EP3044972B1 - Device and method for the decorrelation of loudspeaker signals - Google Patents

Description

The invention relates to an apparatus and method for decorrelating loudspeaker signals by changing the reproduced acoustic scene.

For a three-dimensional listening experience, it may be intended to give the respective listener of an audio piece or viewer of a film a three-dimensional acoustic reproduction a more realistic listening experience by, for example, conveys impressions acoustically, the listener or viewer would be within the reproduced acoustic scene. Psychoacoustic effects can also be used for this. Wavefield Synthesis or Higher Order Ambisonics algorithms are used to generate a particular sound field with a number or plurality of speakers within a playback room. For this purpose, the loudspeakers can be controlled such that the loudspeakers generate wave fields which correspond in whole or in part to acoustic sources which are arranged at a virtually arbitrary location of a reproduced acoustic scene.

Wave Field Synthesis (WFS) or Higher Order Ambisonics (HOA) provides the listener with a high quality spatial listening experience by using a large number of propagation channels to spatially represent virtual acoustic source objects. To provide a more complete user experience, these rendering systems can be supplemented with spatial capture systems to allow for additional applications, such as interactive applications, or to enhance the quality of the playback. The combination of the loudspeaker array, the in-room volume such as a display room and the microphone array is referred to as Speaker Housing Microphone System (LEMS) and identified in many applications by simultaneous observation of the loudspeaker signals and the microphone signals. However, Stereophonic Acoustic Echo Cancellation (AEC) compensation is already known that the typically strong cross-correlation of the loudspeaker signals can prevent adequate system identification, as described for example in [BMS98]. This is called the ambiguity problem. In this case, the result of the system identification is only one of infinitely many solutions determined by the correlation properties of the loudspeaker signals. The result of this incomplete system identification nevertheless describes the behavior of the real LEMS for the instantaneous loudspeaker signals and can therefore be used for various adaptive filtering applications, for example AEC or Listening Room Equalization (LRE). However, this result is no longer correct when the cross-correlation properties of the loudspeaker signals change, which can make the system's behavior based on these adaptive filters unstable. This lack of robustness represents a major hurdle to the applicability of many technologies, such as AEC or adaptive LRE.

For many applications in the field of acoustic reproduction, an identification of a loudspeaker enclosure microphone system (Loudspeaker Enclosure Microphone System), or LEMS may be necessary. With a large number of propagation paths between loudspeakers and microphones, as may be the case for wave-field synthesis (WFS), for example, this problem may be due to the ambiguity problem (i.e. be particularly challenging because of an underdetermined system. If fewer virtual sources are displayed in an acoustic reproduction scene than the loudspeaker system comprises, the ambiguity problem can arise. In such a case, the system can not be uniquely identified, and methods or methods involving system identification suffer from poor or poor robustness to varying correlation characteristics of the loudspeaker signals. A current remedy against the ambiguity problem involves modifying the loudspeaker signals (i.e., a decorrelation) so that the system or LEMS can be uniquely identified and / or increase the robustness under given conditions. However, most known approaches may reduce audio quality or possibly interfere with the synthesized wave field if used in wave-field synthesis.

• Aus [SMH95], [GT98] und [GE98] ist ein Hinzufügen von bezüglich verschiedener Lautsprechersignale unabhängigem Rauschen zu den Lautsprechersignalen vorgeschlagen. In [MHBOI], [BMS98] werden verschiedene nichtlineare Vorverarbeitungen für jeden Wiedergabekanal vorgeschlagen. In [Ali98], [HBK07] werden verschiedene zeitvariante Filterungen für jeden Lautsprecherkanal vorgeschlagen. Obwohl die genannten Techniken die wahrgenommene Klang- oder Schallqualität im Idealfall nicht beeinträchtigen sollten, sind sie im Allgemeinen nicht gut geeignet für WFS: da die Lautsprechersignale für WFS analytisch bestimmt werden, kann eine zeitvariante Filterung das reproduzierte Wellenfeld signifikant stören. Wenn eine hohe Qualität der Audiowiedergabe angestrebt ist, wird ein Hörer möglicherweise eine Hinzufügung von Rauschsignalen oder eine nichtlineare Vorverarbeitung, die beide die Audioqualität reduzieren können, nicht akzeptieren. In [SHK13] wird ein für WFS geeigneter Ansatz vorgeschlagen, bei dem die Lautsprechersignale vorgefiltert werden, so dass eine Veränderung der Lautsprechersignale im Sinne einer zeitvarianten Rotation des wiedergegebenen Wellenfeldes erreicht wird.

For the purpose of decorrelating loudspeaker signals, three ways are known to increase the robustness of the system identification, that is, the identification or estimation of the real LEMS:

• From [SMH95], [GT98] and [GE98], it is proposed to add noise independent of various speaker signals to the speaker signals. In [MHBOI], [BMS98] various non-linear preprocessing is proposed for each playback channel. In [Ali98], [HBK07] different time-variant filters are proposed for each loudspeaker channel. Although these techniques should ideally not compromise the perceived sound or sound quality, they are generally not well suited for WFS: because the loudspeaker signals for WFS are analytically determined, time variant filtering can significantly disturb the reproduced wave field. If high quality audio reproduction is desired, a listener may not accept addition of noise signals or non-linear preprocessing, both of which may reduce audio quality. In [SHK13] a suitable approach for WFS is proposed in which the loudspeaker signals are prefiltered so that a change in the loudspeaker signals in the sense of a time-variant rotation of the reproduced wave field is achieved.

The object of the present invention is therefore to provide an apparatus and a method for generating a plurality of loudspeaker signals, which enables an improved system identification.

This object is solved by the subject matter of the independent patent claims.

The core idea of the present invention is to have recognized that the above object can be achieved in that decorrelated loudspeaker signals can be generated by time-variant modification of meta-information of a virtual source object, such as the position or type of the virtual source object.

According to one embodiment, an apparatus for generating a plurality of loudspeaker signals comprises a modifier configured to time varying modify meta information of a virtual source object. The virtual source object comprises the meta-information and a source signal.

For example, the meta-information determines characteristics such as a position or type of the virtual source object. By modifying the meta-information, for example, the position or type, such as a radiation characteristic, of the virtual source object can be modified. The apparatus further includes a renderer configured to surround the virtual source object and the modified ones Transfer meta information into a variety of loudspeaker signals. By the time-variant modification of the meta-information based on defined rules, which are defined in the further features of claim 1, a decorrelation of the loudspeaker signals can be achieved, so that a stable, ie robust, system identification can be provided, based on the improved system identification a more robust LRE or to enable a more robust AEC, since the robustness of the LRE and / or AEC depends on the robustness of the system identification. A more robust LRE or a more robust AEC can be used for improved sound quality of the speaker signals.

An advantage of this embodiment is that decorrelated loudspeaker signals can be generated by means of the renderer based on the time-varying modified meta-information, so that an additional decorrelation by an additional filtering or an addition of noise signals can be dispensed with.

An alternative embodiment provides a method for generating a plurality of loudspeaker signals based on a source virtual object having a source signal and meta information defining the location or type of the source virtual object. The method comprises time-varying the meta-information by means of defined rules, which are defined in the further features of claim 7, and converting the virtual source object and the modified meta-information into a plurality of loudspeaker signals.

An advantage of this embodiment is that by the modification of the meta information already decorrelated loudspeaker signals are generated, so that compared to a subsequent decorrelation of correlated loudspeaker signals increased reproduction quality of the acoustic playback scene can be achieved because an addition of subsequent noise signals or an application of non-linear operations can be avoided.

Fig. 1

eine Vorrichtung zur Erzeugung einer Mehrzahl von dekorrelierten Lautsprechersignalen basierend auf virtuellen Quellenobjekten;

Fig. 2

eine schematische Aufsicht auf einen Wiedergaberaum, an dem Lautsprecher angeordnet sind;

Fig. 3

eine schematische Übersicht zur Modifikation von Metainformationen verschiedener virtuellen Quellenobjekten;

Fig. 4

eine schematische Anordnung von Lautsprechern und Mikrophonen in einem experimentellen Prototypen;

Fig. 5a

die Ergebnisse erzielbarer Echo Return Loss Enhancement (ERLE) für die Kompensation akustischer Echos (AEC) in vier Plots für vier Quellen mit unterschiedlicher Amplitudenoszillation des Prototypen;

Fig. 5b

den normierten Systemabstand für die Systemidentifikation für die Amplitudenoszillationen;

Fig. 5c

einen Plot an welchem an der Abszisse die Zeit und an der Ordinate die Werte der Amplitudenoszillation angegeben sind;

Fig. 6a

ein Signalmodell zu Identifizierung eines Loudspeaker Enclosure Microphone System (LEMS);

Fig. 6b

ein Signalmodell eines Verfahrens zur Systemschätzung gemäß Fig. 6a und zur Dekorrelation von Lautsprechersignalen;

Fig. 6c

ein Signaimodell einer MIMO Systemidentifikation mit einer Lautsprecherdekorrelation, wie sie in den Fig. 1 und 2 beschrieben ist.

