Method for directional signal processing for a hearing instrument

Abstract

The invention relates to a method for directional signal processing for a hearing instrument (1), wherein a first input signal (E1) is generated from an ambient sound (2) by a first input transducer (M1) of the hearing instrument (1), and a second input signal (E2) is generated from the ambient sound (2) by a second input transducer (M2) of the hearing instrument (1), wherein an angular direction (α) of a sound source (16, 17) relative to a first reference direction (R1), in particular to a frontal direction (12) of a carrier (10) of the hearing instrument (1), is at least approximately detected on the basis of the first input signal (E1) and the second input signal (E2), and wherein an orientation direction (vw1-3, vwk) of said sound source (16, 17), in particular relative to a second reference direction (R2), is at least approximately detected on the basis of the first input signal (E1) and the second input signal (E2).

Description

[0001] The invention relates to a method for directional signal processing for a hearing instrument, wherein, based on a first input signal and a second input signal of the hearing instrument, an angular direction of a sound source relative to a frontal direction of a wearer of the hearing instrument is at least approximately detected.

[0002] In hearing instruments, particularly hearing aids in the narrower sense, an input signal is processed into an output signal, primarily through frequency-band-specific signal processing, and delivered to the ear of the hearing instrument's wearer, for example, via a loudspeaker. This signal processing can also be specifically tailored to the wearer and, in particular, their audiological requirements, such as in hearing aids designed to address the wearer's hearing impairment.

[0003] The aforementioned signal processing is often also directional in the sense that several input signals are processed into the output signal in such a way that sound from different spatial directions of the environment enters the output signal differently, i.e., sound from some spatial directions is suppressed, and sound from other spatial directions is amplified.

[0004] For this type of directional microphone technique, signal processing is usually based on specific models of the environmental situation. For example, a sound source located in the frontal direction of the user is generally considered relevant or a useful signal source, since it is assumed that the user will focus their gaze on the sound sources relevant to them. Another assumption is, for example, that noises from the rear hemisphere of the user, possibly depending on their spectrum, are generally interpreted as background noise.

[0005] However, in complex conversation situations with multiple participants, especially in an environment with many additional background noises (e.g., in a restaurant or similar, a so-called "cocktail party" listening situation), such signal processing reaches its limits, as the distinction between sound signals relevant and irrelevant to the listener is often too coarse or imprecise.

[0006] The invention is therefore based on the objective of providing a method by which an additional possibility is provided for a hearing instrument to evaluate sound sources in their relevance to a wearer of the hearing instrument.

[0007] The aforementioned problem is solved according to the invention by a method for directional signal processing for a hearing instrument, wherein a first input signal is generated from ambient sound by a first input transducer of the hearing instrument, and a second input signal is generated from ambient sound by a second input transducer of the hearing instrument, wherein, based on the first input signal and the second input signal, an angular direction of a sound source relative to a first reference direction, in particular to a frontal direction of a wearer of the hearing instrument, is recognized at least approximately, and wherein, based on the first input signal and the second input signal, an orientation direction of said sound source, in particular relative to a second reference direction, is recognized at least approximately.Advantageous and, in some cases, inventive embodiments are the subject of the dependent claims and the following description.

[0008] A hearing instrument, in this context, generally encompasses any device designed to generate an electrical input signal from ambient sound by means of at least one, in particular, acousto-electrical input transducer; to process said input signal into an output signal by means of, in particular, frequency-band-specific amplification and / or compression; and to generate an audible signal from the output signal and deliver it to the ear of a wearer of this device, in particular by means of an electro-acoustic output transducer (e.g., a loudspeaker, a so-called balanced metal case receiver, or even a bone conduction receiver). Hearing instruments thus include, in particular, headphones (e.g., as "earbuds"), headsets, smart glasses with loudspeakers, etc., which are equipped with a corresponding input transducer.However, a hearing instrument also includes a hearing aid in the narrower sense, i.e., a device for treating a hearing impairment of the wearer, in which, during the processing of the input signal to the output signal, the former is amplified and / or compressed, particularly depending on the frequency band, in order to compensate for the wearer's hearing impairment at least partially by means of an output sound signal generated from the output signal in a user-specific manner.

[0009] In particular, the hearing instrument can also be designed as a binaural hearing system with a first local device and a second local device, in which case the hearing instrument may preferably also have (at least) two further input transducers, wherein the first and second input transducers are arranged in the first local device, and the two further input transducers are arranged in the second local device. The applicability of the described method is generally independent of this.

