energy level at that input, the energy level of which is not allowed to go beyond zero. By dividing that desired output signal level by the attenuated input level at that input, as in a 305 divider, and performing a square root operation, as in block 307, the final scale factor (SF9, in this example) for that exit is obtained. Note that the supervisor derives a single final scale factor for each of these shared inputs regardless of how many modules share the input. An arrangement for determining the total estimated energy of the interior outputs attributable to each of the module inputs is described below in connection with figure 6A.

Due to the fact that the levels are energy levels (a second order quantity), as opposed to amplitudes (a first order quantity), after the division operation, a square root operation is applied, in order to obtain the final scale factor (scale factors are associated with first order quantities). The addition of the interior levels and the subtraction of the total input level are performed in a sen30 of pure energy, due to the fact that interior outputs of different module are assumed to be independent (not correlated). If this hypothesis is not true in an unusual situation, the calculation can

produce more signal left at the entrance than there should be, which can cause a slight spatial distortion in the reproduced sound field (for example, a slight pull of other nearby interior images towards the entrance), but in the same situation, the ear human probably reacts in a similar way. The internal output channel scale factors, such as PSF6 to PSF8 of module 26, are passed by the supervisor as final scale factors (they are not modified). For simplicity, Figure 3 shows only the generation of one of the final end point scale factors. Other end-point scale factors can be derived in a similar way.

Returning to the description of figure 2, as mentioned above, in the variable matrix 203, the variability can be complicated (all variable coefficients) or simple (coefficients varying in groups, as applied to the inputs or outputs of a fixed matrix) . Although any approach can be employed to produce substantially the same results, one of the simplest approaches, that is, a fixed matrix followed by a variable gain for each output (the gain of each output controlled by scale factors) has been shown to produce results satisfactory and is used in the modalities described here. Although a variable matrix in which each matrix coefficient is variable can be used, it has the disadvantage of having more variables and requiring more processing power.

Supervisor 201 also performs a time domain attenuation of the final scale factors, before they are applied to the variable matrix 203. In a variable matrix system, the output channels are never turned off, the coefficients are arranged to reinforce some signals and cancel others. However, a fixed matrix variable gain system, as in the described embodiments of the present invention, however, turns channels on and off, and is more susceptible to desirable vibration artifacts. This can occur, despite the two-stage attenuation described below (see attenuators 219/325, etc.). For example, when a scale factor is close to zero, because only a small change • · · * · is needed to go from 'small' to 'none' and back, transitions to and from zero can cause an audible vibration.

The optional attenuation performed by the supervisor 201, preferably, attenuates the output scaling factors with variable time constants that depend on the size of the absolute difference (abs-diff) between values of newly derived instantaneous scaling factors and a running value of the factor attenuated scale. For example, if the abs-diff is greater than 0.4 (and, obviously, less than or equal to 1.0), there is little or no attenuation applied; a small additional amount of smoothing is applied to abs-diff values between 0.2 and 0.4; and below values of 0.2, the time constant is a continuous inverse function of abs-diff. Although these values are not critical, they have been shown to reduce audible vibration artifacts. Optionally, in a multi-band version of a module, the more attenuated time constants of the scaling factor can also spread with frequency as well as time, in the manner of frequency attenuators 413, 415 and 417 of figure 4A, described below.

As stated above, the variable matrix 203 is preferably a fixed decoding matrix with variable scale factors (gains) in the matrix outputs. Each matrix output channel can have matrix coefficients (fixed) that would have been the coding downmix coefficients if that channel had an encoder with discrete inputs (instead of a mix of source channels directly into the downmix arrangement, which avoids the need for a discrete encoder). The matrix coefficients are fixed once it is known where the output channels are (as discussed above with respect to the master matrix); whereas the scale factors, which control the output gain of each channel, are dynamic.

The inputs comprising the frequency domain transform bins applied to modules 24 to 34 of figure 2 can be grouped into frequency sub-bands for each module, after the initial amounts of energy and common energy are calculated at the bin level, as is further explained below. Thus, there is a scale factor

• · ··· · ♦ · ♦ * · ·· ········· ·· | J \ • · * ♦ · · · · _{ee φ} · · Λ • · · · · ♦ ···· · · · ··· V / ·· · · · ··· · · ····· · v injunction (PSF in figure 2) and a final scale factor (SF in figure 2) for each frequency sub-band. The frequency domain output channels 1 to 23 produced by matrix 203 each comprise a set of transform bins (groups of sized sub-bands of transform bins are treated by the same scale factor). The sets of frequency domain transform bins are converted to a set of PCM output channels 1 to 23, respectively, by a transform or frequency domain transform function for time 205 (from this point, an inverse transform). ), which may be a function of supervisor 201, but is shown separately for clarity. Supervisor 201 can interlace the resulting PCM channels 1 through 23 to provide a single interlaced PCM output stream or leave the PCM output channels as separate streams.

Figures 4A to 4C show a functional block diagram of a module in accordance with an aspect of the present invention. The module receives two or more input signal streams from a supervisor, such as supervisor 201 of figure 2. Each input comprises a set of complex value frequency domain transform bins. Each input, 1 am, is applied to a function or device (such as a function or device 401 for input 1 and a function or device 403 for input m), which calculates the energy of each bin, the what is the sum of the squares of the real and imaginary values of each transform bin (only the paths for two inputs, 1 in, are shown to simplify the drawing). Each of the inputs is also applied to a function or 405 device, which calculates the common energy of each bin through the module input channels. In the case of an FFT modality, this can be calculated by taking the vector 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 bin value of complex L and the complex conjugate of the bin value of complex R). Modes using real values need only a vector multiplication of the real value for each entry. For more than two entries, the special vector multiplication technique described below can

be used, specifically, if all signs are the same, the product is given a positive sign, otherwise it is given a negative sign and scaled by the ratio of the number of possible positive results (always two, they are all positive or all negative) by the number of possible negative results.

Calculation for Common Energy Pair

For example, suppose that a pair of A / B input channels contains a common signal X together with individual uncorrelated signals Y and Z:

A = 0.707Χ + Y

Β = 0.707Χ + Z where the 0.707 = -JÕp> scaling factors provide a power preservation mapping for the nearest input channels.

Energy RMS (A) = ÍA ² 6t = 7 ² = (0.707X + Y) ² (0.5Χ ² + 0.707ΧΥ + Υ ² ) = 0.5)? + 0.707ΧΫ + Ϋ ^2-

Because X and Y are unrelated,

XY = 0

So:

= 0.5 ^ + 7 ² that is, due to the fact that X and Y are uncorrelated, the total energy in input channel A is the sum of the energies of signals X and Y.

Similarly, b ² = 0.5X ² + Z ²

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

ÃB = 0.5)?

Thus, in the case of an output signal shared equally by two neighboring input channels that can also contain independent uncorrelated signals, the average cross product of the signals is equal to the energy of the common signal component in each channel. If the common signal is not shared equally, that is, it has a panorama towards

one of the inputs, the average vector product will be the geometric mean between the energy of the common components in A and B, from which estimates of common energy of the individual channel can be derived by normalizing the square root of the ratio of the channel amplitudes. The real time averages are computed in the subsequent mitigation stages, as described below.

Higher Order Calculation of Common Energy

In order to derive common energy from decoding modules with three or more inputs, it is necessary to form vector products weighted from all input signals. A simple process of processing pair with the inputs fails to differentiate separate output signals between each pair of inputs and a signal common to all.

Consider, for example, three input channels, A, B and C, consisting of uncorrelated signals W, Y, Z and a common signal X:

A = X + W

B = X + Y C = X + Z

If the weighted cross product is calculated, all terms involving combinations of W, Y and Z are canceled, as in the calculation of the second order, leaving the mean of X ³ :

ÃBC = X ^T

Unfortunately, if X is a zero mean time signal, as expected, then your cube average is zero. Unlike calculating the average of X ² , which is positive for any non-zero value of X, X ³ has the same sign as X, so that positive and negative contributions tend to cancel out. Obviously, the same holds for any odd power of X, corresponding to an odd number of module inputs, but even exponents greater than two can also lead to erroneous results; for example, four inputs with components (X, X, -X, -X) will have the same product / average as (X, X, X, X).

This problem can be solved by using a variant of the weighted product technique. Before being weighed, the signal from each

• · · · · · ··· ····· «product is discarded, taking the absolute value of the product. The signs of each product term are examined. If they are all the same, the absolute value of the product is applied to the weight. If any of the signs is different from the others, the negative of the absolute value of the product is weighted. Since the number of possible combinations of the same sign may not be the same as the number of possible combinations of a different sign, a weighting factor comprised by the ratio of the number of combinations of the same sign to a different sign is applied to products of absolute value negative, for compensation. For example, a three-input module has two ways for signals to be the same, with eight possibilities, leaving six possible ways for signals to be different, resulting in a scale factor of 2/6 = 1/3. This compensation causes the integrated or added product to grow in a positive direction if and only if there is a signal component common to all inputs of a decoding module.

However, so that the averages of different order modules are comparable, they must all have the same dimensions. A conventional second-order correlation involves averages of multiplications of two inputs and thus of quantities with the dimensions of energy or power. Thus, the terms to be weighted in correlations must also be modified to have the power dimensions. For a k-order correlation, the absolute values of the individual product must, therefore, be raised to the power of 2 / k, before having the average kitten.