Further advantageous embodiments are the subject of the dependent claims. Preferred embodiments of the present invention will be explained below with reference to the accompanying drawings. Show it:

Fig. 1

a device for generating a plurality of decorrelated loudspeaker signals based on virtual source objects;

Fig. 2

a schematic plan view of a playback room, are arranged on the speaker;

Fig. 3

a schematic overview of the modification of meta-information of various virtual source objects;

Fig. 4

a schematic arrangement of speakers and microphones in an experimental prototype;

Fig. 5a

the results of achievable Echo Return Loss Enhancement (ERLE) for acoustic echo cancellation (AEC) in four plots for four sources with different amplitude oscillation of the prototype;

Fig. 5b

the normalized system distance for the system identification for the amplitude oscillations;

Fig. 5c

a plot at which the abscissa indicates the time and the ordinate the values of the amplitude oscillation;

Fig. 6a

a signal model for identification of a Loudspeaker Enclosure Microphone System (LEMS);

Fig. 6b

a signal model of a method for system estimation according to Fig. 6a and for decorrelating loudspeaker signals;

Fig. 6c

a sign model of a MIMO system identification with a speaker decorrelation, as described in the Fig. 1 and 2 is described.

Before embodiments of the present invention are explained in more detail in detail with reference to the drawings, it is pointed out that identical, functionally identical or equivalent elements, objects and / or structures in the different figures are provided with the same reference numerals, so that shown in different embodiments Description of these elements is interchangeable or can be applied to each other.

Fig. 1 shows a device 10 for generating a plurality of decorrelated speaker signals based on virtual source objects 12a, 12b and / or 12c. A virtual source object can be any type of noise-emitting object, body, or person, such as one or more people, musical instruments, animals, plants, devices, or machines. The virtual source objects 12a-c may be elements of an acoustic playback scene, such as an orchestra performing a performance. For an orchestra, a virtual source object may be, for example, an instrument or a group of instruments. In addition to a source signal, such as a mono signal of a reproduced sound, or a sound or sound sequence of the virtual source object 12a-c, meta information may also be associated with a virtual source object. For example, the meta-information may include a location of the virtual source object within the acoustic playback scene reproduced by a playback system. For example, this may mean a position of a respective instrument within the reproduced orchestra. The meta-information may alternatively or additionally also include a directional or emission characteristic of the respective virtual source object, such as information about the direction in which the respective source signal of the instrument is played. For example, if an instrument of an orchestra is a trumpet, the trumpet sound is preferably radiated in a certain direction (the direction in which the bell is pointed). Alternatively, if the instrument is a guitar, for example, the guitar radiates in a wider viewing angle compared to the trumpet. The meta-information of a virtual source object may include the emission characteristic and the orientation of the emission characteristic in the reproduced reproduction scene. The meta-information may alternatively or additionally also include a spatial extent of the virtual source object in the reproduced playback scene. Based on the meta-information and the source signal, a virtual source object can be described two- or three-dimensionally in space.

A reproduced playback scene can also be, for example, an audio part of a movie, ie the background noise of the movie. For example, a reproduced playback scene may be wholly or partially coincident with a movie scene, such that the virtual source object may be an object positioned in the playback room and directionally speaking, or an object moving in the room of the reproduced scene, such as a train or a car, emitting sounds. can be.

Device 10 is designed to generate loudspeaker signals for driving loudspeakers 14a-e. The speakers 14a-e may be placed on or in a playback room 16. The playback room 16 may be, for example, a concert or cinema hall in which a listener or viewer 17 may be located. By generating and reproducing the speaker signals to the speakers 14a-e, a reproducing scene based on the virtual source objects 12a-c can be reproduced in the reproducing room 16. Apparatus 10 includes a modifier 18 configured to time varying the meta information of one or more of the virtual source objects 12a-c. The modifier 18 is further configured to individually store the metadata of multiple virtual source objects, i. for each virtual source object 12a-c, or for multiple virtual source objects. Modification For example, the modifier 18 is configured to modify the position of the virtual source object 12a-c in the reproduced playback scene or the radiation characteristic of the virtual source object 12a-c.

In other words, application of decorrelation filters may cause an uncontrolled change in the scene being reproduced if speaker signals are decorrelated without regard to the resulting acoustic effects in the playback room, whereas device 10 will be a natural, i. controlled change of the virtual source objects. By a time variant change of the rendered, i. reproduced the acoustic scene by modifying the meta-information such that the position or the radiation characteristic, i. the source type, one or more virtual source objects 12a-c. This can be achieved through access to the playback system, i. by an arrangement of the modifier 18. Modifications of the meta-information of the virtual source objects 12a-c and thus the reproduced acoustic reproduction scene may be intrinsic, i. intrinsically, controlled, so that a limitation of the effects occurring by the modification is possible, for example by the incoming effects of the listener 17 are not perceived or perceived as disturbing.

Apparatus 10 includes a renderer 22 configured to translate the source signals of the virtual source objects 12a-c and the modified meta-information into a plurality of loudspeaker signals. The renderer 22 includes component generators 23a-c and signal component renderers 24a-e. The renderer 22 is designed to use the component generators 23a-c to generate the source signal of the virtual source object 12a-c and the modified meta-information into signal components such that a wave field can be generated by the loudspeakers 14a-e and the wave source field represents the virtual source object 12a-c at a position 25 within the reproduced acoustic reproduction scene. The reproduced acoustic reproduction scene may be at least partially disposed inside or outside of the reproduction room 16. The signal component renderers 24a-e are configured to render the signal components of one or more virtual source objects into speaker signals for driving the speakers 14a-e. On or in a playback room 16, for example, depending on the reproduced playback scene and / or a size of the playback room 16, a plurality of speakers, of, for example, more than 10, 20, 30, 50, 300 or 500 arranged or attachable. In other words, the renderer can be described as a Multiple Input (Mimo) multiple output (loudspeaker signals) MIMO system that converts input signals of one or more virtual source objects into loudspeaker signals. Alternatively, the component generators and / or the signal component processors may be arranged in two or more separate components.

The renderer 22 may alternatively or additionally implement a pre-equalization such that in the reproduction room 16 the reproduced reproduction scene is rendered as if it were reproduced in a free-field environment or other environment such as a concert hall, i. the renderer 22 may partially or completely compensate for distortions of acoustic signals caused by the playback room 16, such as by pre-equalization. In other words, the renderer 22 is designed to generate loudspeaker signals for the virtual source object 12a-c to be displayed.

If a plurality of virtual source objects 12a-c are converted into loudspeaker signals, a loudspeaker 14a-e may at one time reproduce drive signals based on a plurality of virtual source objects 12a-c.

Device 10 comprises microphones 26a-d which may be attached to or in the display room 16 so that the wave fields generated by the loudspeakers 14a-e can be detected by the microphones 26a-d. A system calculator 28 of the apparatus 10 is configured to estimate a transmission characteristic of the playback room 16 based on the microphone signals of the plurality of microphones 26a-d and the loudspeaker signals. A transfer characteristic of the reproduction room 16, ie, a characteristic of how the reproduction room 16 influences the wave fields generated by the loudspeakers 14a-e can be represented, for example, by a varying number of persons residing in the reproduction room 16 by changes in furniture such as a variable scenery of the reproduction room 16 or by caused a variable position of persons or objects within the playback room 16. For example, by an increasing number of persons or objects in the playback room 16, reflection paths between speakers 14a-e and microphones 26a-d may be blocked or generated. The estimation of the transfer characteristic can also be represented as system identification. If the loudspeaker signals are correlated, the ambiguity problem can occur during system identification.

The renderer 22 may be configured to implement a time-varying rendering system based on the time-varying transmission characteristic of the rendering room 16 such that a changed transmission characteristic can be compensated and a reduction in audio quality can be avoided. In other words, the renderer 22 may enable adaptive equalization of the playback room 16. Alternatively or additionally, the renderer 22 may be configured to superimpose the generated loudspeaker signals with noise signals to add attenuation to the loudspeaker signals and / or to delay the loudspeaker signals by, for example, filtering the loudspeaker signals using a decorrelation filter. A decorrelation filter can, for example, be used for a time-variant phase shift of the loudspeaker signals. By decorrelation filtering and / or the addition of noise signals, additional decorrelation of the loudspeaker signals can be achieved, for example, if meta-information on a virtual source object 12a-c is modified only slightly by the modifier 18, so that the loudspeaker signals generated by the renderer 22 are to an extent are correlated, which is to be reduced for a playback scene.