[0010] A first or second input converter, in this context, includes in particular any device designed to generate a corresponding electrical signal from an acoustic signal. In particular, the generation of the first or second input signal by the respective input converter may also involve preprocessing, e.g., in the form of linear pre-amplification and / or analog-to-digital conversion. The input signal generated is, in particular, an electrical signal whose current and / or voltage fluctuations essentially represent the sound pressure fluctuations in the air.

[0011] The angular direction of a sound source relative to the wearer's frontal direction is understood to mean, in particular, that by wearing the hearing aid as intended on the head (even in the case of a binaural hearing system), especially on or in one ear (in the case of a monaural hearing aid), a clear relationship is established between the two input transducers and the wearer's frontal direction. Through appropriate directional processing of the two input signals, it is thus possible to determine, for example, via time-of-arrival differences of corresponding signal components in the first and second input signals, the angular direction in which a sound source is positioned. This angle is referenced to the first reference direction, which is itself given by, or can be defined based on, the described frontal direction.

[0012] The orientation direction of a sound source is understood here to be, in particular, the direction in which the sound source emits its maximum sound energy, or in which the emission maximum of the sound energy of the sound source lies. In the case of a speaker as the sound source, this direction is normally equivalent to the speaker's frontal orientation. In the case of a loudspeaker, the orientation direction is usually defined by an axis of symmetry of the diaphragm arrangement (or diaphragms). The orientation direction can be specifically related to the first reference direction, preferably the frontal orientation of the person wearing the hearing instrument, as a vectorial direction (i.e., shifted towards the sound source), such that the vectors of the first and second reference directions are parallel. Preferably, however, the second reference direction is defined by the aforementioned angular direction.However, due to the fixed relationship between the first and second reference directions, the definition of the second reference direction can preferably be chosen depending on a subsequent application, since all alternative definitions are equivalent to each other (except for a corresponding angular transformation).

[0013] Approximate recognition of the angular direction and / or orientation direction includes, in particular, specifying a plurality of at least three (and preferably more) discrete angle values ​​as a possible range of values ​​for the respective direction, determining the angle value that most closely corresponds to the actual angular direction or orientation direction, and outputting the angle value as the recognized angular direction or orientation direction.

[0014] In this context, at least approximate recognition means that the recognition of the respective direction can take place either in the aforementioned discrete angle values ​​or continuously and, in particular, exactly (within the scope of the respective possibilities, which are determined in particular by any effects of discretization, sampling rates and finite computing and storage capacities, etc.).

[0015] The detection of the angle or orientation direction based on the first and second input signals particularly includes the fact that the detection is performed directly on the signal components of the first and second input signals, i.e., that a corresponding angle-dependent filter (e.g., a notch filter) for detecting the direction in question is applied directly to the first and second input signals, and / or a propagation delay difference of signal components is determined directly in the first and second input signals. Preferably, a shadowing effect of the head can also be taken into account for such a filter for detection, in particular by means of one or more head-related transfer functions, especially those dependent on the direction of orientation.

[0016] However, the detection of the angle or orientation direction based on the first and second input signals also includes applying the filter used to detect the direction to a first intermediate signal and a second intermediate signal, and / or determining the propagation delay difference of signal components in the first and second intermediate signals. The first intermediate signal is preferably derived solely from the first input signal (i.e., without signal components from the second input signal being directly incorporated into the second intermediate signal). (enter), and the second intermediate signal preferably directly from the second input signal (i.e., preferably without signal contributions from the first input signal in the second intermediate signal).

[0017] The intermediate signals can be generated from the respective input signals via single-channel preprocessing. Alternatively, the intermediate signals can be generated from the first and second input signals respectively via directional preprocessing, for example as forward and reverse cardioid signals.

[0018] The detection of the angular direction of the sound source can precede the detection of the orientation direction, or occur together with it.

[0019] By identifying the orientation of the sound source, which also includes the direction of maximum sound energy emission, and its relationship to the angular direction, it becomes easier to distinguish whether the sound source is a source of unwanted noise or a potential source of useful signal for the wearer. In particular, a sound source can be interpreted as a source of unwanted noise if, due to its orientation (given a specific angular direction), only a small portion of the sound energy is emitted towards the wearer of the hearing instrument. This can be differentiated, for example, using appropriate angular thresholds originating from the sound source.

[0020] Preferably, the orientation of the sound source is used to identify it as a relevant sound source for the carrier. This can also be determined using appropriate angular thresholds originating from the sound source. For example, it can be checked whether the orientation, relative to the angular direction (i.e., the, preferably inverse, angular direction of the sound source as a second reference direction), is greater than, for example, ±10°, ±22.5°, or ±45° (i.e., whether the magnitude of the deviation of the orientation from the angular direction is greater than 10°, 22.5°, or 45°, respectively).