Obviously, regardless of the order, the individual input energies of a module, if necessary, can be calculated as the average of the square of the corresponding input signal, and need not first be raised to the k-th power and then reduced to a second order quantity.

Returning to the description of figure 4A, the transform bin outputs of each of the blocks can be grouped in sub-bands by a respective function or device 407, 409 and 411. The sub-bands can be

approach the critical bands of the human ear, for example. The remainder of the module modality of figures 4A to 4C operates separately and independently in each subband. In order to simplify the design, only the operation of a subband is shown.

Each subband of blocks 407, 409 and 411 is applied to a frequency attenuator or frequency attenuation function 413, 415 and 417 (from this point on, frequency attenuator), respectively. The purpose of frequency attenuators is explained below. Each attenuated frequency subband of a frequency attenuator is applied to optional fast attenuators or attenuation functions 419, 421 and 423 (from this point on, fast attenuators), respectively, which provide a time domain attenuation. Although preferred, fast attenuators can be omitted when the time constant of the fast attenuators is close to the block length of the forward transformation that generated the input bins (for example, a forward transform in the supervisor 201 of figure 1 ). The fast attenuators are fast in relation to the attenuators or the slow variable time constant attenuation functions 425, 427 and 429 (from this point on, the slow attenuators) that receive the respective outputs from the fast attenuators. Examples of fast and slow attenuator time constant values are given below.

Thus, if a fast attenuation is provided by the inherent operation of a forward transform or by the fast attenuator, a two-stage attenuation action is preferred, in which the second, slower, stage is variable. However, a single attenuation stage can provide acceptable results.

The time constants of the slow attenuators are preferably in sync with each other in a module. This can be accomplished, for example, by applying the same control information to each slow attenuator and by configuring each slow attenuator to respond in the same way to applied control information. The derivation of information for the control of slow attenuators is described below

xo.

Preferably, each pair of attenuators is in series, in the manner of pairs 419/425, 421/427 and 423/429, as shown in figures 4A and 4B, in which a fast attenuator feeds a slow attenuator. A series arrangement has the advantage that the second stage is resistant to short rapid signal peaks at the input of the pair. However, similar results can be obtained by configuring the attenuator pairs in parallel. For example, in a parallel arrangement, the resistance of the second stage in a series arrangement to short fast signal peaks can be dealt with in the logic of a time constant controller.

Each two-stage attenuator stage can be implemented by a single-pole low-pass filter (a leaking integrator), such as an RC low-pass filter (in an analogous mode) or, equivalently, a low-pass filter. low pass of first or15 dem (in a digital mode). For example, in a digital modality, first order filters can be made, each, as a bicad filter, a general second order IIR filter, in which some of the coefficients are set to zero so that the filter functions as a filter first order. Alternatively, the two attenuators can be combined in a single second order bikini stage, although it is simpler to calculate coefficient values for the second (variable) stage if it is separated from the first (fixed) stage.

It should be noted that in the form of figures 4A, 4B and 4C, all signal levels are expressed as energy levels (at square 25), unless an amplitude is required by taking a square root. The attenuation is applied to the energy levels of the applied signals, causing the attenuators to detect RMS, instead of detecting average (the attenuators of average detection are fed by linear amplitudes). Due to the fact that the signals applied to the attenuators are squared levels, the attenuators react to sudden increases in the signal level more quickly than the average attenuators, since the increases are magnified by the elevation squared function.

The two-stage attenuators thus provide an average time for each subband of each input channel energy (that of the first channel is provided by the slow attenuator 425 and that of the mth channel by the slow attenuator 427) and the average of each subband of the common energy of input channels (provided by the slow attenuator 429).

The average energy outputs of the slow attenuators (425, 427, 429) are applied to combiners 431, 433 and 435, respectively, in which (1) the neighbor energy levels (if any) (from supervisor 201 in figure 2, for example) are subtracted from the attenuated energy level of each of the input channels, and (2) the higher-order neighbor energy levels (if any) (from supervisor 201 in figure 2, for example) are subtracted from each of the slow attenuator average power outputs. For example, each module receiving an input 3 '(FIG. 1 and 2) has two neighboring modules and receives neighbor energy level information that compensates for the effect of those two neighboring modules. However, none of those modules is a higher-order module (that is, all modules sharing input channel 3 'are two-input modules). In contrast, module 28 (FIG. 1 and 2) is an example of a module that has a higher-order module sharing one of its inputs. As20 yes, for example, in module 28, the average energy output from a slow attenuator to an input 13 'receives higher level neighbor level compensation.

The resulting compensated neighbor energy levels for each sub-range of each of the module inputs are applied to a 437 function or device, which calculates a nominal primary direction in progress for those energy levels. The direction indication can be calculated as the vector sum of the weighted energy inputs. For a two-input module, this simplifies to be the L / R ratio of the attenuated and neighboring compensated input signal energy levels.

Assume, for example, a flat surround arrangement in which the channel positions are given as 2-doubles representing x, y coordinates for two inputs. The listener in the center is assumed to be, say, at (0, 0). The left front channel, in normalized spatial coordinates, is at (1, 1). The right front channel is at (-1.1). If the left input amplitude (Lt) is 4 and the right input amplitude (Rt) is 3, then, using those amplitudes as weighting factors, the primary direction in nominal travel is:

(4 * (1.1) + 3 * (-1, 1)) / (4 + 3) = (0.143, 1), or slightly to the left of the center of a horizontal line connecting the Left and Right.

Alternatively, once a master matrix is defined, the spatial direction can be expressed in matrix coordinates, instead of physical coordinates. In this case, the input amplitudes, normalized to the quadratic sum to one, are the effective matrix coordinates of the direction. In the example above, the left and right levels are 4 and 3, which normalize to 0.8 and 0.6. Consequently, the direction is (0.8,

0.6). In other words, the primary direction in nominal travel is the quadratic sum version for a normalized square root of the compensated neighbor's attenuated input energy levels. Block 337 produces the same number of outputs, indicating a spatial direction, as there are inputs for the module (two in this example).

The offset energy levels of compensated neighbor for each subband of each of the module inputs applied to the direction determination device or device 337 are also applied to a function or device 339 that calculates the compensated neighbor cross correlation ( neighbor-compensated_xcor), Block 339 also receives as input the weighted common energy of the module inputs of each subband from the slow variable attenuator 329, which was compensated in combiner 335 by the order neighbor energy levels higher, if any. The cross-correlation of compensated neighbor is calculated in block 339 as the highest-order compensated common energy divided by the root M-th, where Μ is the number of inputs, of the product of the attenuated energy levels of compensated neighbor for each of the module input channels, to derive a value from

true mathematical correlation in the range of 1.0 to -1.0. Preferably, the values from 0 to -1.0 are taken to be zero. Neighbor-compensated_xcor provides an estimate of the cross-correlation that exists in the absence of other modules.

The neighbor-compensated_xcor of block 339 is then applied to a device or a weighting function 341, which weighs neighborcompensated_xcor with the compensated neighbor direction information to produce a weighted direction compensated neighbor cross correlation (direction-weighted_xcor ). Weighting increases as the primary direction in nominal travel deviates from a centralized condition. In other words, unequal input amplitudes (and thus energies) cause a proportional increase in direction-weighted_xcor. Direction-weighted_xcor provides an estimate of image compactness. Thus, in the case of a two-input module having, for example, the left L and right R inputs, the weighting increases as the direction deviates from the center towards the left or right (that is, the weighting is the same in any direction for the same degree of deviation from the center). For example, in the case of a two input module, the neighborweighted_xcor value is weighted by an L / R or R / L ratio, so that a non-uniform force distribution-forceed_xcor signal distribution to 1.0. For a two input module like this:

when R> L, direction-weighted_xcor = (1 - ((1 - neighbor-weighted_xcor) * (L / R))), and when R <L, direction-weighted_xcor = (1 - (((1 - neighborweighted_xcor) * ( R / L))).

For modules with more than two inputs, calculating directionweighted_xcor from neighbor-weighted_xcor requires, for example, replacing L / R or R / L above with a uniformity measure, which varies between 1.0 and 0. For example, to calculate the uniformity measure for any number of inputs, normalize the input signal levels by the total input power, resulting in normalized input levels

that add up in a sense of energy (squared) up to 1.0. Divide each normalized input level by the normalized input level similarly to a signal centered in the array. The smallest relation becomes the measure of uniformity. So, for example, for a three-input module with an input having a level of zero, the uniformity measure is zero, and the direction-weighted_xcor is equal to one. (In this case, the signal is at the edge of the three-input module, on a line between two of its inputs, and a two-input module (lower in the hierarchy) decides where on the line is the primary direction in nominal progress, and how much along the line the output signal must be spread.)

Returning to the description of figure 4B, direction-weighted_xcor is weighted further by its application to a function or device 443, which applies a random_xcor weighting to produce an effective_xcor. Effective_xcor provides an estimate of the input signal distribution format.