By modifying the meta-information of the virtual source objects 12a-c by means of the modifier 18, a decorrelation of the loudspeaker signals and thus a reduction or avoidance of system instabilities can be achieved. A system identification can be improved, for example, by taking advantage of a change, ie modification of the spatial properties of the virtual source objects 12a-c.

In contrast to a change in the loudspeaker signals, the modification of the meta-information can be targeted and, according to psychoacoustic criteria, carried out in such a way that the listener 17 of the reproduced reproduction scene does not perceive the modification or does not find it disturbing. Thus, a shift of the position 25 of a virtual source object 12a-c in the reproduced playback scene may result in changed loudspeaker signals and thus in a partial decorrelation of the loudspeaker signals, thus adding noise signals or applying nonlinear filtering operations such as in decorrelation filters , can be avoided. For example, if a train is displayed in the reproduced playback scene, it may, for example, go unnoticed by the listener 17 if the corresponding train is at a great distance from the listener 17, such as 200, 500 or 1000 m, for example 1, 2 or 5 m in the room is shifted.

Multi-channel reproduction systems, such as WFS, as suggested in [BDV93], Higher-Order Ambisonics (HOA), as proposed in [Dan03], for example, or similar methods can provide wave fields with multiple virtual sources or source objects, inter alia, by displaying the reproduce virtual source objects in the form of point sources, dipole sources, kidney-shaped radiation sources, or plane wave emitting sources. If these sources have stationary spatial characteristics, such as fixed positions of the virtual source objects or fixed radiation or directional characteristics, a constant acoustic reproduction scene can be identified if a corresponding correlation matrix has full rank, as shown in FIG Fig. 6 is explained in detail.

Device 10 is configured to generate a decorrelation of the loudspeaker signals by a modification of the metadata of the virtual source objects 12a-c and / or to take into account a time-varying transmission characteristic of the playback room 16.

The device represents a time variant variation of the reproduced acoustic reproduction scene for WFS, HOA or similar reproduction models to decorrelate the loudspeaker signals. Such a decorrelation can be a remedy if the problem of system identification is underdetermined. In contrast to prior art solutions, device 10 allows a controlled change of the reproduced playback scene to obtain a high quality of WFS or HOA reproduction.

Fig. 2 shows a schematic plan view of a playback room 16, are arranged on the speaker 14a-h. Device 10 is configured to generate loudspeaker signals based on one or more virtual source objects 12a and / or 12b. Perceptible modification of the metadata of the virtual source objects 12a and / or 12b may be distracting to the listener. If, for example, a location or a position of the virtual source object 12a and / or 12b is changed too much, the listener may, for example, have the impression that an instrument of an orchestra is moving in space. Alternatively, if the reproduced playback scene belongs to a movie, the acoustic impression may arise that the virtual source object 12a and / or 12b differs at an acoustic velocity different from an optical speed of an object implied by the image sequence virtual source object, for example, moves at different speeds or in a different direction. By changing the meta-information of a virtual source object 12a and / or 12b within certain intervals or tolerances, a perceivable or annoying impression can be reduced or prevented.

For a perception of acoustic scenes, a spatial hearing in a median plane, that is in a horizontal plane of the earpiece 17, may be significant, whereas a spatial hearing in the sagittal plane, ie a left and right body half of the listener 17 center separating plane, play a minor role. For playback systems designed to render three-dimensional scenes, the playback scene may additionally be changed in the third dimension. A localization of acoustic sources by the listener 17 may be more inaccurate in the sagittal plane than in the median plane. It is conceivable to maintain or extend the third dimensional limits defined below for two dimensions (horizontal plane), since limits derived from a two-dimensional wave field represent very conservative lower limits for possible changes of the rendered scene in the third dimension. Although the following explanations are focused on perceptual effects in two-dimensional rendering scenes in the median plane, which are an optimization criterion for many rendering systems, the explanations also apply to three-dimensional systems.

In principle, different types of wave fields can be reproduced, such as wave fields of point sources, plane waves or wave fields of general multipole sources, such as dipoles. In a two-dimensional plane, i. considering only two dimensions, the perceived position of a point source or a multipole source is describable by a direction and a distance, whereas plane waves are describable by an incident direction. The listener 17 can locate the direction of a sound source by two spatial triggering stimuli, interaural level differences (ILDs) and interaural time differences (ITDs). The modification of the meta-information of a respective virtual source object may lead to a change of the respective ILDs and / or to a change of the respective ITDs for the listener 17.

The removal of a sound source can already be perceived by the absolute monaural level, as described in [Bla97]. In other words, the distance can be perceived by a volume and / or a distance change by a volume change.

The interaural level difference describes a level difference between the two ears of the listener 17. An ear facing a sound source may be exposed to a higher sound pressure level than an ear facing away from the sound source. If the earpiece 17 turns its head until both ears are exposed to approximately the same sound pressure level and the interaural level difference is only slight, then the listener can face the sound source or alternatively be positioned with his back to the sound source. For example, modifying the meta-information of the virtual source object 12a or 12b such that the virtual source object is displayed at a different location or has a different directional characteristic can result in a different change in the respective sound pressure levels at the ears of the listener 17 and hence in a change in the interaural Level difference lead, this change may be perceptible to the listener 17.

Interaural time differences may result from different transit times between a sound source and an ear of a listener 17 located at a shorter distance or at a greater distance such that a sound wave emitted by the sound source requires a longer time to the ear further away. A modification of the metadata of the virtual source object 12a or 12b, for example, so that the virtual source object is displayed at a different location, can lead to a different change in the distances between the virtual source object and both ears of the listener 17 and thus to a change in the interaural time difference, this change for the listener 17 can be perceived.

An imperceptible or non-annoying change in the ILD may be between 0.6 dB and 2 dB, depending on the scenario being reproduced. A 0.6 dB ILD variation corresponds to a decrease in the ILD of approximately 6.6% or an increase of approximately 7.2%. A 1 dB change in ILD corresponds to a percentage increase in ILD of approximately 12% and a percentage decrease of 11%, respectively. An increase in ILD by 2 dB corresponds to a percentage increase in ILD of approximately 26%, whereas a decrease of 2 dB corresponds to a percentage decrease of 21%. A perception threshold for an ITD may be dependent on a particular scenario of the acoustic playback scene, such as 10, 20, 30, or 40 μs. By modifying the meta-information of the virtual source object 12a or 12b may be low, i. In the range of a few 0.1 dB, changed ILDs, a change in the ITDs may possibly be perceived earlier by the handset 17 or be perceived as disturbing than a change in the ILD.

The modification of the meta-information may only slightly affect the ILDs if the distance from one sound source to the listener 17 is slightly shifted. ITDs may present a more stringent constraint for inaudible or non-disturbed alteration of the reproduced playback scene due to the earlier perceptibility and linear change in position change. If ITDs of 30 μs are permitted, this can lead to a maximum change of a source direction between the sound source and the listener 17 of up to α ₁ = 3 ° for frontally, ie in a viewing direction 32 or a frontal area 34a, 34b of the listener 17, arranged sound sources and / or a change of up to α ₂ = 10 ° for laterally, ie laterally arranged sound sources. A laterally disposed sound source may be located in one of the side regions 36a or 36b extending between the frontal regions 34a and 34b. The frontal areas 34a and 34b may be defined, for example, such that at an angle of ± 45 ° with respect to the viewing direction 32, the frontal area 34a of the earpiece 17 and ± 45 ° counter to the viewing direction, the frontal area 34b extends, so that the frontal area 34b in the back of the listener can be arranged. Alternatively or additionally, the frontal regions 34a and 34b may also comprise a smaller or larger angle or comprise different angular ranges from one another, such that, for example, the frontal region 34a comprises a larger angular range than the frontal region 34b. In principle, frontal regions 34a and 34b and / or side regions 36a and 36b can be arranged independently of one another or spaced from one another. The viewing direction 32 can, for example, be seated on or in which the handset 14 is seated or by a chair in which the handset 17 looks at a screen.

In other words, device 10 may be designed to take into account the viewing direction 32 of the earpiece 17, such that frontally arranged sound sources such as the virtual source object 12a are up to α ₁ = 3 ° and laterally arranged sound sources such as the virtual source object 12b up to α ₂ = 10 ° be modified with respect to their direction. As opposed to a system as proposed in [SHK13], apparatus 10 may enable a displacement of a source object individually with respect to the virtual source objects 12a and 12b, whereas in [SHK13], only the reproduced playback scene as a whole can be rotated. In other words, a system as described, for example, [SHK13] has no information about the rendered scene but takes into account information about the generated speaker signals. Device 10 changes the rendered scene known to device 10.