[0021] If this is not the case for the set limit, and the deviation is therefore smaller, then the sound source is assumed to be essentially directed towards the wearer of the hearing system and is classified as a sound source relevant to the wearer.

[0022] The preferred method for identifying a sound source is a speaker. While the procedure is generally applicable to other types of sound sources, it demonstrates particular advantages when dealing with a speaker who can move during a conversation and, in particular, change their orientation relative to the speaker. The speaker can be identified as such based on spectral characteristics; that is, the sound containing their speech is recognized as speech based on characteristic spectral features (such as formants).

[0023] The speaker is then particularly favorably identified as a conversation partner of the carrier based on their orientation. This means, in particular, that the speaker identified as a relevant sound source is then also recognized as being in conversation with the carrier, especially as long as the orientation of said conversation partner is sufficiently aligned with the carrier.

[0024] Advantageously, the directionality of the sound from the sound source is determined, whereby the orientation of the sound source is only determined if the determined directionality does not fall below a lower limit. This prevents computing power from being wasted on detecting a relevant sound source if, for example, the sound source is only diffusely perceptible. Determining the directionality can be done, in particular, based on spectral characteristics of the sound, for example, by detecting a diffuse sound component.

[0025] Advantageously, it is determined whether the sound source lies in a first region located in the front hemisphere of the support, whereby the orientation of the sound source is only detected if the sound source lies in the first region. This prevents computing power from being wasted on detecting a relevant sound source when the sound source lies in a region of the space that is a priori considered irrelevant to the support. The first region can preferably be selected as the region [-67.5°, 67.5°], particularly preferably [-60°, 60°] with respect to the first reference direction or the frontal direction of the support.

[0026] Advantageously, the orientation direction is approximately detected by selecting from a plurality of, preferably at least three, discrete core orientation directions. The orientation direction can also be detected by identifying one of three orientation areas: orientation direction facing the carrier, orientation direction past the carrier in front, or orientation direction past the carrier at the rear. To determine whether a sound source is relevant to the carrier, a higher resolution, i.e., a larger range of values ​​for the orientation directions (with respect to the angular direction or the inverse angular direction), is often unnecessary. Therefore, more complex determination and, in particular, more complex calculations can be avoided.

[0027] It is further advantageous to provide a first plurality of angle-dependent filters, particularly two- or multi-channel filters, whose underlying angles cover at least a partial area of ​​the room. The first plurality of angle-dependent filters are each applied to the first and second input signals and / or to a first and second intermediate signal, which are derived from the first and second input signals, respectively. A set of corresponding angle-dependent features is then determined, and the angular direction of the sound source is recognized based on these angle-dependent features. In particular, the angle-dependent features are quantitative features such that the features for different angular directions can be ordered according to their magnitude. The angle-dependent filters thereby "scan" the partial area of ​​the room.the entire space, so that, based on a maximum or minimum of the angle-dependent features, the angle corresponding to the feature in question can be determined as the angular direction.

[0028] As an angle-dependent feature, a sound level or a degree of attenuation of sound from a sound source arranged in the relevant angular direction is preferably determined. The angle-dependent filters are particularly designed as notch filters, which are applied directly to the first and second input signals, or to a first and second intermediate signal derived from each of these, and effect a corresponding spatial filtering with the relevant angular range, so that a maximum emphasis or a maximum attenuation can be determined at one of the angles.

[0029] InIn a further advantageous embodiment, a second plurality of orientation-direction-dependent filters is provided for at least one angular direction, each corresponding to an orientation direction at said angular direction. The second plurality of orientation-direction-dependent filters are applied to the first and second input signals and / or to the first and second intermediate signals, and a second set of corresponding orientation-direction-dependent features is determined from this. The orientation direction is then recognized based on these orientation-direction-dependent features. In particular, the orientation-direction-dependent features are quantitative features such that the features for different orientation directions can be ordered according to their magnitude.

[0030] The orientation-direction-dependent filters "scan" the space around the sound source at a given angular direction, so that, in particular, the orientation direction associated with the relevant feature can be determined based on a maximum or minimum of the orientation-direction-dependent features.

[0031] Preferably, the degree of attenuation of sound from a sound source oriented in the relevant direction is determined as an orientation-direction-dependent characteristic. The orientation-direction-dependent filters are particularly designed as notch filters, which are applied directly to the first and second input signals, or to a first and second intermediate signal derived from each of these.