Random_xcor is the average vector product of the input magnitudes divided by the square root of the average input energies. The value of random_xcor can be calculated assuming that the output channels were originally module input channels, and calculating the value of xcor that results from all those channels having independent signals, but of equal level, going through a downmix passively. According to this approach, for a three-output module with two inputs, random_xcor is set to 0.333 and, for a five-output module (three interior outputs) with two inputs, random_xcor is calculated for 0.483 . The value of random_xcor needs to be calculated only once for each module. Although such random_xcor values have shown the provision of satisfactory results, the values are not critical, and other values can be used at the discretion of the system designer. A change in the value of random_xcor affects the dividing line between the two operating regimes of the signal distribution system, as described below. The precise location of that dividing line is not critical.

The weighting of random_xcor performed by the function or by

device 343 can be considered as the normalization of the direction-weightedxcor value, so that an effectivexcor is obtained:

effective_xcor = (direction-weighted_xcor - randomxcor) / (1 - random_xcor), if direction-weighted_xcor> random_xcor, effective_xcor = 0, otherwise.

The weighting of random_xcor accelerates the reduction in directionweighted xcor, as direction-weighted_xcor decreases below 1.0, so that when direction-weighted_xcor equals random xcor, the value of effective xcor is zero. Because the outputs of a module represent directions along an arc or line, values of effective_xcor less than zero are treated as equal to zero.

Information on the control of the slow attenuators 325, 327 and 329 is derived from the energies of slow and fast attenuated input channels of the un compensated neighbor and from the common energy of the slow and fast attenuated input channels. In particular, a function or device 345 calculates a fast compensated non-neighbor cross-correlation in response to the fast attenuated input channel energies and the common energy of the fast attenuated input channels. A function or device 347 calculates a fast compensated non-neighbor direction (relation or vector, as discussed above in relation to the description of block 337), in response to the fast attenuated input channel energies. A function or device 349 calculates a slow compensated non-neighbor cross-correlation in response to slow attenuated input channel energies and the common energy of slow attenuated input channels. Fast non-compensated neighbor cross correlation, fast non-compensated neighbor direction, slow non-compensated neighbor cross correlation and slow non-compensated neighbor direction, together with direction-weighted xcor from block 341, are applied to a device or function 353, which provides the information for controlling the slow variable attenuators 325, 327 and 329, for the adjustment of their time constants (from this point on time adjustment constants). Preferably, the same control information is applied to each variable slow attenuator. Different-

't. mind the other quantities fed to the time constant selection box, which compare a fast measure with a slow one, direction-weighted_xcor is preferably used without reference to any fast value, so that the absolute value of direction5 weighted_xcor is greater than a threshold, this can cause the setting time constants 353 to select a faster time constant. The rules for operating 353 set time constants are set out below.

Generally, in a dynamic audio system, it is desirable to use slow time constants as much as possible, staying at a quiescent value, to minimize audible disturbance of the reproduced sound field, unless a new event occurs in the audio signal, in which case it is desirable for a control signal to change rapidly to a new quiescent value, then remain at that value until another new event occurs. Typically, audio processing systems have equal amplitude changes with a new event. However, when dealing with vector products or a cross correlation, the novelty and breadth are not always equivalent: a new event can cause a decrease in the cross correlation. By detecting changes in parameters relevant to the module's operation, specifically cross-correlation and direction measurements, the time constants of a module can be accelerated and quickly assume a new control state, as desired.

The consequences of improper dynamic behavior include image migration, vibrations (a channel quickly activating and deactivating), pumping (unnatural changes in level) and, in a multiple band mode, humming (vibrations and pumping on a band by band). Some of these effects are especially critical for the quality of isolated channels.

A modality such as that of figures 1 and 2 employs a lattice of decoding modules. Such a configuration results in two classes of dynamic problems: inter and in47 dynamics

I »ι» I ι Λ n II · » ^H il c ⁿ n»> - r »(·>

C ** A I OH tramodule. In addition, the various ways of implementing audio processing (for example, broadband, single module using a linear FFT or MDCT filter bank, or a discrete filter bank, critical band or otherwise), each, requires your own dynamic behavior optimization.

The basic decoding process in each module depends on a measure of the energy relationships of the input signals and a measure of the cross correlation of the input signals (in particular, the direction-weighted_xcor) described above; output of block 341 in figure 4B), which together control the distribution of control signal among the outputs of a module. The derivation of these basic quantities requires attenuation, which, in the time domain, requires the computation of a time-weighted average of the instantaneous values of those quantities. The range of time constants required is quite large: very short (1 ms, for example) for rapid transient changes in signal conditions, very long (150 ms, for example) for low correlation values, where the instantaneous variation it is likely to be much higher than the true average value.

A common method of implementing variable time constant behavior is, in analogical terms, the use of an acceleration diode. When the instantaneous level exceeds the average level by a limit quantity, the diode conducts, resulting in a shorter effective time constant. A drawback of this technique is that a momentary peak at an otherwise steady input can cause a large change in the attenuated level, which then decays very slowly, providing an unnatural emphasis on isolated peaks that would otherwise have little audible consequence.

The correlation calculation described in relation to the modality of figures 4A to 4C makes the use of acceleration diodes (or their DSP equivalent) problematic. For example, all attenuators in a particular module, preferably, have synchronized time constants, so that their attenuated levels are comparable. Therefore, a

Global time constant (conjugated) switching mechanism is preferred. In addition, a rapid change in signal conditions is not necessarily associated with an increase in the common energy level. The use of an acceleration diode for this level is prone to produce inaccurate oriented estimates of correlation. Therefore, the aspect modalities of the present invention, preferably, use a two-stage attenuation, without an equivalent diode acceleration. The correlation and direction estimates can be derived at least from the first and second stages of the attenuators, for the regulation of the time constant of the second stage.

Each pair of attenuators (eg 319/325), the first stage, the fixed fast stage, the time constant can be set to a fixed value, such as 1 ms. The second stage, the variable slow stage, the time constants can be selectable, for example, between 10 ms (fast), 30 ms (average) and 150 ms (slow). Although such time constants have been shown to provide satisfactory results, their values are not critical and other values can be used at the discretion of the system designer. In addition, the second stage time constant values can be continuously variable, rather than discrete. The selection of time constants can be based not only on the signal conditions described above, but also on a hysteresis mechanism using a rapid indicator, which is used to ensure that once a genuine rapid transition is found, the system remains in fast mode, avoiding the user of the average time constant, until the signal conditions allow the slow time constant again. This can help ensure quick adaptation to new signal conditions.

The selection of which of the three possible second stage time constants to use can be carried out by the setting time constants 353, according to the following rules for the case of two entries:

If the absolute value of direction-weighted_xcor is less than a first reference value (0.5, for example) and the absolute difference between non-neighbor-compensated_xcor fast and non-neighbor-compensates49

ted_xcor slow is less than the same first reference value, and the absolute difference between fast and slow direction ratios (each of which has a range of +1 to -1) is less than the same first reference value , then, the slow second stage time constant is used, and the fast indicator is set to True, allowing a subsequent selection of the average time constant.

Otherwise, if the fast indicator is True, the absolute difference between non-neighbor-compensated_xcor fast and slow is greater than the first reference value and less than a second reference value (0.75, for example) , the absolute digital signal between the fast and slow temporary L / R ratios is greater than the first reference value and less than the second reference value, and the absolute value of directionweighted_xcor is greater than the first reference value and less than the second reference value, then the average second stage time constant is selected.

Otherwise, the fast second stage time constant is used, and the fast indicator is set to False, not allowing the subsequent use of the average time constant until the slow time constant is selected again.

In other words, the slow time constant is chosen when all three conditions are less than a first reference value, the average time constant is chosen when all conditions are between a first reference value and a second reference value , and the previous condition was the slow time constant, and the fast time constant is chosen when either condition is greater than the second reference value.

Although the newly established rules and reference values have been shown to produce satisfactory results, they are not critical, and variations in rules or other rules that take into account fast and slow cross correlation and fast and slow direction can be employed at the discretion of the system designer. In addition, other changes can be made. For example, it can be simpler, but also effective 50

Φ · φφφ φφ φφφφ φφ φφφφφφφφφ φφ • · φ φ φφφφ φφ φ * φφφ φ φ φφφφ φφφ φφφ φφ φφφ φφ φ φ φ φφφφφ φ • φφ φφ Φφ φφ φφ φ φ φφ tive use a processing type of diode acceleration, but with a combined operation, so that if any attenuator in a module is in a fast mode, all other attenuators will also be switched to a fast mode. It may also be desirable to have separate attenuators for determining the time constant and signal distribution, with the attenuators for determining the time constant maintained with fixed time constants, and only the varied signal distribution time constants.

Due to the fact that, even in fast mode, the attenuated signal levels require several milliseconds to adapt, a time delay can be built into the system, to allow control signals to adapt, before applying them to a signal path. . In a broadband mode, this delay can be perceived as a discrete delay (5 ms, for example) in the signal path. In multiple-band (transformed) versions, the delay is a natural consequence of block processing, and if an analysis of a block is performed before the signal path of that block is registered, no explicit delay can be required.

The multiple band modalities of aspects of the invention can use the same time constants and rules as the broadband versions, except that the sampling rate of the attenuators can be set to the signal sampling rate divided by the size of block (for example, the rate of biocoil), so that the coefficients used in the attenuators are adjusted accordingly.