While changes in the reproduced playback scene by changing the source direction by 3 ° or 10 ° may not be perceptible to the listener 17, it is also conceivable to accept perceptible changes in the reproduced playback scene that may be perceived as non-irritating. For example, a change of the ITD of up to 40 μs or 45 μs can be permitted. In addition, for example, a rotation of the entire acoustic scene by up to 23 ° can be perceived by many or most listeners as not disturbing [SHK13]. This limit can be increased by a few to several degrees by independently modifying the individual sources or directions from which the sources are perceived, so that its displacement of the acoustic playback scene by up to 28 °, 30 ° or 32 ° can be possible.

The distance 38 of an acoustic source, such as a virtual source object, may possibly be inaccurately perceived by a listener. Experiments show that a variation of the distance 38 of up to 25% for listeners is usually not perceived or bothersome, which allows a rather strong variation of the source distance, as described for example in [Bla97] is.

A period between changes in the reproduced playback scene may have a constant or variable interval between individual changes, such as 5 seconds, 10 seconds, or 15 seconds, to ensure high audio quality. The high audio quality can be achieved by allowing an interval of about 10 seconds between scene changes or changes in meta-information of one or more virtual source objects, a sufficiently high decorrelation of the speaker signals and the rarity of the changes or modifications contributes to changes in the playback scene not perceptible or disturbing.

Variation or modification of the radiation characteristics of a general multipole source can leave the ITDs unaffected, whereas the ILDs can be affected. This may allow for any modifications to the radiation characteristics that are unnoticed or unnoticed by a listener 17 as long as the ILDs at the listener location are less than or equal to the respective threshold (0.6 dB to 2 dB).

The same limits may be used for a monaural level change, i. with respect to an ear of the earpiece 17.

Device 10 is configured to overlay an original virtual source object 12a, with an additional mapped virtual source 12'a that emits the same or a similar source signal. In other words, the modifier 18 is configured to create an image of the virtual source object (12a). The imaged virtual source 12'a can be arranged approximately at a virtual position P ₁ at which the virtual source object 12a is originally arranged. The virtual position P ₁ has a distance 38 to the handset 17. In other words, the additional mapped virtual source 12'a may be an imaged version of the virtual source object 12a created by the modifier 18 such that the mapped virtual source 12'a is the virtual source object 12. In other words, the virtual source object 12a may have been imaged by the modifier 18 into the mapped virtual source object 12'a. The virtual source object 12a by the modification of the meta information for example. At a virtual position P ₂ at a distance 42 to the mapped virtual source object 12'a and a distance 38 'are moved to the handset 17. Alternatively or additionally, it is conceivable that the modifier 18 modifies the meta-information of the image 12'a.

A region 43 can be represented as a partial area of a circle with a distance 41 around the imaged virtual source object 12'a, which has a distance of at least the distance 38 from the receiver 17. If the distance 38 'between the modified virtual source object 12a is greater than the distance 38 between the imaged virtual source 12'a, so that the modified source object 12a is located within the area 43, the virtual source object 12a may be in the area 43 around that shown virtual source object 12'a without the imaged virtual source object 12'a and the virtual source object 12 being perceived as separate acoustic objects. The region 43 can extend up to 5, 10 or 15 m around the imaged virtual source object 12'a and be bounded by a circle with the radius R ₁ corresponding to the distance 38.

Alternatively or additionally, device 10 may be configured to take advantage of the precedence effect, also known as the Haas effect, as described in [Bla97]. According to a Haas observation, an acoustic reflection of a sound source arriving at the listener 17 up to 50 ms after the direct, for example unreflected, portion of the sound can be almost perfectly recorded in the spatial perception of the original source. That is, two separate acoustic sources are perceptible as one.

Fig. 3 shows a schematic overview of the modification of meta-information of various virtual source objects 121-125 in a device 30 for generating a plurality of decorrelated speaker signals. Even though Fig. 3 and the explanations for a clear representation are kept two-dimensional, all examples also apply to the three-dimensional case.

The virtual source object 121 is a spatially limited source, such as a point source. The meta-information of the virtual source object 121 can be modified, for example, such that the virtual source object 121 is moved on a circular path over a plurality of interval steps.

The virtual source object 122 is also a spatially limited source such as a point source. A change in the metadata of the virtual source object 122 may, for example, take place such that the point source is moved irregularly in a limited area or volume over a plurality of interval steps. The wave field of the virtual source objects 121 and 122 may be modified in general by modifying the meta information so that the position of the respective virtual source object 121 or 122 is modified. In principle, this is possible for any virtual source object with a limited spatial extent, such as a dipole or a source with a kidney-shaped radiation characteristic.

The virtual source object 123, representing a planar sound source, may be varied with respect to the excited plane wave. By modifying the meta-information, an emission angle of the virtual source object 123 and / or an angle of incidence on the receiver 17 can be influenced.

The virtual source object 124 is a virtual source object having a limited spatial extent, such as a dipole source having a directional radiation characteristic, as indicated by the circles. For changing or modifying the meta-information of the virtual source object 124, the direction-dependent emission characteristic can be rotated.

For directional virtual source objects, such as the virtual source object 125 having a kidney-shaped radiation characteristic, the meta-information may be modified so that the radiation pattern is modified depending on the particular time. For the virtual source object 125, this is exemplified by a change from a kidney-shaped radiation characteristic (solid line) to a hypercardioid directional characteristic (dashed line). For omnidirectional virtual source objects or sound sources, an additional, time-variant direction-dependent directional characteristic can be added or generated.

The various possibilities, such as changing the position of a virtual source object such as a point source or source with limited spatial extent, a change in the angle of incidence of a plane wave, a change of the radiation pattern, a rotation of the radiation characteristic or adding a directional directional characteristic to an omnidirectional radiating Source object, can be combined with each other. Here, the parameters which are selected or determined to be modified for the respective source object may be any and different. Further, the manner of changing the spatial characteristics as well as a speed of change may be chosen such that the change of the reproduced scene of reproduction either goes unnoticed by a listener or is acceptable in the perception by the listener. In addition, the spatial characteristics for temporally individual frequency ranges can be varied differently.

The following is based on Fig. 4 with reference to Fig. 5c and Fig. 6c one of a variety of possible structures for verification of the findings of the invention described. Fig. 5c shows an exemplary course of an amplitude oscillation of a virtual source object over time. In the Fig. 6c For example, a signal model of decorrelated speaker signal generation by a modification of the acoustic reproducing scene will be explained. It is a prototype to represent the effects. The prototype is, for example, constructed experimentally with regard to the loudspeakers and / or microphones used, the dimensions and / or distances between components.

Fig. 4 shows a schematic arrangement of speakers and microphones in an experimental prototype. An exemplary number of N _L = 48 loudspeakers is arranged in a loudspeaker system 14S. The loudspeakers are arranged equidistantly on a circular line with a radius of, for example, 1.5 m, so that an exemplary angular distance of 2π / 48 = 7.5 ° results. An exemplary number of N _M = 10 microphones is arranged equidistantly in a microphone system 26S on a circular line with a radius R _M of, for example, 0.05 m, so that the microphones can have an angle of 36 ° to one another. For test purposes, the setup is arranged in a room (enclosure of the LEMS) with a reverberation time T ₆₀ of about 0.3 seconds. The impulse responses can be measured at a sampling frequency of 44.1 kHz, converted to a sampling rate of 11025 Hz and cut to a length of 1024 measurement points, which is the length of the adaptive filters for the AEC. The LEMS is simulated by convolution of received impulse responses without noise on the microphone signal (near-end noise) or local sound sources within the LEMS. These ideal laboratory conditions are selected to separate the influence of the proposed method on the convergence of the adaptation algorithm from other influences. Further experiments, for example with modeled near-end noise can lead to equivalent results.

The signal model is in FIG. 6c explained. There, the decorrelated loudspeaker signals x '(k) are input to the LEMS H, which can then be identified by a transfer function H _est (n) based on the observations of the decorrelated loudspeaker signals x ' (k) and the resulting microphone signals d (k). The error signals e (k) can detect reflections from speaker signals on the enclosure, such as the remaining echo. For the AEC, a generalized adaptive filter algorithm in the frequency domain with an exponential memory factor λ = 0.95, a step size μ = 0.5 (with 0 ≤ μ ≤ 1) and a frame shift of L _F = 512 can be used, as described in [SHK13 ], [BBK03] is proposed to be applied.

berechnet werden, wobei ∥·∥_F die Frobenius-Norm bezeichnet und N der Blockzeitindex ist. Ein geringer Wert des Systemabstandes bezeichnet eine Systemidentifikation (Schätzung) mit einer geringen Abweichung zum realen System.A measure of the achieved system identification is referred to as Normalized Misalignment (NMA) and can be determined by the calculation rule $${\mathrm{\Delta }}_H\left(n\right)=20{\mathrm{<figure-callout\; id="10"\; label="log"\; filenames="imgf0001.png,imgf0002.png"\; state="\{\{state\}\}">log</figure-callout>}}_{10}\left(\frac{{\Vert {\mathbf{H}}_{\mathrm{est}}\left(n\right)-\mathbf{H}\Vert }_{\mathrm{F}}}{{\Vert \mathbf{H}\Vert }_{\mathrm{F}}}\right).$$

where ∥ · ∥ _F denotes the Frobenius norm and N is the block time index. A small value of the system distance denotes a system identification (estimation) with a small deviation from the real system.