[0032] Particularly preferred for determining the degree of attenuation is a comparison signal, generated by applying the relevant angle-dependent and / or orientation-direction-dependent filter to the first and second input signals and / or to the first and second intermediate signals, which is compared to a reference signal that preferably has an omnidirectional directional characteristic. Such a comparison is particularly easy to implement, and the reference signal also provides a comparison value for the overall sound level as a reference for the attenuation by the respective filter.

[0033] In particular, the reference signal is derived solely from the first input signal. This ensures omnidirectional directional characteristics without further signal processing steps.

[0034] Advantageously, a plurality of discrete configurations are specified, each defined by an associated angular direction from a first plurality of discrete angular directions for the sound source and by an associated orientation direction from a second plurality of discrete orientation directions, wherein an orientation-direction-dependent filter is provided for each of the configurations, which is applied to the first and second input signals and / or to the first and second intermediate signals, and from this a corresponding orientation-direction-dependent feature for the respective configuration is determined, and wherein the orientation direction is recognized on the basis of the determined orientation-direction-dependent features.

[0035] This means, in particular, that only a specific number of possible configurations deemed relevant are checked, where each configuration is defined by one of several possible angular directions and an associated orientation (specifically, relative to the respective angular direction or inverse angular direction). For each angular direction, there can be exactly one or more orientations (i.e., one configuration or multiple configurations for that angular direction). This allows for minimal overhead by considering only a reduced number of configurations (compared to a potentially much larger number of possible configurations), providing appropriate filters, and determining the associated features.

[0036] It is expedient to define a head-direction-related transfer function (orientation) for each orientation direction relative to a given angular direction. The pending HeadA related transfer function (ODHRTF) is provided, where each ODHRTF represents a transmission path for sound from the sound source, which is arranged at a given angle with the stated orientation direction, and where the orientation-dependent filters are formed based on the respective associated ODHRTF. This means, in particular, that the respective filter, e.g., a notch filter, considers or preferably incorporates the associated ODHRTF from the sound source (given the angle and orientation direction) to the first or second input transducer. This ensures a precise representation of the spatial propagation of the sound from the sound source to the respective input transducer, thus minimizing spatial distortion effects that could lead to an imprecise determination of the angle or orientation direction.The ODHRTF represents an extension of the concept of the head-related transfer function (HRTF), which represents a transmission path for sound from a sound source arranged at a specific angle, but unlike the ODHRTF, does not depend on the orientation of the sound source. Preferably, the individual ODHRTFs are measured and / or simulated in a calibration beforehand (i.e., before the actual procedure).

[0037] Advantageously, the orientation direction is detected using an artificial neural network based on orientation-direction-dependent features. This means that the orientation-direction-dependent features are passed as input to an artificial neural network, e.g., a Recurrent Neural Network (RNN), and the neural network outputs the orientation direction, determined at least approximately, as an output. In particular, simultaneous processing of the aforementioned inputs is possible, leading to an improvement in runtime.

[0038] Advantageously, for each orientation direction relative to a given angular direction, a head-direction-related transfer function is provided, whereby input variables for the artificial neural network are used, which are derived from the first and / or second input signal filtered with the corresponding head-direction-related transfer function. This enables, on the one hand, the calculation of the orientation direction using a particularly suitable artificial neural network, since this type of task can be performed relatively efficiently even by smaller artificial neural networks, and these can also be configured for said task with comparatively little training effort (since only a small number of parameters are required for small artificial neural networks). On the other hand, the signal processing process for recognizing the orientation direction can be simplified by using acoustic features such as...The input signals, filtered as described, are significantly shortened as input quantities. The invention further describes a hearing instrument comprising a first input converter for generating a first input signal from ambient sound, a second input converter for generating a second input signal from the ambient sound, and a signal processing unit, wherein the hearing instrument is configured to perform the aforementioned method, and wherein, in particular, the signal processing unit is configured (by means of appropriate processor power and addressable working memory, as well as by program instructions) to perform the signal processing steps of the aforementioned method.

[0039] The hearing instrument according to the invention shares the advantages of the method according to the invention. The advantages stated for the method and for its further developments can be transferred analogously to the hearing instrument.