For frequencies below 400 Hz, in multiple band modes, the time constants are preferably scaled inversely to the frequency. In the broadband version, this is not possible as there are no separate attenuators at different frequencies, so that, as a partial compensation, a 30-band / pre-emphasis passband filter can be applied to the input signal for the control path, for emphasizing upper and lower middle frequencies. This filter can have, for example, a high pass feature of two

poles with a 200 Hz corner frequency, plus a 2-pole low-pass characteristic with a 8000 Hz corner frequency, plus a pre-emphasis network applying 6 dB of intensification from 400 Hz to 800 Hz and 6 others dB from 1600 Hz to 3200 Hz. Although such a filter has proved to be adequate, the filter characteristics are not critical, and other parameters can be used at the discretion of the system designer.

In addition to the time domain attenuation, the multi-band versions of aspects of the invention preferably also employ frequency domain attenuation, as described above in relation to figure 4A (frequency attenuators 413, 415 and 417). For each block, non-compensated neighbor energy levels can be weighted with a sliding frequency window, adjusted to approach a 1/3 octave bandwidth (critical band), before being applied to domain processing subsequent time period described above. Since transform-based filter banks have intrinsically linear frequency resolution, their window width (in number of transform coefficients) increases with increasing frequency and is usually just a wide to low transform coefficient. frequencies (below about 400 Hz). Therefore, the full attenuation applied to the multi-band processor is based more on a time domain attenuation at low frequencies, and a frequency domain attenuation at higher frequencies, where a quick time response is more likely to be needed more. sometimes.

Going back to the description in figure 4C, the preliminary scaling factors (shown as PSFs in figure 2), which ultimately affect the dominant / fill / endpoint signal distribution, can be produced by a combination of devices or functions 455, 457 and 459, which calculate the dominant scale factor components, the fill scale factor components and the excess end point energy scale components, respectively, respective normalizers or normalizing functions 361, 363 and 365, and a device or function 367, which takes the largest among the components of

dominant and fill scale factor and / or additive combination of fill scale factor components and excess end point energy. Preliminary scale factors can be sent to a supervisor, such as supervisor 201 of figure 2, if the module is one of a plurality of modules. Preliminary scale factors can also each range from zero to one.

Dominant Scale Factor Components

In addition to effective_xcor, a 355 device or function (calculating dominant scale factor components) receives the compensated neighbor direction information from block 337 and information regarding local matrix coefficients of a local matrix 369, so that it can determine the nearest N output channels (where N = number of inputs), which can be applied to a weighted sum for the production of primary direction coordinates in nominal progress and application of dominant scale factor components to them for the production of dominant coordinates. The output of block 355 is a scaling factor component (per subband), if the primary direction in nominal travel is likely to coincide with an output direction or, otherwise, multiple scaling factor components (one by the number of inputs per subband) en20 revolving the primary direction in nominal progress and applied in appropriate proportions, in order to form a panorama or map the dominant signal to the correct virtual location in a sense of power preservation (that is, for N = 2, the two dominant channel scale factor components assigned must have the quadratic sum calculated up to effecti25 ve_xcor).

For a two-input module, all output channels are in a line or an arc, so there is a natural ordering (from left to right), and it is readily apparent which channels are close to each other. For the hypothetical case discussed above having two input channels and five output channels with sin / cos coefficients as shown, the primary direction in nominal progress can be assumed to be (0.8, 0.6) between the Left Middle channel ML (0.92, 0.38) and

the central channel C (0.71, 0.71). This can be done by finding two consecutive channels in which the L coefficient is greater than the primary direction L coordinate in nominal travel and the channel to its right has a lower L coefficient than the dominant L coordinate.

The dominant scale factor components are prorated to the two closest channels in a direction of constant power. To do this, a system of two equations and two unknowns is solved, the unknowns being the scale factor component of the dominant component of the channel to the left of the dominant direction (SFL), and the corresponding scale factor component to the right of the primary direction in nominal progress (SFR) (these equations are solved for SFL and SFR).

first_dominant_coord = SFL * left channel matrix value 1 + SFR * right channel matrix value 1 second_dominant_coord = SFL * left channel matrix value 2 + SFR * right channel matrix value 2

Note that left and right channel means the channels involving the primary direction in nominal travel, not the input channels L and R for the module.

The solution is the anti-dominant level calculations for each channel, normalized to the quadratic sum up to 1.0, and used as the components of the dominant distribution scale factor (SFL, SFR), each for the other channel. In other words, the antidominant value of an output channel with coefficients A, B for a signal with coordinates C, D is the absolute value of AD - BC. For the numerical example under consideration:

Anti-sound (ML channel) = abs (0.92 * 0.6 - 0.38 * 0.8) = 0.248 Anti-sound (channel C) = abs (0.71 * 0.6 - 0.71 * 0.8) = 0.142 (where abs indicates taking the absolute value).

The normalization of the last two numbers to a dramatic sum of up to 1.0 leads to values of 0.8678 and 0.4949, respectively. So, when switching these values to opposite channels, the components of the dominant scale factor are (note that the value of the dominant scale factor,

before direction weighting, it is the square root of effective_xcor):

ML sun sf = 0.4969 * square root (effective_xcor)

C dom sf = 0.8678 * square root (effective_xcor) (the dominant signal is closer to Count than to MidLout).

The use of a normalized antidom component of one channel, such as the dominant scale factor component of the other channel, can be better understood considering what happens if the primary direction in nominal travel happens to point exactly at one of the two channels selected. Suppose the coefficients of one channel are [A, B] and the coefficients of the other channel are [C, D] and the primary direction coordinates in nominal progress are [A, B] (pointing to the first channel), then:

Antidom (first channel) = abs (AB - BA)

Antidom (second channel) = abs (CB - DA)

Note that the first antidom value is zero. When the two antidom signals are normalized to a quadratic sum up to 1.0, the second antidom value is 1.0. When switched, the first channel receives a dominant scale factor component of 1.0 (times the square root of effective_xcor) and the second channel receives 0.0, as desired.

When this approach is extended to modules with more than two inputs, there is no longer a natural ordering that occurs when the channels are in line or in an arc. Again, block 337 of figure 4B, for example, calculates the primary direction coordinates in nominal travel by taking the input amplitudes, after a neighbor offset, and normalizing them to a quadratic sum up to one. Block 455 of figure 4B, for example, then identifies the nearest N channels (where N = number of inputs) that can be applied to a weighted sum to lead to the dominant coordinates. (Note: distance or proximity can be calculated as the sum of the squared coordinate differences, as if there were spatial coordinates (x, y, z)). Thus, the nearest N channels are not always captured, because they have to be weighted to add up to the primary direction in nominal travel.

• ··· • «• ·«

For example, suppose you have a three input module powered by a triangle of channels: Ls, Rs and Topo, as in figure 5. Assume that there are three inner output channels close together close to the bottom of the triangle, with coefficients of local module matrix [0.71,

0.69, 0.01], [0.70, 0.70, 0.01], and [0.69, 0.71, 0.01], respectively. Assume that the primary direction in nominal travel is slightly below the center of the triangle, with coordinates [0.6, 0.6, 0.53], (Note: half of the triangle has coordinates [0.5, 0.5, 0.707]). The three channels closest to the primary direction in nominal travel are those three inner channels at the bottom, but they do not add up to the dominant coordinates using scale factors between 0 and 1, so instead, two are chosen from the bottom and top end point channel for the distribution of the dominant signal, and the three equations for the three weighting factors are solved, in order to complete the dominant calculation and proceed to the fill and end point calculations .

In the examples in figures 1 and 2, there is only one module with three inputs, and it is used for the derivation of only one interior channel, which simplifies the calculations.

Fill Scale Factor Components

In addition to effective_xcor, the device or function 357 (calculate fill scale factor components) receives random_xcor, direction-weighted_xcor from block 341, EQUIAMPL (EQUIAMPL is defined and explained below), and information regarding local matrix coefficients a from the local matrix (in this case, the same filling factor factor is not applied to all outputs, as explained below in relation to figure 14B). The output of block 457 is a scale factor component for each module output (per subband).

As explained above, effective_xcor is zero when directionweighted_xcor is less than or equal to random_xcor. When direction30 weighted_xcor> random_xcor, the fill scale factor component for all output channels is:

fill scale factor component = square root

da (1 - effective_xcor) * EQUIAMPL

Thus, when direction-weighted_xcor = random_xcor, the effective_xcor is 0, so that (1 - effective_xcor) is 1.0, so that the fill-scale factor component is equal to EQUI5 AMPL (ensuring that output = input power in that condition). That point is the maximum value that the fill scale factor components reach.

When weighted_xcor is less than random_xcor, the dominant scale factor component (s) is (are) zero, and the fill scale factor components are reduced to zero as direction-weighted_xcor approaches zero :

fill scale factor component = square root (direction-weighted_xcor / random_xcor) * EQUIAMPL

Thus, at the border, when direction-weighted_xcor = ran15 dom_xcor, the preliminary fill factor factor is again equal to EQUIAMPL, ensuring continuity of the results of the above equation for the case of direction-weighted_xcor greater than random_xcor.

Associated with each decoder module are not only a random_xcor value, but also an EQUIAMPL value, which is a scale factor value that all scale factors should have, if the signals are distributed equally, so that power is preserved, specifically:

EQUIAMPL = square root of (number of decoder module 25 input channels / number of decoder module output channels) of:

For example, for a two input module with three outputsEQUIAMPL = square root (2/3) = 0.8165 where square root () means square root of ().