The relation between n and k can be given by n = floor (k / L _F ), where floor (·) is the "floor" operator or the Gaussian bracket, ie the quotient is rounded off. In addition, an achieved echo suppression can be considered, which can be described, for example, by means of the Echo Return Loss Enhancement (ERLE), in order to allow better comparability with [SHK13].

wobei ∥·∥₂ die Euklidische Norm beschreibt.The ERLE is defined as $$\mathrm{ALDER}\left(k\right)=20{\mathrm{<figure-callout\; id="10"\; label="log"\; filenames="imgf0001.png,imgf0002.png"\; state="\{\{state\}\}">log</figure-callout>}}_{10}\left(\frac{{\Vert \mathrm{d}\left(k\right)\Vert }_2}{{\Vert \mathrm{e}\left(k\right)\Vert }_2}\right).$$

where ∥ · ∥ _{2 describes} the Euclidean norm.

beschrieben werden, wobei ϕ_a die Amplitude der Einfallswinkeloszillation und L_P die Periodendauer der Einfallswinkeloszillation ist, wie sie exemplarisch in Fig. 5c veranschaulicht wird. Für die Quellensignale wurden untereinander unkorrelierte Signale weißen Rauschens verwendet, so dass alle 48 Lautsprecher mit einer gleichen durchschnittlichen Leistung betrieben werden können.In a first experiment, the loudspeaker signals are determined according to the theory of wave field synthesis, as proposed for example in [BDV93], in order to synthesize four plane waves simultaneously with angles of incidence varying by α _q . α _q is given by 0, π / 2, π and 3 π / 2 for the sources q = 1,2, ...., N _S = 4. The resulting time-variant angles of incidence can by $${\phi }_q\left(n\right)={\alpha }_q+{\phi }_{\mathrm{a}}\cdot \mathrm{<figure-callout\; id="2"\; label="sin"\; filenames="imgf0005.png"\; state="\{\{state\}\}">sin</figure-callout>}\left(2\pi \frac{n}{L_P}\right).$$

where φ _{a is} the amplitude of the incident angle oscillation and L _{P is} the period of the incidence angle oscillation, as exemplified in FIG Fig. 5c is illustrated. For the source signals, white noise uncorrelated signals were used among each other, so that all 48 speakers can be operated with the same average power.

Although noise signals used to drive loudspeakers may not be relevant in practice, this scenario may allow a clear and concise assessment of the influence of φ _a . In view of the fact that, by way of example, only four independent signal sources (N _S = 4) and 48 loudspeakers (N _L = 48) are arranged or used, the task and the system of equations of the system identification are massively underdetermined, so that a high normalized system distance ( NMA) can be expected.

The prototype can achieve results of the NMA that can surpass the state of the art and can thus lead to a better acoustic reproduction of WFS or HOA.

In the following Fig. 5 The results of the experiment are shown graphically.

Fig. 5a shows the ERLE for the four sources of the prototype. Plot 1 shows: φ _a = π / 48, plot 2: φ _a = 4π / 48, plot 3: φ _a = 8π / 48 and plot 4: φ _a = 0. For plot 4 and therefore for φ _a = 0, the ERLE can be achieved up to about 58 dB.

Fig. 5b shows the achieved normalized system spacing with the identical values for φ _a in plots 1 to 4. The system spacing can reach values of up to about -16 dB, which compares to values of -6 dB achieved in [SHK13] a significant improvement in the system description of the LEMS.

Fig. 5c shows a plot at which the time on the abscissa and the values of the amplitude oscillation φ _a are indicated on the ordinate, so that the period L _{P can be} read.

The improvement over [SHK13] of up to 10 dB in terms of normalized system spacing can be explained, at least in part, by the approach proposed in [SHK13] using spatially bandlimited loudspeaker signals. The spatial bandwidth of a natural acoustic scene is generally too large for the scene to be reproduced perfectly by the (to a limited extent) provided loudspeaker signals and loudspeakers, ie without deviations. Artificial, ie controlled, band limitation, such as in HOA, allows a spatially band limited scene to be obtained. In alternative methods, such as WFS, an occurrence of aliasing effects can be tolerated to obtain a band-limited scene. Devices like those in the Fig. 1 and 2 are proposed, can work with a spatially unstable or hardly bandlimited virtual playback scene. In [SHK13], WFS aliasing artifacts already created or captured in the speaker signals are simply rotated with the reproduced playback scene so that aliasing effects can persist between the virtual source objects. In the Fig. 5 and 6 For example, the portions of the individual WFS aliasing spawn in the loudspeaker signals may vary with a rotation of the virtual playback scene by an individual modification of the metadata of individual source objects. This can lead to a stronger decorrelation. The Fig. 5a-c show that the system identification can be improved with a larger rotation amplitude φ _{a of} a virtual source object of the acoustic scene, as shown in Plot 3 of the Fig. 5b A reduction of the NMA may possibly be achieved at the expense of reduced echo cancellation, as is plots 1-3 in FIG Fig. 5a opposite plot 4 (without rotation amplitude). However, the echo cancellation for decorrelated loudspeaker signals (φ _a > 0) improves over time, whereas the system identification for unchanged loudspeaker signals (φ _a = 0) does not do this.

Below are in the Fig. 6a-c various types of system identification described. In Fig. 6a describes a signal model of a system identification of a multiple input multiple output (MIMO) system, where the ambiguity problem can occur. In Fig. 6b For example, a signal model of a MIMO system identification with a decorrelation of the loudspeaker signals according to the prior art will be described. Fig. 6c shows a signal model of a MIMO system identification with a decorrelation of loudspeaker signals, such as with a device of the Fig. 1 or the Fig. 2 is achievable.

$${\mathrm{x}}_l\left(k\right)={\left(x_l\left(k-L_X+1\right),x_l\left(k-L_X+2\right),\dots ,x_l\left(k\right)\right)}^T;$$

wobei L_x die Länge der individuellen Komponentenvektoren x _l(k) beschreibt, welche die Abtastungen x _l(k) des Lautsprechersignals l zum Zeitpunkt k erfassen. Gleichermaßen können die Vektoren, die die erfassten Mikrophonsignale L_D beschreiben, als Aufnahmen zu bestimmten Zeitpunkten für jeden Kanal und definiert sein als $$\mathrm{d}\left(k\right)={\left({\mathrm{d}}_1\left(k\right),{\mathrm{d}}_2\left(k\right),\dots ,{\mathrm{d}}_{N_M}\left(k\right)\right)}^T\mathrm{.}$$

$${\mathrm{d}}_m\left(k\right)={\left(d_m\left(k-L_D+1\right),d_m\left(k-L_D+2\right),\mathrm{...},d_m\left(k\right)\right)}^T\mathrm{.}$$

In Fig. 6a For example, the LEMS H is estimated by H _est (n), where H _est (n) is estimated by observing the loudspeaker signals x (k) and the microphone signals d (k). H _est (n) may, for example, be a possible solution of an underdetermined system of equations. The vectors that capture the loudspeaker signals are defined by $$\mathrm{x}\left(k\right)={\left({\mathrm{x}}_1\left(k\right).{\mathrm{x}}_2\left(k\right).....{\mathrm{x}}_{N_L}\left(k\right)\right)}^T.$$

$${\mathrm{x}}_l\left(k\right)={\left(x_l\left(k-L_X+1\right).x_l\left(k-L_X+2\right).....x_l\left(k\right)\right)}^T;$$

where L _{x describes} the length of the individual component vectors x ₁ (k) which detect the samples x ₁ (k) of the loudspeaker signal 1 at time k. Likewise, the vectors describing the acquired microphone signals L _{D may} be defined as recordings at particular times for each channel and $$\mathrm{d}\left(k\right)={\left({\mathrm{d}}_1\left(k\right).{\mathrm{d}}_2\left(k\right).....{\mathrm{d}}_{N_M}\left(k\right)\right)}^T\mathrm{,}$$

$${\mathrm{d}}_m\left(k\right)={\left(d_m\left(k-L_D+1\right).d_m\left(k-L_D+2\right).\mathrm{...}.d_m\left(k\right)\right)}^T\mathrm{,}$$

wobei die individuellen Aufnahmen der Mikrophonsignale durch $$d_m\left(k\right)={\displaystyle \sum _{l=1}^{N_L}}{\displaystyle \sum _{\kappa =0}^{L_H-1}}{x}_l\left(k-\kappa \right)h_{m,l}\left(\kappa \right)\mathrm{.}$$

erhalten werden können. Die Impulsantworten h_m,l(k) des LEMS mit der Länge L_H können das zu identifizierende LEMS beschreiben. Um die individuellen Aufnahmen der Mikrophonsignale durch die lineare MIMO Filterung auszudrücken, kann die Beziehung von L_X und L_D mit L_X = L_D + L_H - 1 definiert werden. Die Lautsprechersignale x(k) können durch ein Wiedergabesystem basierend auf WFS, Higher-Order Ambisonics oder einem ähnlichen Verfahren erhalten werden. Das Wiedergabesystem kann eine bspw. lineare MIMO Filterung einer Anzahl von N_S virtuellen Quellensignalen s̊ (k). Die virtuellen Quellensignale s̊ (k) können durch den Vektor