[0040] An embodiment of the invention is explained in more detail below with reference to the drawings. The drawings schematically depict: Fig. 1 shows a block diagram of a hearing instrument, Fig. 2 shows a top view of a hearing situation of a wearer of the hearing instrument. Fig. 1 Fig. 3 shows in a block diagram the sequence of a procedure for determining the orientation direction of the speaker, using a speaker as the sound source. Fig. 2 , Fig. 4 in a top view several configurations, each with specifically oriented speakers around the wearer of the hearing instrument according to Fig. 1 Given, Fig. 5 shows in a diagram the effect of individual filters for the process according to Fig. 3, applied to a configuration according to Fig. 4 , Fig. 6 each shows in top view various possible configurations for three different angular directions of a speaker.

[0041] Corresponding parts and sizes are marked with the same reference symbols in all figures.

[0042] In Figure 1A block diagram schematically depicts a hearing instrument 1, which has a first input transducer M1 and a second input transducer M2. The first input transducer M1 and the second input transducer M2 are each represented by a corresponding microphone. The first input transducer M1 is configured to generate a first input signal E1 from ambient sound 2 when the hearing instrument 1 is in operation. Similarly, the second input transducer M2 is configured to generate a second input signal E2 from ambient sound 2 when the hearing instrument 1 is in operation.The first input signal E1 and the second input signal E2 are fed to a signal processing unit 4, in which both input signals E1 and E2 are processed into an output signal A1, and in particular, are amplified and / or compressed in a frequency-band-specific manner. This signal processing of the two input signals E1 and E2 into the output signal A1 is direction-dependent, meaning that contributions from individual sound sources in different directions can be amplified to varying degrees in the ambient sound. Furthermore, this signal processing can be performed according to the audiological requirements of a user of the hearing aid 1.

[0043] The hearing instrument 1 further comprises an output transducer L1, which is configured to generate an output sound signal 6 from the output signal A1. The schematic representation of the hearing instrument 1 is shown in Figure 1Figure 1 shows a so-called behind-the-ear (BTE) hearing aid with an earpiece 8 in which the output transducer L1 is located. However, the hearing instrument 1 can also be configured in other ways, in particular as an in-the-ear (ITE), in-the-canal (ITC), completely-in-the-canal (CIC), or receiver-in-the-canal (RIC) hearing aid, or even as an earphone not exclusively or primarily intended for the treatment of hearing loss. In particular, the hearing instrument 1 can also be configured as a binaural hearing system with a first local device and a second local device (not shown). In this case, the hearing instrument 1 can also have two additional input transducers, with the two additional input transducers being located in the second local device. The applicability of the described method is generally independent of this configuration.

[0044] In Figure 2A schematic top-down view illustrates the listening situation of a wearer 10 of the hearing instrument 1. The hearing instrument 1 can also be a binaural hearing system with a first local device 1a and a second local device 1b, wherein the two input transducers M1, M2 can be distributed between both local devices 1a, 1b (one input transducer M1, M2 in each local device 1a, 1b), or each local device 1a, 1b can have two input transducers M1, M2 (the input transducers for local device 1b are not shown). The arrangement of the two input transducers M1, M2 in the hearing instrument 1 defines a first reference direction R1 for the intended use of the hearing instrument 1, which is preferably chosen in a frontal direction 12 of the wearer 10. In a front hemisphere 14, a sound source 16 is arranged in an angular direction α with respect to the first reference direction R1, which is given by a speaker S1.In order to determine whether sound source 16 is relevant to carrier 10, and thus whether speaker S1 is a conversation partner of carrier 10, the listening instrument 1 uses the first and second input signals E1, E2 to determine, in a manner to be described later, an orientation direction vw1 of sound source 16 in which sound source 16 emits the maximum of its sound energy. For speaker 1 as sound source 16, this orientation direction vw1 is equivalent to the speaking or looking direction of speaker S1.

[0045] The orientation direction vw1 is preferably defined relative to a suitably chosen second reference direction R2. In this case, the second reference direction R2 is chosen as the (reverse) angular direction α of the speaker S1, but can also be chosen to be identical to the first reference direction R1. Figure 2Another sound source 17 is shown, which is provided by another speaker S2. Speaker S2 is not looking towards the support 10, meaning his orientation vw2 is directed away from the support.

[0046] In Figure 3The sequence of a method for determining the orientation direction vw1 of the sound source 16 is schematically illustrated in a block diagram. The individual signal processing and / or signal analysis steps are preferably performed in the signal processing unit 14 of the hearing instrument 1, in particular on a suitably configured, preferably programmable, signal processor and / or on an application-specific integrated circuit (e.g., an ASIC). In the individual analysis steps, the first and second input signals E1, E2 can be used directly; that is, spatial filters or similar for detecting the angular and / or orientation direction of the sound source 16 are applied directly to the first and second input signals.The first and second input signals E1, E2 can, however, be further processed to at least two intermediate signals (not shown) while preserving at least some of their spatial information (e.g., a cardioid and an anticardioid signal), and said filters can be applied to these intermediate signals.