For a two input module with 4 outputs:

EQUIAMPL = square root (2/4) = 0.7071.

For a two input module with 5 outputs:

EQUIAMPL = square root (2/5) = 0.6325.

Although such EQUIAMPL values have been shown to provide satisfactory results, the values are not critical, and other values can be used at the discretion of the system designer. Changes in the EQUIAMPL value affect the output channel levels for the fill condition (intermediate correlation of the input signals) with respect to the output channel levels for the dominant condition (maximum condition of the input signals) and the condition all end points (minimal correlation of input signals).

Endpoint Scale Factor Components

In addition to neighbor-compensated_xcor (from block 439, figure 4B), a device or function 359 (calculating excess endpoint energy scale factor components) receives the respective non-compensated neighbor energy attenuated from the first to the last th input (from blocks 325 and 327) and, optionally, information regarding the local matrix coefficients from the local matrix (in case one or both end points of the module do not match an input and the module applies the excess endpoint energy to the two exits having directions closer to the input direction, as further discussed below). The output of block 359 is a scale factor component for each endpoint output, if the directions match the input directions, otherwise, two scale factor components, one for each of the outputs closest to the end, as explained below.

However, the excess endpoint energy scale factor components produced by block 359 are not the only endpoint scale factor components. There are three other sources of end point scale factor components (two in the case of a single stand-alone module):

First, in preliminary scale factor calculations for a particular module, end points are possible candidates for dominant signal scale factor components for block 355 (and the

normalizer 361).

Second, in the filling calculation ”of block 357 (and normalizer 363) of figure 4C, the end points are treated as possible filling candidates, along with all the inner channels. Any non-zero scale factor component can be applied to all outputs, even the end points and the chosen dominant outputs.

Third, if there is a multi-module lattice, a supervisor (such as supervisor 201 in the example in figure 2) performs a fourth final assignment of the endpoint channels, as described above with respect to figures 2 and 3.

So that block 459 calculates the excess end-point energy scale factor components, the total energy in all interior outputs is reflected back to the module inputs, based on neighbor-compensated_xcor, to estimate how much of the energy from the interior outputs contributes from each input (interior energy at the input 'n'), and that energy is used to compute the excess end point energy scale factor component on each module output that matches a entry (that is, a full stop).

Reflection of the inner energy back to the inputs is also required in order to provide the necessary information by a supervisor, such as supervisor 201 in figure 2, for the calculation of neighbor levels and higher order neighbor levels. . One way to calculate the interior energy contribution at each of the module inputs and to determine the excess end point scale factor component for each end point output is shown in figures 6A and 6B.

Figures 6A and 6B are diagrams of functional blocks that show, respectively, in a module, like any of the modules 24 to 34 of figure 2, a suitable arrangement for (1) the generation of the total estimated interior 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 neighbor-compensated_xcor (see figure 4B, the output of block 439), generating

an excess endpoint energy scale factor component for each of the module endpoints. The total estimated interior energy for each input of a module (FIG. 6A) is required by the supervisor, in the case of a multiple module arrangement and, in any case, by the module itself, in order to generate the scale factor components of excess end point energy.

Using the scale factor components derived in blocks 455 and 457 of figure 4C, together with other information, the arrangement in figure 6A calculates the total estimated energy at each interior output (but not at its end point outputs). Using the calculated indoor output energy levels, it multiplies each output level by the matrix coefficient relative to that output for each input [m inputs, m multipliers], which provides the energy contribution of all indoor output channels to the obtaining the total interior energy contribution of that 15 entry. The total interior energy contribution of each input is reported to the supervisor and is used by the module to calculate the excess end point energy scale factor component for each end point output.

Referring to figure 6A, in detail, the total attenuated energy level for each module input (preferably not compensated neighbor) is applied to a set of multipliers, a multiplier for each of the module's internal outputs. For simplicity of representation, figure 6A shows two inputs, 1 in, and two interior outputs, X and Z. The total attenuated energy level for each module input is multiplied by a matrix coefficient (from the local module matrix) , which relates the input in particular to one of the module's internal outputs (note that the matrix coefficients are their own inverses, because the matrix coefficients have the quadratic sum calculated up to one). This is done for each interior input and output combination. Thus, as shown in figure 6A, the total energy level attenuated at input 1 (which can be obtained, for example, at the output of slow attenuator 425 in figure 4B) is applied to a multiplier 601, which multiplies that energy level by a co-

efficient matrix relating the interior output X to input 1, providing a stepped output energy level component Xi at output X. Similarly, multipliers 603, 605 and 607 provide stepped energy level components X _m , Zi and Z _m .

The energy level components for each interior output (for example, Xi and X _m ; Zy and Z _m ) are added in combinators 611 and 613 in an amplitude / power manner according to neighbor-compensated_xcor. If the inputs for a combiner are in phase, indicated by the weighted neighbor correlation of 1.0, their linear amplitudes are added. If they are uncorrelated, indicated by a weighted neighbor cross-correlation of zero, their energy levels are added. If the cross-correlation is between one and zero, the sum is partly a sum of amplitude and partly a sum of power. In order to properly add the inputs for each combiner, both the sum of amplitude and the sum of power are calculated and weighted by neighbor-compensated_xcor and (1 - neighbor-weighted_xcor), respectively. In order to obtain the weighted sum, the square root of the power sum is taken, to obtain an equivalent amplitude, or the sum of linear amplitude is squared, to obtain its power level, before being weighted sum is made. For example, taking the latter approach (weighted sum of powers), if the amplitude levels are 3 and 4 and neighbor-weighted_xcor is the sum of amplitude 3 + 4 = 7, or a power level of 49 and the sum power energy is 9 + 16 = 25. Thus, the weighted sum is 0.7 * 49 + (1 - 0.7) * 25 = 41.8 (power energy level) or by taking the square root , 6.47.

The sum products (X ₁ + X _m ; Zi + Z _m ) are multiplied by the scale factor components for each of the outputs, X and Z, in multipliers 613 and 615, for the production of the total energy level in each interior exit, which can be identified as X 'and Z'. The scale factor component for each of the interior outputs is obtained from block 467 (FIG. 4C). Note that the excess end point energy scale components of block 459 (FIG. 4C) do not affect the outputs

and the calculations made by the arrangement in figure 6A are not involved.

The total energy level at each interior output X 'and Z' is reflected back to the respective module inputs, by multiplying each one by a matrix coefficient (from the local module matrix) that relates the particular output to each one of the module inputs. This is done for each combination of interior outlet and entrance. Thus, as shown in figure 6A, the total energy level X 'at the interior output X is applied to a multiplier 617, which multiplies the energy level by a matrix coefficient relating the interior output X to an input 1 (which is the same as its inverse, as mentioned above), providing a scaled energy level component X / at input 1.

It should be noted that when a second order value, such as the total energy level X ', is weighted by a first order value, such as a matrix coefficient, a second order weight is required. This is equivalent to taking the square root of the energy to obtain an amplitude, multiplying that amplitude by the matrix coefficient and squaring the result to obtain an energy value back.

Similarly, multipliers 619, 621 and 623 provide staggered energy levels X _m ', Z <and Z _m '. The energy components related to each output (for example, X / and ZT, X _m 'and Z _m ') are added in combinators 625 and 627 in an amplitude / power manner, as described above in relation to combiners 611 and 613 , according to neighborcompensated_xcor. The outputs of the combiners 625 and 627 represent the total estimated interior energy for inputs 1 in, respectively. In the case of a multiple module lattice, this information is sent to the supervisor, such as the supervisor 201 of figure 2, so that the supervisor can calculate neighbor levels. The supervisor requests all total interior energy contributions for each input from all modules connected to that input, then inform each module, for each of its inputs, what was the sum of all other total interior energy contributions from all other modules connected to that input.

»· * ····

The result is the neighbor level for that entry in that module. The generation of neighbor level information is further described below.

The total estimated indoor energy contributed by each of the inputs 1 and m is also required by the module, in order to calculate the excess end point energy scale factor component for each end point output. Figure 6B shows how such scale factor component information can be calculated. For simplicity of representation, only the calculation of the scale factor component information for an end point is shown, it being understood that a similar calculation is performed for each end point output. The total estimated interior energy contributed by an input, such as input 1, is subtracted in a combiner or combination function 629 from the total input energy attenuated for the same input, input 1 in this example (the same total energy level attenuated at input 1, obtained, for example, at the output of the slow attenuator 425 of figure 4B, which is applied to a multiplier 601). The subtraction result is divided into the divisor or division function 631 by the total energy level attenuated for the same input 1. The square root of the division result is taken in a square root extractor or a square root extraction function 633 It should be noted that the operation of the divisor or division function 631 (and other divisors described here) must include a test for a null denominator. In this case, the quotient can be set to zero.

If there is only a single independent module, the preliminary end point scale factor components are thus determined by virtue of having determined the dominant scale, fill and excess end point energy factors.

Thus, all output channels including the end points have been assigned scale factors, and you can proceed to use them to perform signal path enrollment. However, if there is a lattice of multiple modules, each one assigned an end point scaling factor to each input feeding it, so that each input having more than one module connected to it has multiple factor assignments of * * ·

scale, one from each connected module. In this case, the supervisor (like supervisor 201 in the example in figure 2) performs a fourth final assignment of the endpoint channels, as described above with respect to figures 2 and 3, which the supervisor determines the end point scaling factors that suppress all scale factor assignments made by individual modules as end point scale factors.