The LEMS can then be described by a linear MIMO filtering, which can be expressed as: $$\mathrm{d}\left(k\right)=\mathrm{Hx}\left(k\right).$$

wherein the individual recordings of the microphone signals by $$d_m\left(k\right)={\displaystyle \underset{l=1}{\overset{N_L}{\Sigma }}}{\displaystyle \underset{\kappa =0}{\overset{L_H-1}{\Sigma }}}{x}_l\left(k-\kappa \right)H_{m.l}\left(\kappa \right)\mathrm{,}$$

can be obtained. The impulse responses h _{m, l} (k) of the LEMS of length L _H can describe the LEMS to be identified. To express the individual recordings of the microphone signals by linear MIMO filtering, the relationship of L _X and L _D with L _X = L _D + L _H - 1 can be defined. The speaker signals x (k) can be obtained by a reproducing system based on WFS, higher order ambisonics or the like. The rendering system may include, for example, linear MIMO filtering of a number of N _S virtual source signals s̊ (k). The virtual source signals s̊ (k) can be passed through the vector

die Faltung der Quellensignale s̊_q(k) mit der Impulsantwort g_l,q(k) beschreibt. Dies kann genutzt werden, um die Lautsprechersignale x_l(k) aus den Quellensignalen s̊ _q(k) gemäß der Berechnungsvorschrift

zu beschreiben. Die Impulsantworten g_l,q(k) haben bspw. eine Länge von L_R Abtastungen und repräsentieren R(l,q,ω) im diskreten Zeitbereich.For example, where L _{s is,} for example, a length of the signal segment of the individual component s̊ _q (k) and s̊ _q (k) is the result of sampling the source q at time k. A matrix G can represent the rendering system and be structured such that

describes the convolution of the source signals s̊ _q (k) with the impulse response g _{l, q} (k). This can be used to derive the loudspeaker signals x _l (k) from the source signals s _q (k) according to the calculation rule

to describe. The impulse responses g _{l, q} (k) have, for example, a length of L _R samples and represent R (1, q, ω) in the discrete time domain.

bestimmbar sein kann und bezüglich einer entsprechenden Norm, wie etwa der Euklidischen oder einer geometrischen Norm, minimiert wird. Wird die Euklidische Norm ausgewählt, können die bekannten Wiener-Hopf Gleichungen resultieren. Werden lediglich Finite Impulse Response (FIR)-Filter für die Systemantworten betrachtet, können die Wiener-Hopf Gleichungen in Matrixnotation in der Form $${\mathrm{R}}_{\mathit{xx}}{\mathrm{H}}_{\mathrm{est}}^H\left(n\right)={\mathrm{R}}_{\mathit{xd}}$$

mit $${\mathrm{R}}_{\mathit{xd}}=\epsilon \left\{\mathrm{x}\left(k\right){\mathrm{d}}^H\left(k\right)\right\}$$

geschrieben bzw. dargestellt werden, wobei R _xd bspw. die Korrelationsmatrix der Lautsprecher- und Mikrophonsignale ist. H _est(n) kann nur eindeutig sein, wenn die Korrelationsmatrix R _xx der Lautsprechersignale vollen Rang hat. Für R _xx kann die folgende Relation erhalten werden: $${\mathrm{R}}_{\mathit{xx}}=\epsilon \left(\mathrm{x}\left(k\right){\mathrm{x}}^H\left(k\right)\right)={\mathrm{GR}}_{\mathit{ss}}{\mathrm{G}}^H,$$

wobei R _ss bspw. die Korrelationsmatrix der Quellensignale gemäß

ist. Daraus kann L_S = L_X + L_R - 1 folgen, so dass R _ss die Dimension N_S(L_X + L_R - 1) x N_S(L_X + L_R - 1) hat, während R _xx die Dimension N_LL_X x N_LL_X hat. Eine notwendige Bedingung dafür, dass R _xx vollen Rang hat, ist $$N_L{L}_X\le N_S\left(L_X+L_R-1\right),$$

wobei die virtuellen Quellen zumindest unkorrelierte Signale tragen und an verschiedenen Positionen positioniert sind.The LEMS can be identified such that an error e (k) of the system estimate H _est (n) $$\mathrm{e}\left(k\right)=\mathrm{d}\left(\mathit{k}\right)-{\mathrm{H}}_{\mathrm{est}}\left(\mathit{n}\right)\mathrm{x}\left(\mathit{k}\right)$$

can be determinable and minimized with respect to a corresponding standard, such as the Euclidean or a geometric standard. If the Euclidean norm is chosen, the well-known Wiener-Hopf equations can result. If only finite impulse response (FIR) filters are considered for the system responses, the Wiener-Hopf equations can be expressed in matrix notation in the form $${\mathrm{R}}_{\mathit{xx}}{\mathrm{H}}_{\mathrm{est}}^H\left(n\right)={\mathrm{R}}_{\mathit{xd}}$$

With $${\mathrm{R}}_{\mathit{xd}}=\epsilon \left\{\mathrm{x}\left(k\right){\mathrm{d}}^H\left(k\right)\right\}$$

For example, R _xd, for example, is the correlation matrix of the speaker and microphone signals. H _est (n) can only be unique if the correlation matrix R _{xx of} the loudspeaker signals has full rank. For R _xx the following relation can be obtained: $${\mathrm{R}}_{\mathit{xx}}=\epsilon \left(\mathrm{x}\left(k\right){\mathrm{x}}^H\left(k\right)\right)={\mathrm{GR}}_{\mathit{ss}}{\mathrm{G}}^H.$$

where R _ss, for example, the correlation matrix of the source signals according to

is. From this L _S = L _X + L _R - 1 can follow, so that R _{ss has} the dimension N _S (L _X + L _R - 1) x N _S (L _X + L _R - 1), while R _xx has the dimension N x N _{X L} L _L L _X. A necessary condition for R _{xx to have} full rank is $$N_L{L}_X\le N_S\left(L_X+L_R-1\right).$$

wherein the virtual sources carry at least uncorrelated signals and are positioned at different positions.

If the number of speakers N _L exceeds the number of virtual sources N _S , the ambiguity problem may arise. In the following consideration, the influence of the impulse response lengths L _X and L _{R is} neglected.

The ambiguity problem can result, at least in part, from the strong mutual cross-correlation of the loudspeaker signals, which may be due, inter alia, to the smaller number of virtual sources. Occurrence of the ambiguity problem may be more likely the more channels are used for the rendering system, inter alia if the number of virtual source objects is less than the number of speakers used in the LEMS. Auxiliary solutions according to the prior art aim at a change of the loudspeaker signals, so that the rank of R _{xx is} increased or the condition number of R _{xx is} improved.

Fig. 6b shows a signal model of a method for system estimation and decorrelation of loudspeaker signals. Correlated loudspeaker signals x (k) can be converted, for example, by decorrelation filters and / or noise-based approaches into decorrelated loudspeaker signals x '(k). The two approaches can be used together or separately. A block 44 (Decorr Fig. 6b describes a filtering of the loudspeaker signals x _l (k), which differs for each loudspeaker with index I and may be non-linear, as described, for example, in [MHB01, BMS98]. Alternatively, the filtering may be linear, but time varying, as suggested, for example, in [SHK13, Ali98, HBK07, WWJ12]. The noise-based approaches proposed in [SMH95, GT98, GE98] can be represented by an addition of uncorrelated noise, indicated by n (k). These approaches have in common that they neglect or leave unchanged the virtual source signals s (k) and the rendering system G. They only process the loudspeaker signals x (k).