[0047] In a first process step V1, an angular direction α of the sound source 16 is determined based on the first and second input signals E1, E2. This corresponds to determining a so-called "Direction of Arrival" (DoA) and can be achieved, for example, by determining time-of-arrival and / or level differences in the two input signals E1, E2, and / or by using spatial first filters F1 (αj), which are applied to both input signals E1, E2 (or correspondingly derived intermediate signals) and which are designed as notch filters such that they exhibit maximum attenuation at an angle αj. The individual first filters F1 (αj) can then "scan" the space by varying the angular argument αj.

[0048] Once the angular direction α has been determined, it can optionally be checked in a process step V2, based on the angular direction α, whether the sound source 16 is located in a first region 18, which occupies the front hemisphere 14. The first region 18 can, in particular, be defined by a symmetrical angular range [-y,y] around the frontal direction 12, where γ is preferably selected from the (semi-open) interval [45°, 90°], and especially preferably from the interval [60°, 80°]. The subsequent process steps can then, in particular, only be carried out if the sound source 16 is located in the first region 18, i.e., if the angular direction α lies within the angular range [-y,y].

[0049] Furthermore, in an optional procedure step V3, the directionality d of the sound from sound source 16 can be determined, for example based on spectral features in the input signals E1, E2. The subsequent procedure steps can then only be carried out if the distance d of the sound source 16 does not fall below a predetermined lower limit (not shown).

[0050] In the next process step V4, an orientation direction vw1 is determined for the sound source 16. For this purpose, orientation-direction-dependent second filters F2 (α, vwk) are provided in a first intermediate step V4.1 for the determined angular direction α; that is, different sets of second filters F2 (αi, vwk), F2 (αj, vwk) exist for different angular directions αi, αj. For the second filters F2, an orientation-direction-dependent ODHRTF (α, vwk, M1 / M2) from the sound source 16 arranged in the angular direction α to the first or second input transducer M1, M2 can be used in particular. In a further intermediate step V4.2, the second filters F2 (α, vwk) are applied to the first and second input signals E1, E2 for the individual orientation directions vwk (application ° of the second filter F2 to the set {E1, E2} of input signals), and if necessary normalized via a reference signal N1.The first or second input signal E1, E2 can be used directly as the reference signal N1.

[0051] Finally, in an intermediate step V4.3, the minimum of the input signals E1, E2, filtered as described above with the respective second filters F2 (α, vwk) and normalized with the reference signal N1, is determined via the individual orientation directions vwk, and the corresponding argument arg min is determined as the orientation direction vw1 of the sound source 16. This determination can be approximate, in particular by checking only a limited number of orientation directions vwk (e.g., k=1...3 or k=1...5). Specifically, the orientation direction vw1 can be referenced to the (inverse) angular direction α, so that the range of values ​​for the orientation direction for k =1...3 is limited to the results "aligned towards the support" (vw1 = α or 180° + α), "aligned in front of the support," and "aligned behind the support."

[0052] The orientation direction vw1 can also be determined by a suitably trained artificial neural network DNN (dashed signal path), which receives as input the input signals E1, E2 filtered with the orientation direction-dependent ODHRTF (α, vwk, M1 / M2), and outputs the orientation direction vw1 as a result.

[0053] In Figure 4The diagram schematically depicts several configurations in a conversational situation involving the subject 10 in a top-down view, each configuration being defined by a speaker S1, S2, S3 and their specific orientation vw1, vw2, vw3. Speaker S1 sits diagonally opposite the subject 10 at a table 20, facing forward in their own frontal direction 22 (and thus their orientation vw1 is not directed towards the subject 10, but runs parallel to the subject's frontal direction 12; first configuration). Speaker S2 sits next to the subject 10 with their gaze and thus their orientation vw2 directed towards Speaker S1 (second configuration). Speaker S3 stands diagonally behind the subject 10, with their orientation vw3 directed towards the subject 10 (third configuration).In a real, lively conversation situation, speakers S1, S2, and S3 speak at different times, sometimes interrupting each other.