In practical arrangements, there is no certainty that there is actually an exit channel direction corresponding to an end point position, although this is often the case. If there is no physical endpoint channel, but there is at least one physical channel beyond the endpoint, the endpoint energy has a picture for the physical channels closest to the end, as if there were a dominant signal component. In a horizontal arrangement, these are the two channels closest to the end point position, preferably using a constant energy distribution (the two scale factors have the quadratic sum calculated up to 1.0). In other words, when a direction of sound does not correspond to the position of a real sound channel, even if that direction is an end point signal, it is preferred to have an overview of it for the closest available pair of real channels, because if the sound moved slowly, it suddenly jumps from one output channel to another. Thus, when there is no physical endpoint sound channel, it is not appropriate to pan from an endpoint signal to a sound channel closer to the endpoint location, unless there is no physical channel beyond the endpoint end, in which case there is no choice but a sound channel that approaches the end point location.

Another way to implement such a use of panorama is that the supervisor, like supervisor 201 in figure 2, generates final scale factors, based on the hypothesis that each input also has a corresponding output channel (that is, each corresponding entry and exit coincide, representing the same location). Then, an output matrix, such as the variable 203 matrix in Figure 2, can map an output channel to one or more appropriate output channels, if there is no

actual output channel that corresponds directly to an input channel.

As mentioned above, the outputs of each of the devices or functions to calculate scale factor component 455, 457 and 459 are applied to the respective normalization devices or functions 461,

463 and 465. Such standardizers are desirable because the scale factor components calculated by blocks 455, 457 and 459 are based on compensated neighbor levels, whereas the final signal path matrix (in the master matrix, in the case of multiple modules, or in the local matrix, in the case of an independent module) involves non-compensated neighbor levels (the input signals applied to the matrix are not compensated neighbor). Typically, scale factor components are reduced in value by a normalizer.

A suitable way of implementing standards is as follows. Each normalizer receives the compensated neighbor input energy for each of the module inputs (as in combinators 331 and 333), the non-neighbor compensated input energy i for each of the module inputs (as in blocks 325

W and 327), the local matrix gum composition information of the local matrix, and the respective outputs of blocks 355, 357 and 359. Each normalizer cal20 provides a desired output for each output channel and an actual output level for each output channel, assuming a scale factor of 1. It then divides the desired output calculated for each output channel by the actual output level calculated for each output channel and takes the square root of the quotient for the provision of a factor potential preliminary scale for application to sum and / or greater than 367. Consider the following example.

Assume that the attenuated non-compensated neighbor energy levels of a two-input module are 6 and 8, and that the corresponding compensated neighbor energy levels are 3 and 4. Also assume a central interior output channel having matrix coefficients = (0.71.0.71), or squared: (0.5, 0.5). If the module selects an initial scale factor for this channel (based on compensated neighbor levels) of 0.5 or squared = 0.25, then the desired output level

• · · · · · · · · ·> · · · ·········· · · of this channel (assuming the sum of pure energy for simplicity and using compensated neighbor levels) is:

0.25 * (3 * 0.5 + 4 * 0.5) = 0.875.

Due to the fact that the actual input levels are 6 and 8, if the scale factor above (squared) 0.25 is used for the final signal path registration, the output level is:

0.25 * (6 * 0.5 + 8 * 0.5) = 1.75 instead of the desired output level of 0.875. The normalizer adjusts the scale factor to obtain the desired output level when non-compensated neighbor levels are used.

Actual output, assuming SF = 1 = (6 * 0.5 + 8 * 0.5) = 7.

(desired output level) / (actual output assuming SF = 1) = 0.875 / 7.0 = 0.125 = final squared scale factor

Final scale factor for that output channel = square root (0.125) = 0.354, instead of the initially calculated value of 0.5.

The sum and / or greater of 367, preferably, adds the fill scale factor and end point components for each output channel per subband, and selects the largest of the fill scale factor and end point components for each output channel by subband. The sum and / or greater function of block 367, in this preferred form, can be characterized as shown in figure 7. Specifically, the dominant scale factor components and the fill scale factor components are applied to a device or to a 701 function, which selects the largest of the scale factor components for each output (greater than 701) and applies them to an additive combiner or to a 703 function, which adds the scale factor components of the greater than 701 with the excess end point energy scale factors for each output. Alternatively, acceptable results can be obtained when the sum and / or greater than 467: (1) sum in both Region 1 and Region 2, (2) takes the largest in both Region 1 and Region 2, or (3 ) selects the largest in Region 1 and adds up in Region 2.

Figure 8 is an idealized representation of the way in which

an aspect of the present invention generates scale factor components in response to a cross-correlation measure. The figure is particularly useful for reference to the examples in figures 9A and 9B to figures 16A and 16B. As mentioned above, the generation of scale factor components can be considered to have two regions or operating regimes: a first region, Region 1, limited by everything dominant and uniformly populated, in which the available scale factor components are one mixture of dominant and fill scale factor components, and a second region, Region 2, bounded by uniformly filled and all end points, in which the available scale factor components are a mixture of fill scale factor components and of excess end point energy. The all-dominant boundary condition occurs when direction-weighted_xcor is one. Region 1 (dominant plus padding) extends from that boundary to the point where direction-weighted_xcor is equal to random_xcor, the condition uniformly fulfilled. The boundary condition for all endpoints occurs when direction-weighted_xcor is zero. Region 2 (padding plus endpoint) extends from the uniformly filled boundary condition to the boundary condition all endpoint. The uniformly filled border point ”can be considered to be in Region 1 or Region 2. As mentioned below, the precise border point is not critical.

As illustrated in figure 8, as the scale factor component (s) declines in value, the fill scale factor components increase in value, reaching a maximum according to the component (s) scale factor reaches (m) a zero value, at which point as the fill scale factor components decline in value, the excess end point energy scale factor components increase in value. The result, when applied to an appropriate matrix, which receives the module input signals, is an output signal distribution that provides a compact sound image, when the input signals are highly correlated, diffused (widened).

• *

from compact to wide as the correlation decreases, and progressively dividing or bulging out to multiple sound images, each at an end point, from wide, as the correlation continues to decrease until highly correlated.

Although it is desirable to have a single spatially compact sound image (in the nominal primary direction of the input signals) for the case of a full correlation and a plurality of spatially compact sound images (each at an end point) for the in case of no full correlation, the spatially di10 sound image fused between those extremes can be obtained in other ways than the one shown in the illustration in figure 8. It is not critical, for example, that the fill scale factor component values reach a maximum for the case of random_xcor = direction-weighted_xcor, nor that the values of the three scale factor components change linearly as shown. The modifications of the relationships in figure 8 (and the equations expressed here that underlie the figure) and other relationships between a suitable cross-correlation measure and scale factor values that are capable of producing a compact dominant for wide diffusion for a distribution of compact endpoint signal for a cross correlation measure from highly correlated to highly uncorrelated are also contemplated by the present invention. For example, instead of obtaining a signal distribution of endpoints from compact dominant to broad to diffused by using a doubled region approach, as described above, such results can be obtained by a mathematical approach, such as one employing the resolution of an equation based on pseudo-inverse.

Output Scale Factor Examples

A series of idealized representations, figures 9A and 9B through figures 16A and 16B, illustrates the output scale factors of a module for various examples of input signal conditions. For simplicity, a single independent module is assumed, so that the scale factors it produces for a variable matrix are the final scale factors. O

• · ·

module and an associated variable matrix have two input channels (such as left L and right R) that coincide with two end point output channels (which can also be designated L and R). In this series of examples, there are three interior output channels (such as left middle Lm, center C and right middle Rm).

The meanings of everything dominant, dominant and mixed fill, uniformly filled, of filled and mixed end points, and all end points are further illustrated in relation to the examples of. 9A and 9B to figures 16A and 16B. In each pair of figures (9A and 9B, for example), figure A shows the energy levels of two inputs, left L and right R and figure B shows scale factor components for the five outputs, left L, left mean LM, center C, right mean RM and right R. The figures are not to scale.

In figure 9A, the input energy levels, shown as two vertical arrows, are the same. In addition, direction-weighted_xcor (and effective_xcor) is 1.0 (full correlation). In this example, there is only one non-zero scale factor, shown in figure 9B as a single vertical arrow in C, which is applied to the central inner C channel output, resulting in a compact dominant signal. In this example, the output is centered (L / R = 1) and, thus, the channel coincides with the central inner output channel C. If there is no coincident output channel, the dominant signal is applied in proportions appropriate to the input channels. outputs, in order to create a panorama of the dominant signal for the correct virtual location between them. If, for example, there is no central output channel C, the left average output channels LM and right average RM would have a non-zero scale factor, causing the dominant signal to be applied equally to the outputs LM and RM. In this case of full correlation (all dominant signal), there are no filler and endpoint components. Thus, the preliminary scale factors produced by block 467 (FIG. 4C) are the same as the normalized dominant scale factor components produced by block 361.

In figure 10A, the energy levels are the same, but direction-

weighted_xcor is less than 1.0 and greater than random_xcor. Consequently, the scale factor components are those of Region 1 dominant and mixed fill factor components. Most of the normalized dominant scale factor component (from block 361) and the normalized fill scale factor component (from block 363) are applied to each output channel (through block 367), so that the dominant scale is located on the same central output channel C as in figure 10B, but is smaller, and the fill scale factors appear in each of the other output channels, L, LM,

RM and R (including end points L and R).