gegeben. Diese Bedingung gilt unabhängig von den tatsächlichen räumlichen Eigenschaften, wie physikalische Abmessungen oder Abstrahlcharakteristik der virtuellen Quellenobjekte. Die jeweiligen virtuellen Quellenobjekte sind dabei an von einander verschieden Positionen in dem jeweiligen Wiedergaberaum positioniert. Jedoch können verschiedene räumliche Eigenschaften der virtuellen Quellenobjekte verschiedene Impulsantworten benötigen, die in G darstellbar sind. Gemäß $${\mathrm{R}}_{\mathit{xx}}=\epsilon \left(\mathrm{x}\left(k\right){\mathrm{x}}^H\left(k\right)\right)={\mathrm{GR}}_{\mathit{ss}}{\mathrm{G}}^H,$$

bestimmt G die Korrelationseigenschaften der Lautsprechersignale x(k), beschrieben durch R _xx. Dadurch können wegen der Mehrdeutigkeit verschiedene Mengen von Lösungen für H_est(n) gemäß $${\mathrm{R}}_{\mathit{xx}}{\mathrm{H}}_{\mathrm{est}}^H\left(n\right)={\mathrm{R}}_{\mathit{xd}}$$

existieren, abhängig von den räumlichen Eigenschaften der virtuellen Quellenobjekte. Da alle Lösungen aus dieser Menge von Lösungen die perfekte Identifikation H _est(n) = H beinhalten, unabhängig von R _xx, kann ein variierendes R _xx für eine Systemidentifikation, wie sie in [SHK13] beschrieben ist, vorteilhaft sein. Fig. 6c shows a signal model of a MIMO system identification with a loudspeaker decorrelation as shown in the Fig. 1 and 2 is described. A necessary condition for a clear system identification is with $$N_L{L}_X\le N_S\left(L_X+L_R-1\right).$$

given. This condition applies regardless of the actual spatial properties, such as physical dimensions or radiation characteristics of the virtual source objects. The respective virtual source objects are positioned at positions different from one another in the respective reproduction space. However, different spatial properties of the virtual source objects may require different impulse responses that are representable in G. According to $${\mathrm{R}}_{\mathit{xx}}=\epsilon \left(\mathrm{x}\left(k\right){\mathrm{x}}^H\left(k\right)\right)={\mathrm{GR}}_{\mathit{ss}}{\mathrm{G}}^H.$$

G determines the correlation properties of the loudspeaker signals x (k) described by R _xx . As a result, because of the ambiguity, different amounts of solutions for H _est (n) according to $${\mathrm{R}}_{\mathit{xx}}{\mathrm{H}}_{\mathrm{est}}^H\left(n\right)={\mathrm{R}}_{\mathit{xd}}$$

exist, depending on the spatial characteristics of the virtual source objects. Since all solutions from this set of solutions involve the perfect identification H _est (n) = H , regardless of R _xx , a varying R _xx for a system identification as described in [SHK13] may be advantageous.

A change in the spatial properties of virtual source objects can be exploited to improve system identification. This is made possible by implementing a time-varying rendering system, represented by G '(k). The time-variant rendering system G '(k) comprises the modifier 18, as it is, for example, in Fig. 1 is explained to modify the metadata of the virtual source objects and thus the spatial properties of the virtual source objects. The Rendering systems of the renderers 22 provide loudspeaker signals based on the meta-information modified by the modifier 18 to reflect the wavefields of various virtual source objects, such as point sources, dipole sources, planar sources, or kidney-shaped radiation sources.

In contrast to the descriptions regarding the rendering system G in the Fig. 6a and 6b G '(k) is the Fig. 6c depends on the time step k and can be variable for different time steps k. The renderer 22 produces the decorrelated loudspeaker signals x '(k) directly, so that it is possible to dispense with the addition of noise or a decorrelation filter. The matrix G '(k) can be determined for each time step k in accordance with the selected display scheme, wherein the times k have a temporal difference from one another.

Although some aspects have been described in the context of a device, it will be understood that these aspects also constitute a description of the corresponding method, so that a block or a component of a device is also to be understood as a corresponding method step or as a feature of a method step. Similarly, aspects described in connection with or as a method step also represent a description of a corresponding block or detail or feature of a corresponding device.

Depending on particular implementation requirements, embodiments of the invention may be implemented in hardware or in software. The implementation may be performed using a digital storage medium, such as a floppy disk, a DVD, a Blu-ray Disc, a CD, a ROM, a PROM, an EPROM, an EEPROM or FLASH memory, a hard disk, or other magnetic disk or optical memory are stored on the electronically readable control signals that can cooperate with a programmable computer system or cooperate such that the respective method is performed. Therefore, the digital storage medium can be computer readable. Thus, some embodiments according to the invention include a data carrier having electronically readable control signals capable of interacting with a programmable computer system such that one of the methods described herein is performed.

In general, embodiments of the present invention may be implemented as a computer program product having a program code, wherein the program code is operable to perform one of the methods when the computer program product runs on a computer. The program code can also be stored, for example, on a machine-readable carrier.

Other examples include the computer program for performing any of the methods described herein, wherein the computer program is stored on a machine-readable medium.

In other words, an embodiment of the method according to the invention is thus a computer program which has a program code for performing one of the methods described herein when the computer program runs on a computer. Another example of the inventive methods is thus a data carrier (or a digital storage medium or a computer-readable medium) on which the computer program is recorded for carrying out one of the methods described herein.

Another embodiment includes a processing device, such as a computer or a programmable logic device, that is configured or adapted to perform one of the methods described herein.

Another embodiment includes a computer on which the computer program is installed to perform one of the methods described herein.

In some embodiments, a programmable logic device (eg, a field programmable gate array, an FPGA) may be used to perform some or all of the functionality of the methods described herein. In some embodiments, a field programmable gate array may include a Microprocessor cooperate to perform any of the methods described herein. In general, in some embodiments, the methods are performed by any hardware device. This may be a universal hardware such as a computer processor (CPU) or hardware specific to the process, such as an ASIC.

The embodiments described above are merely illustrative of the principles of the present invention. It will be understood that modifications and variations of the arrangements and details described herein will be apparent to others of ordinary skill in the art. Therefore, it is intended that the invention be limited only by the scope of the appended claims and not by the specific details presented in the description and explanation of the embodiments herein.

literature

• [Ali98] ALI, M .: Stereophonic Acoustic Echo Cancellation System Using Time Varying All-pass filtering for signal decorrelation. In: IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP) Vol. 6. Seattle, WA, May 1998, pp. 3689-3692
• [BBK03] BUCHNER, H .; BENESTY, J .; KELLERMANN, W .: Multichannel Frequency Domain Adaptive Algorithms with Application to Acoustic Echo Cancellation. In: BENESTY, J. (ed.); HUANG, Y. (ed.): Adaptive Signal Processing: Application to Real-World Problems. Berlin: Springer, 2003
• [BDV93] BERKHOUT, AJ; DE VRIES, D .; VOGEL, P .: Acoustic control by wave field synthecsis. In: J. Acoust. Soc. At the. 93 (1993), May, pp. 2764-2778
• [BLA97] Blauert, Jens: Spatial Hearing: The Psychophysics of Human Sound Localization. MIT press, 1997
• [BMS98] BENESTY, J .; MORGAN, DR; SoNDHI, MM: A better understanding and an improved solution to the specific problems of stereophonic acoustic echo cancellation. In: IEEE Trans. Speech Audio Process. 6 (1998), March, No. 2, pp. 156-165
• [Dan03] DANIEL, J .: Spatial sound encoding including near field effect: Introducing distance coding filters and a variable, new ambisonic format. In: 23rd International Conference of the Audio Eng. Soc. 2003
• [GE98] GÄNSLER, T .; ENEROTH, P .: Influence of audio coding on stereophonic acoustic echo cancellation. In: IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP) Vol. 6. Seattle, WA, May 1998, pp. 3649-3652
• [GT98] GILLOIRE, A .; TURBIN, V .: Using auditory properties to improve the behavior of stereophonic acoustic echo cancellers. In: IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP) Vol. 6. Seattle, WA, May 1998, pp. 3681-3684
• [HBK07] HERRE, J .; BUCHNER, H .; KELLERMANN, W .: Acoustic Echo Cancellation for Surround Sound Using Perceptually Motivated Convergence Enhancement. In: IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP) Vol. 1. Honolulu, Hawaii, April 2007, pp. I-17 - I-20
• [MHBOI] MORGAN, DR; HALL, JL; BENESTY, J .: Investigation of several types of nonlinearities for use in stereo acoustic echo cancellation. In: IEEE Trans. Speech Audio Process. 9 (2001), September, No. 6, pp. 686-696
• [SHK13] SCHNEIDER, M .; HUEMMER, C .; KELLERMANN, W .: Wave-Domain Loudspeaker Signal Decorrelation for System Identification in Multichannel Audio Reproduction Scenarios. In: IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP). Vancouver, Canada, May 2013
• [SMH95] SoNDHI, MM; MORGAN, DR; HALL, JL: Stereophonie acoustic echo cancellation - An overview of the fundamental problem. In: IEEE Signal Process. Lett. 2 (1995), August, No. 8, pp. 148-151
• [WWJ12] WUNG, J.; WADA, TS; JUANG, BH: Inter-channel decorrelation by sub-band resampling in frequency domain. In: International Workshop on Acoustic Signal Enhancement {IWAENC). Kyoto, Japan, March 2012, pp. 29-32
• [Bla97] Blauert, Jens: Spatial Hearing: The Psychophysics of Human Sound Localization. MIT press, 1997 ]

used abbreviations

AEC

Acoustic echo cancellation

FIR

finite impulse response

HOA

Higher-order Ambisonics

ILD

interaural level difference

ITD

interaural time difference

LEMS

Speaker Housing Microphone System (Ioudspeaker-enclosure-microphone system)