[0054] In Figure 5 is now for the procedure according to Figure 3 the effect of individual second filters F2 applied to the individual configurations according to Figure 4The diagram in the upper left shows the output values ​​of the second filter for each of the three configurations K1, K2, and K3 (i.e., with respect to the speaker in the given angular direction and orientation), applied to the first and second input signals, plotted against a time axis. As can be seen, the first filter for configuration K3 produces a faint but perceptible minimum up to a time just before 6 seconds, whereas after this time, the second filter for configuration K2 produces a noticeable minimum. This indicates that speaker S1 of the first configuration is active up to a time just before 6 seconds, and speaker S2 of the second configuration is active thereafter. The corresponding result is shown against the time axis in the second diagram (upper right).

[0055] In a third diagram (lower left), the output values ​​of the second filter for each of the three configurations K1, K2, and K3 are plotted against the time axis. As can be seen, the second filter for configuration K3 yields a clear minimum up to a point just before 6 seconds, whereas after this point, the second filter for configuration K2 yields a clear minimum. This suggests that speaker S3 of the third configuration is active up to a point just before 6 seconds, and speaker S2 of the second configuration is active thereafter. The corresponding result is shown against the time axis in the fourth diagram (lower right).

[0056] In Figure 6 For three different angular directions α of the speaker S1 (relative to the frontal direction 12 of the carrier 10), various possible configurations with respect to the orientation direction vw1-3 are shown.

[0057] The upper configurations show the speaker S1 at an angle of α = 90° to the first reference direction R1, which is defined by the frontal direction 12 of the carrier 10 of the hearing instrument 1. In the configuration shown on the left, the speaker S1 has an orientation vw1 of 0° relative to the second reference direction R2, which is defined by the (inverse) angle α. In the configurations shown next to it, the speaker S1 has an orientation vw2 of 45° and vw3 of 90° relative to the second reference direction R2.

[0058] The middle configurations show the speaker S1 at an angle of α = 45° to the first reference direction R1. In the configuration shown on the left, the speaker S1 has an orientation vw1 of 0° relative to the second reference direction R2. In the configurations shown next to it, the speaker S1 has an orientation vw2 of 45° or vw3 of 90° relative to the second reference direction R2.

[0059] The lower configurations show the speaker S1 at an angle of α = 0° to the first reference direction R1. In the lowest configuration, the speaker S1 has an orientation vw1 of 0° relative to the second reference direction R2, which is defined by the (inverse) angle α. In the configurations above, the speaker S1 has an orientation vw2 of 45° and vw3 of -45° relative to the second reference direction R2.

[0060] Although the invention has been illustrated and described in detail by the preferred embodiment, the invention is not limited by the disclosed examples and other variations can be derived by the person skilled in the art without leaving the scope of protection of the invention. Reference symbol list

[0061] 1 Hearing instrument 1a / b Local device (of the hearing instrument) 2 Ambient sound 4 Signal processing unit 6 Output sound signal 8 Earpiece 10 Carrier 12 Frontal direction 14 Anterior hemisphere 16 Sound source 17 Sound source 18 First area 20 Table 22 Frontal direction A1 Output signal DNN artificial neural network E1 / 2 First / second input signal F1 (angle-dependent) first filter F2 (orientation-direction-dependent) second filter L1 Output converter K1-3 Output value of the second filter (for configuration 1-3) M1 / 2 First / second input converter N1 Reference signal R1 / 2 First / second reference direction S1-3 Speaker V1-4 Process step V4.1-3 Intermediate step vw1-3, vwk Orientation direction α, αj Angle direction

Claims

1. Method for directional signal processing for a hearing instrument (1), - wherein a first input signal (E1) is generated from an ambient sound (2) by a first input transducer (M1) of the hearing instrument (1), and a second input signal (E2) is generated from the ambient sound (2) by a second input transducer (M2) of the hearing instrument (1), - wherein, based on the first input signal (E1) and the second input signal (E2), an angular direction (α) of a sound source (16, 17) relative to a first reference direction (R1), in particular to a frontal direction (12) of a carrier (10) of the hearing instrument (1), is at least approximately detected, and - wherein, based on the first input signal (E1) and the second input signal (E2), an orientation direction (vw1-3, vwk) of said sound source (16, 17), in particular relative to a second reference direction (R2), is at least approximately detected.

2. Method according to claim 1, wherein the sound source (16, 17) is recognized as a sound source relevant to the carrier (10) based on its orientation direction (vw1-3, vwk).

3. Method according to claim 1 or claim 2, wherein said sound source (16, 17) is identified as a speaker (S1-3).

4. Method according to claim 3, referring back to claim 2, wherein the speaker (S1-3) is recognized as a conversation partner of the carrier (10) based on the orientation direction (vw1-3).