In figure 11A, the energy levels remain the same, but direction-weighted_xcor = random_xcor. Consequently, the scale factors, figure 11B, are those of the boundary condition between Regions 1 and 2 - the uniformly fulfilled condition in which there is no dominant or end point scale factor, only the scale factors having the same value in each exit (thus uniformly filled), as indicated by the identical arrows at each exit. The fill scale factor levels reach their highest value in this example. As discussed below, the fill scale factors can be applied non-uniformly, as well as in a tapered manner, depending on the input signal conditions.

In figure 12A, the input energy levels remain the same, but random_xcor and greater than zero (Region 2). Consequently, as shown in figure 12B, there are fill scale and end point factors, but not dominant scale factors.

In figure 13A, the input energy levels remain the same, but direction-weighted_xcor is zero. Consequently, the scale factors, shown in figure 13B, are those of the boundary condition at all end points. There are no interior output scaling factors, only 30 end-point scaling factors.

In the examples in figures 9A / 9B to 13A / 13B, due to the fact that the energy levels of the two inputs are equal, the direction-weigh-

·· ·· ted_xcor (as produced by block 441 of figure 4B) is the same as neighbor-compensated_xcor (as produced by block 439 of figure 4B). However, in figure 14A, the input energy levels are not the same (L is greater than R). Although neighbor-compensated_xcor is equal to random_xcor in this example, the resulting scale factors, shown in figure 14B, are not fill scale factors applied uniformly to all channels, as in the example in figures 11A and 11B. Instead, unequal input energy levels cause a proportional increase in direction-weighted_xcor (proportional to the degree to which the primary direction in nominal swing deviates from its central position), so that it becomes greater than neighbor-compensated_xcor , thereby causing the scale factors to be weighted more towards everything dominant (as illustrated in figure 8). This is a desired result because strongly weighted L or R signals must not be wide in width; they must have a compact width close to the end point of channel L or R. The resulting output, shown in figure 14B, is a non-zero dominant scale factor located closer to output L than to output R (the direction information of compensated neighbor, in this case, chute to locate the dominant P mode coding precisely in the left middle position LM), reduced fill scale factor amplitudes and no end point scale factor (direction weighting pushes the operation to the Region 1 of figure 8 (dominant and mixed padding)).

For the five outputs corresponding to the addressability data in figure 14B, the outputs can be expressed as:

Lout = Lt (SF _L )

MidLout + ((0.92) Lt + (0.38) Rt)) (SFmícíl)

Cout = ((0.45) Lt + (0.45) Rt)) (SF _C )

MidRout = ((0.38) Lt + (0.92) Lt)) (SF _Mid R)

Rout = Rt (SF _r )

Thus, in the example in figure 14B, although the scale factors (SF) for each of the four outputs in addition to MidLout are equal (padding), the corresponding signal outputs are not the same because Lt is greater than Rt (resulting in more exit sign to the left) and the dominant output in MidLeft is greater than the scale factor indicates. Due to the fact that the primary direction in nominal travel coincides with the output channel MidLeft, the ratio of Lt to Rt is the same as the matrix coefficients for the output channel MidLeft, specifically, 0.92 and 0.38 . Assume that those are the general amplitudes for Lt and Rt. To calculate the output levels, these levels are multiplied by the corresponding matrix coefficients, added and scaled by the respective factors:

output amplitude (output_channel_sub_i) = sf (i) * (Lt_Coeff (i) * Lt + Rt_Coeff (i) * Rt)

Although the mix between amplitude and energy addition is preferably taken into account (as in the calculations related to figure 6A), in this example the cross correlation is reasonably large (large dominant scale factor) and a common sum can be performed:

Lout = 0.1 * (1 * 0.92 + 0 * 0.38) = 0.092

MidLout = 0.9 * (0.92 * 0.92 + 0.38 * 0.38) = 9.00 Cout = 0.1 * (0.71) * 0.92 + 0.71 * 0.38 ) = 0.092 MidRout = 0.1 * (0.38 * 0.92 + 0.92 * 0.38) = 0.070 Rout = 0.1 * (0 * 0.92 + 1 * 0.38) = 0.038

So, this example demonstrates that the output signals at Lout,

Cout, MidRout and Rout are unequal, because Lt is greater than Rt, although the scale factors for those outputs are the same.

The fill scale factors can be equally distributed to the output channels, as shown in the examples in figures 10B, 11 Β, 12B and 14B. Alternatively, the fill scale factor components, instead of being uniform, can be varied with position in some way, as a function of the dominant (correlated) and / or end point (uncorrelated) input signal components (or , equivalently, as a function of the direction-weighted_xcor value). For example, for moderately high values of direction-weighted_xcor, the fill scale factor component ranges can be curved convexly, so that the channels

outputs near the primary direction in nominal travel receive more signal level than channels farther away. For directionweighted_xcor = random_xcor, the fill scale factor component amplitudes can flatten to a uniform distribution, and for direction-weighted_xcor <random_xcor, the amplitudes can be concave curved, favoring channels close to the end point directions.

Examples of such curved scale factor amplitudes are set out in figures 15B and 16B. The figure output

15B results from an entry (FIG. 15A) which is the same as in figure 10A, described above. The output of figure 16B results from an entrance (FIG. 16A) which is the same as in figure 12B, described above.

Communication between Module and Supervisor with Respect to Neighbor Levels and Neighbor Levels of Higher Order

Each module in a multiple module arrangement, as in the example in figures 1 and 2, requires two mechanisms, in order to support communication between it and a supervisor, such as the supervisor 201 in figure 2:

(a) one to choose and report the information required by the su20 expert to calculate neighbor levels and higher order neighbor levels (if any). The information required by the supervisor is the total estimated interior energy attributable to each of the module inputs, as generated, for example, by the arrangement in figure 6A.

(b) another to receive and apply the neighboring levels (if any) and higher-order neighboring levels (if any) from the supervisor. In the example in figure 4B, the neighbor levels are subtracted in respective combinators 431 and 433 from the attenuated energy levels of each entry, and the higher order neighbor levels (if any) are subtracted in the respective combiners 431, 433 and 435 of the attenuated energy levels of each input and the energy

through channels.

Once a supervisor knows all the total estimated interior energy contributions for each input in each module:

(1) it determines whether the total estimated indoor energy contributions of each input (added from all modules connected to that input) exceeds the total available signal level at that input. If the sum exceeds the total available, the supervisor schedules back each interior energy reported by each module connected to that input, so that they add up to the total input level.

(2) it informs each module of its neighboring levels at each input as the sum of all other energy inputs from that input (if any).

Higher-order neighbor (HO) levels are neighbor levels of one or more higher-order modules that share the inputs of a lower-level module. The calculation above neighbor levels refers only to modules in a particular entry that have the same hierarchy: all three-entry modules (if any), then all two-entry modules, etc. A HO neighbor level of a module is the sum of all neighbor levels of all higher order modules in that input (that is, the HO neighbor level in an input of a two-input module is the sum of all third, fourth order and higher order modules, if any, sharing the node of a two input module). Once a module knows what its HO neighbor levels are at a particular input from its inputs, it subtracts them, along with neighboring levels from the same hierarchy level, from the total input energy level of that input, for obtaining the compensated neighbor level at that input node. This is shown in Fig. 30B, where the neighbor levels for input 1 and input m are subtracted in combiners 431 and 433, respectively, from the outputs of the variable slow attenuators 425 and 427, and the order neighbor levels

highest for input 1, input m and common energy are subtracted in combiners 431, 433 and 435, respectively, from the outputs of variable slow attenuators 425, 427 and 429.

A difference between the use of neighboring levels and the neighboring levels of HO for compensation is that the neighboring levels of HO are also used to compensate for common energy through the input channels (for example, performed by subtracting a level of neighbor to HO in combiner 435). The rationale for this difference is that the common level of a module is not affected by adjacent modules of the same hierarchy10, but can be affected by a higher-order module sharing all the inputs of a module.

For example, assume input channels Ls (left surround), Rs (right surround) and Top, with an inner output channel in the middle of the triangle between them (rear high ring), plus an inner output channel in a row between Ls and Rs (rear of main horizontal ring), the first output channel needs a three input module to recover the signal common to all three inputs. So, the last output channel, being on a line between two inputs (Ls and Rs), needs a two input module. However, the total common signal level observed by the two-input module includes common elements of the three-input module that do not belong to the last output channel, so that the square root of the products is subtracted alongside the neighboring levels of HO of the common energy of the two-input module, to determine how much common energy is due solely to its inner channel (the last mentioned). Thus, in Figure 4B, the common attenuated energy level (from block 429) subtracted from it the common derived HO level to lead to a common compensated neighbor energy level (from combiner 435) that is used by the module to calculate (in block 439) neighbor-compensated_xcor.

The present invention and its various aspects can be implemented in an analog circuit, or, more likely, as software functions performed on digital signal processors, programmed general purpose digital computers, and / or digital final computers.

special quality. The interfaces between analog and digital signal streams can be performed on appropriate hardware and / or as functions in software and / or firmware. Although the present invention and its various aspects may involve analog or digital signals, in practical applications most or all of the processing functions tend to be performed in the digital domain in digital signal streams, in which audio signals are represented by samples .