LRE

Listening room equalization

MIMO

multi-input multi-output

WFS

Wave field synthesis (wave field synthesis)

Claims (8)

1. A device (10, 30) for generating a multitude of loudspeaker signals (x'(k)) based on at least one virtual source object (12a-c) which comprises a source signal and meta information determining a position (P₁; P₂) or type of the at least one virtual source object (12a-c), comprising:

   a modifier (18) configured to time-varyingly modify the meta information; and

   a renderer (22) configured to transfer the at least one virtual source object (12a-c) and the modified meta information in which the type or position (P₁, P₂) of the at least one virtual source object (12a-c) is modified time-varyingly, to form a multitude of loudspeaker signals (x'(k)),

   wherein the modifier (18) is configured to modify the meta information of the at least one virtual source object (12a-c) such that a virtual position (P₁, P₂) of the at least one virtual source object (12a-c) is modified from one time instant to a later time instant and thereby a distance between the virtual position (P₁, P₂) of the at least one virtual source object (12a-c) relative to a position in a playback space (16) is altered by at most 25 %, or

   wherein the modifier (18) is configured to modify the meta information of the at least one virtual source object (12a-c) from one time instant to a later time instant such that, relative to a position (P₁, P₂) in a playback space (16), an interaural level difference is increased by at most 26 % or decreased by at most 21 %, or

   wherein the modifier (18) is configured to modify the meta information of the at least one virtual source object (12a-c) from one time instant to a later time instant such that, relative to a position (P₁, P₂) in a playback space (16), a monaural level difference is increased by at most 26 % or decreased by at most 21 %, or

   wherein the modifier (18) is configured to modify the meta information of the at least one virtual source objet (12a-c) from one time instant to a later time instant such that, relative to a position (P₁, P₂) in a playback space (16), an interaural time difference is modified by at most 30 µs, or

   wherein the at least one virtual source object (12a-c) is arranged in the front (34a, 34b) relative to a listener (17) in a playback space (16) and the modifier (18) is configured to modify the meta information of the at least one virtual source object (12a-c) from one time instant to a later time instant such that a direction of the at least one virtual source object (12a-c) relative to the listener (17) is altered by less than 3° (α₁), or

   wherein the at least one virtual source object (12a-c) is arranged in a lateral direction (36a, 36b) relative to a listener (17) in a playback space (16) and the modifier (18) is configured to modify the meta information of the at least one virtual source object (12a-c) from one time instant to a later time instant such that a direction of the at least one virtual source object (12a-c) relative to the listener (17) is altered by less than 10° (α₂), or

   wherein the modifier (18) is configured to modify the meta information of the at least one virtual source object (12a-c) at a time interval of at least 10 seconds, or

   wherein the modifier (18) is configured to produce an image (12'a) of the at least one virtual source object (12a), wherein the image at least partly comprises the meta information of the at least one virtual source object (12a); to time-varyingly modify the meta information such that the at least one virtual source object (12a) and the image (12'a) comprise mutually different meta information, and to position the image (12'a) at a distance (41) of at most 10 meters to the at least one virtual source object (12a), or

   wherein the renderer (22) is additionally configured to add to the loudspeaker signals (x'(k)) an attenuation or a delay such that a correlation of the loudspeaker signals (x'(k)) is reduced.
2. The device (10, 30) in accordance with claim 1, further comprising:

   a system calculator (28) configured to estimate, based on a plurality of microphone signals (d(k)) and the multitude of loudspeaker signals (x'(k)), a transmission characteristic (H _est(n)) of a playback space (16) where a plurality of loudspeakers which the multitude of loudspeaker signals (x'(k)) is determined for and a plurality of microphones which the plurality of microphone signals (d(k)) originate from may be applied;

   wherein the renderer (22) is configured to calculate the multitude of loudspeaker signals (x'(k)) based on the estimated transmission characteristic (H _est(n)) of the playback space (16).
3. The device (10, 30) in accordance with claim 1 or 2, wherein the renderer (22) is configured to calculate the multitude of loudspeaker signals (x'(k)) in accordance with the rule of a wave-field synthesis algorithm or a high-order ambisonic algorithm, or wherein the renderer (22) is configured to calculate at least 10 loudspeaker signals (x'(k)).
4. The device (10, 30) in accordance with any of the preceding claims, wherein the modifier (18) is configured to modify at least two virtual source objects (12a-c) such that the meta information of a first virtual source object (12a-c) are modified differently as regards position or type of the virtual source object (12a-c) compared to the meta information of a second virtual source object (12a-c); and
   wherein the renderer (22) is configured to calculate the multitude of loudspeaker signals (x'(k)) based on the first modified meta information and the second modified meta information.
5. The device (10, 30) in accordance with any of the preceding claims, wherein the modifier (18) is additionally configured to produce an image (12'a) of the at least one virtual source object (12a), wherein the image at least partly comprises the meta information of the at least one virtual source object (12a); and wherein the modifier is configured to time-varyingly modify the meta information such that the at least one virtual source object (12a) and the image (12'a) comprise mutually different meta information.
6. The device (10, 30) in accordance with any of the preceding claims, wherein the modifier (18) is configured to modify the meta information of the at least one virtual source object (12a-c) of a playback scene reproduced as regards the position or type of the at least one virtual source object (12a-c) partly such that the modification of the playback scene reproduced is not noticeable by a listener (17) in a playback space (16) or not perceived as being disturbing.
7. A method for generating a multitude of loudspeaker signals (x'(k)) based on at least one virtual source object (12a-c) which comprises a source signal and meta information determining the position or type of the at least one virtual source object (12a-c), comprising:

   time-varyingly modifying the meta information; and

   transferring the at least one virtual source object (12a-c) and the modified information in which the type or position of the at least one virtual source object (12a-c) is modified time-varyingly, to form a multitude of loudspeaker signals (x'(k))

   wherein modifying modifies the meta information of the at least one virtual source object (12a-c) such that a virtual position (P₁, P₂) of the at least one virtual source object (12a-c) is modified from one time instant to a later time instant and thereby a distance between the virtual position (P₁, P₂) of the at least one virtual source object (12a-c) relative to a position in a playback space (16) is altered by at most 25 %, or

   wherein modifying modifies the meta information of the at least one virtual source object (12a-c) from one time instant to a later time instant such that, relative to a position (P₁, P₂) in a playback space (16), an interaural level difference is increased by at most 26 % or decreased by at most 21 %, or

   wherein modifying modifies the meta information of the at least one virtual source object (12a-c) from one time instant to a later time instant such that, relative to a position (P₁, P₂) in a playback space (16), a monaural level difference is increased by at most 26 % or decreased by at most 21 %, or

   wherein modifying modifies the meta information of the at least one virtual source object (12a-c) from one time instant to a later time instant such that, relative to a position (P₁, P₂) in a playback space (16), an interaural time difference is modified by at most 30 µs, or

   wherein the at least one virtual source object (12a-c) is arranged in the front (34a, 34b) relative to a listener (17) in a playback space (16) and modifying modifies the meta information of the at least one virtual source object (12a-c) from one time instant to a later time instant such that a direction of the at least one virtual source object (12a-c) relative to the listener (17) is altered by less than 3° (α₁), or

   wherein the at least one virtual source object (12a-c) is arranged in a lateral direction (36a, 36b) relative to a listener (17) in a playback space (16) and modifying modifies the meta information of the at least one virtual source object (12a-c) from one time instant to a later time instant such that a direction of the at least one virtual source object (12a-c) relative to the listener (17) is altered by less than 10° (α₂), or

   wherein modifying modifies the meta information of the at least one virtual source object (12a-c) at a time interval of at least 10 seconds, or

   wherein modifying produces an image (12'a) of the at least one virtual source object (12a), wherein the image at least partly comprises the meta information of the at least one virtual source object (12a); and time-varyingly modifies the meta information such that the at least one virtual source object (12a) and the image (12'a) comprise mutually different meta information, and the image (12'a) is positioned at a distance (41) of at most 10 meters to the at least one virtual source object (12a), or

   wherein transferring adds to the loudspeaker signals (x'(k)) an attenuation or a delay such that a correlation of the loudspeaker signals (x'(k)) is reduced.
8. A computer program comprising a program code for performing the method in accordance with claim 7 when the program runs on a computer.

Images