5. Method according to one of the preceding claims, wherein a directionality (d) of the sound of the sound source (16, 17) is detected, and wherein the orientation direction (vw1-3, vwk) of the sound source (16, 17) is determined only if the detected directionality (d) does not fall below a lower limit.

6. Method according to one of the preceding claims, wherein it is detected whether the sound source (16, 17) is located in a first area (18) which is located in the front hemisphere (14) of the support (10), and wherein the detection of the orientation direction (vw1-3, vwk) of the sound source (16, 17) is only carried out if the sound source (16, 17) is located in the first area (14).

7. Method according to one of the preceding claims, wherein the approximate detection of the orientation direction (vw1-3, vwk) is carried out by a selection from a plurality of, preferably at least three, discrete core orientation directions.

8. A method according to any of the preceding claims, wherein a first plurality of angle-dependent filters (F1) is provided, the underlying angles (αj) of which cover at least a partial area of ​​the space, wherein the first plurality of angle-dependent filters (F1) are each applied to the first and second input signals (E1, E2) and / or to a first and a second intermediate signal, which are each derived from the first and second input signals (E1, E2), respectively, and a set of corresponding angle-dependent features is determined therefrom, and wherein the angular direction (α) of the sound source (16, 17) is recognized on the basis of the angle-dependent features.

9. A method according to one of the preceding claims, wherein for at least one angular direction (α) a second plurality of orientation-direction-dependent filters (F2) is provided, each corresponding to an orientation direction (vw1-3, vwk) at said angular direction (α), wherein the second plurality of orientation-direction-dependent filters (F2) are applied to the first and second input signals (E1, E2) and / or to the first and second intermediate signals, and a second set of corresponding orientation-direction-dependent features is determined from this, and wherein the orientation direction (vw1-3, vwk) is recognized on the basis of said orientation-direction-dependent features.

10. A method according to any of the preceding claims, wherein a plurality of discrete configurations is specified, each of which is defined by an associated angular direction (α) from a first plurality of discrete angular directions (αj) for the sound source (16, 17) and by an associated orientation direction (vw1-3) from a second plurality of discrete orientation directions (vwk), wherein an orientation-direction-dependent filter (F2) is provided for each of the configurations, which is applied to the first and second input signals (E1, E2) and / or to the first and second intermediate signals, and from this a corresponding orientation-direction-dependent feature for the respective configuration is determined, and wherein the orientation direction (vw1) is recognized on the basis of the determined orientation-direction-dependent features.

11. Method according to claim 9 or claim 10, wherein for each orientation direction (vw1-3, vwk) to a given angular direction (α) a head-direction-related transfer function (ODHRTF) is provided, wherein each of the head-direction-related transfer functions (ODHRTF) represents a transmission path for sound from the sound source (16, 17) which is arranged in a relevant angular direction (α) with said orientation direction (vw1-3, vwk), and wherein the orientation-direction-dependent filters (F2) are each formed based on the respective head-direction-related transfer function (ODHRTF).

12. Method according to one of claims 8 to 11, wherein as an angle-dependent and / or orientation-direction-dependent feature a degree of attenuation of a sound from a sound source (16, 17) is determined which is arranged in the relevant angular direction (α) and / or is aligned in the associated orientation direction (vw1-3, vwk).

13. Method according to claim 12, wherein for the degree of attenuation a comparison signal (N1) which is generated by applying the relevant angle-dependent and / or orientation-direction-dependent filter (F1, F2) to the first and second input signals (E1, E2) and / or to the first and second intermediate signals is compared with a reference signal (N1) which preferably has an omnidirectional directional characteristic.

14. Method according to claim 13, wherein the reference signal (N1) is derived only from the first input signal (E1).

15. Method according to one of claims 8 to 14, wherein the orientation direction (vw1-3, vwk) is detected based on the orientation direction-dependent features by means of an artificial neural network (DNN).

16. Method according to claim 15, wherein for each orientation direction (vw1-3, vwk) to a given angular direction (α) a head-direction-related transfer function (ODHRTF) is provided, and wherein input variables for the artificial neural network (DNN) are used which are derived from the first and / or second input signal (E1, E2) filtered with the associated head-direction-related transfer function (ODHRTF).

17. Hearing instrument (1) comprising - a first input transducer (M1) for generating a first input signal (E1) from an ambient sound (2), - a second input transducer (M2) for generating a second input signal (E2) from the ambient sound (2), and - a signal processing unit (4), wherein the hearing instrument (1) is configured to perform the method according to one of the preceding claims.

Images