It should be understood that an implementation of other variations and modifications of the invention and its various aspects will be evident to those skilled in the art, and that the invention is not limited by these specific modalities described. Therefore, it is contemplated to cover by the present invention any and all modifications, variations or equivalents that fall in the true spirit and scope of the basic underlying principles shown and claimed here.

Claims (10)

1. Process for translating M audio input signals, each associated with a direction, into N audio output signals, each associated with a direction, where N is greater than M, M is two or more and N is a positive integer number equal to three or more, comprising: providing an M: N variable matrix, applying the M audio input signals to the variable matrix; derive the N audio output signals from the variable matrix, and 10 characterized by the fact that it still comprises controlling the variable matrix in response to the input signals, so that a sound field generated by the output signals has a compact sound image in the nominal primary direction of the input signals, when the input signals are highly correlated, the image diffuses15 from compact to wide, as the correlation decreases and progressively divides into multiple compact sound images, each in a direction associated with an input signal, as the correlation continues to decrease until highly uncorrelated. 2. Process according to claim 1, characterized 20 by the fact that the variable matrix M: N is a variable matrix that has variable coefficients or is a variable matrix that has fixed coefficients and variable outputs, and the variable matrix is controlled by the variation of the variable coefficients or by the variation variable outputs. 3. Process according to claim 1, characterized 25 by the fact that 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. 4. Process according to claim 3, characterized 30 due to the fact that for a cross-correlation measurement of the input signals having values in a first range, limited by a maximum value and a reference value, the sound field has a sound image with Petition 870180005820, from 01/23 / 2018, p. 4/14 pacta, when the cross correlation measure is the maximum value and has a widespread image when the cross correlation measure is the reference value, and for a cross correlation measure of the input signals having values in a second range , limited by the value of 5 reference and for a minimum value, the sound field has a widely spread image when the cross correlation measure is the reference value, and has a plurality of compact sound images, each in a direction associated with an input signal , when the cross-correlation measure is the minimum value. 10 5. Process, according to claim 4, characterized by the fact that the reference value is almost the value of a cross-correlation measure of the input signals for the case of equal energy in each of the outputs. 6. Process according to claim 3, characterized 15 by the fact that a measure of the relative levels of the input signals is in response to an attenuated energy level of each input signal. 7. Process according to claim 3 or 6, characterized by the fact that a measure of the relative levels of the input signals is a primary direction in nominal progress of the input signals. 20 8. Process according to claim 3, characterized by the fact that a measure of the cross correlation of the input signals is in response to a common attenuated energy of the input signals divided by the Mth root of the product of the attenuated energy level of each input signal, where M is the number of inputs. 9. Process according to any one of claims 6 to 8, characterized by the fact that the attenuated energy level of each input signal is obtained by the attenuation of the variable time constant time domain. 10. Process according to any one of claims 6 30 to 8, characterized by the fact that the attenuated energy level of each input signal is obtained by the attenuation of the frequency domain and by the attenuation of the variable time constant time domain. Petition 870180005820, of 23/01/2018, p. 5/14 11. Process according to claim 8, characterized by the fact that the common energy of the input signals is obtained by the vector multiplication of the input amplitude levels. 12. Process according to claim 11, characterized by the fact that the common attenuated energy of the input signals is obtained by attenuating the variable time constant time domain of the common energy of the input signals. 13. Process according to claim 12, characterized by the fact that the attenuated energy level of each input signal is 10 obtained by attenuation of time domain of variable time constant. 14. Process according to claim 11, characterized by the fact that the common attenuated energy of the input signals is obtained by the attenuation of the frequency domain and the attenuation of the variable time constant time domain of the common energy of the signal signals. entry. 15. Process according to claim 14, characterized by the fact that the attenuated energy level of each input signal is obtained by attenuating the frequency domain and attenuating the time domain of the variable time constant. 16. Process according to any of the claims 9, 10, 12, 13, 14 or 15, characterized by the fact that the time domain attenuation of the variable time constant is performed by an attenuation having both a fixed time constant and a variable time constant. 17. Process according to any of the claims 9, 10, 12, 13, 14 or 15, characterized by the fact that the time domain attenuation of variable time constant is performed by an attenuation having only one variable time constant. 18. Process according to claim 16 or 17, characterized by the fact that the variable time constant is variable in stages. 19. Process, according to claim 16 or 17, characterPetition 870180005820, of 23/01/2018, p. 6/14 caused by the fact that the variable time constant is continuously variable. 20. Process according to claim 16 or 17, characterized by the fact that the variable time constant is controlled in response to measurements of the relative levels of the input signals and their cross-correlation. 21. Process according to claim 6, characterized by the fact that the attenuated energy level of each input signal is obtained by a time domain attenuation of the variable time constant of the energy levels of each input signal with substantially the same time constant. 22. Process, according to claim 3, characterized by the fact that the measurements of the relative levels of the input signals and their cross-correlation are each obtained by an attenuation of time domain of variable time constant, in the which the same time constant is applied to each attenuation. 23. Process according to claim 8, characterized by the fact that the cross-correlation measure is a first cross-correlation measure of the input signals and an additional cross-correlation measure is obtained by applying a measure of the relative levels from input signals to the first cross-correlation measure to produce a weighted cross-correlation direction measure. 24. Process according to claim 23, characterized by the fact that an additional measure of cross-correlation of the 25 input is also obtained by applying a scale factor almost equal to a value of a cross correlation measure of the input signals for the case of equal energy in each of the outputs. 25. Method for translating M audio input signals, each associated with a direction, into N audio output signals, each associated with a direction, where N is greater than M, and M is three or more, comprising the steps of: provide a plurality of arrays of variable m: n, where Petition 870180005820, of 23/01/2018, p. 7/14 m is a subset of M and n is a subset of N, apply a respective subset of the M audio input signals to each of the variable arrays, derive a respective subset of the N audio output signals from each one of the variable arrays, characterized by the fact that it still comprises: controlling each of the variable arrays in response to the subset of input signals applied to it, so that a sound field generated by the respective subset of output signals derived from it have a compact sound image in the primary direction in nominal travel of the subset of input signals applied to it, when such input signals are highly correlated, the image diffuses from compact to wide, as the correlation decreases and progressively divides into multiple compact sound images, each in a direction associated with an input signal applied to it, as the correlation continues to decrease until is highly uncorrelated, and derive the N audio output signals from the subsets of N audio output channels. 26. Method according to claim 25, characterized by the fact that the variable matrices are also controlled in response to information that compensates for the effect of one or more other variable matrices receiving the same input signal. 27. Method according to claim 25 or claim 26, characterized by the fact that deriving the N audio output signals from the subsets of N audio output channels includes compensating for the multiple variable arrays producing the same exit sign. 28. Method according to any one of claims 25 to 27, characterized by the fact that each of the variable matrices is controlled in response to measures of: (a) the relative levels of the input signals applied to it, and (b) the cross-correlation of the input signals. 29. Apparatus for translating M audio input signals, each Petition 870180005820, of 23/01/2018, p. 8/14 one associated with a direction, in N audio output signals, each associated with a direction, where N is greater than M, and M is three or more, comprising one or more processors and a memory configured to provide a matrix of variable M: N in response to scale factors that 5 control the matrix coefficients or control the matrix outputs, apply the M audio input signals to the variable matrix, provide a plurality of variable matrix scale factor generators m: n, where m is a subset of M and n is a subset of N, apply a respective subset of the M audio input signals to each of the factor generators 10 of variable matrix scale, derive a set of variable matrix scale factors for the respective subsets of the N audio output signals from each of the variable matrix scale factor generators, characterized by the fact that which further comprises controlling each of the variable matrix scale factor generators in response to the sub15 set of input signals applied to it, so that when the scale factors generated by it are applied to the variable matrix M: N, a sound field generated by the respective subset of output signals produced having a compact sound image in the primary direction in nominal travel of the subset of input signals that produced the factors 20 scale applied, when such input signals are highly correlated, the image diffuses from compact to wide as the correlation decreases and progressively divides into multiple compact sound images, each in a direction associated with an input signal it produced the scale factors applied, as the correlation continues to decrease until highly uncorrelated; and derive the N audio output signals from the variable matrix. 30. Apparatus according to claim 29, characterized by the fact that the variable matrix scale factor generators are also controlled in response to information that compensates for the effect 30 of one or more other variable matrix scale factor generators receiving the same input signal. 31. Apparatus, according to claim 29 or 30, characterPetition 870180005820, of 23/01/2018, p. 9/14 due to the fact that the derivation of the N audio output signals from the variable matrix includes a compensation of the multiple variable matrix scale factor generators producing scale factors for the same output signal. 32. Apparatus according to any of the claims 29 to 31, characterized by the fact that each of the variable matrix scale factor generators is controlled in response to measures of: (a) the relative levels of the input signals applied to it, and (b) the cross-correlation of the input signals. Petition 870180005820, of 23/01/2018, p. 10/14: ··. ·: ·. ··. ···:. ··.: ...... ::. ·· U • · ·· · ·· ···· ······ / / 1 AI: ··:::. :::. :: ··: ·:: v \ IV / λ ................ 17 ' 1/12