Methods for reducing echo in a hearing instrument

Abstract

Methods for reducing echo in a hearing instrument (1), - wherein a first input signal (x1) is generated from an ambient sound (8) by an electroacoustic first input transducer (6) of the hearing instrument (1), - wherein an external input signal (xe) is received from an external communication device (4) by means of a communication device (24) of the hearing instrument (1), - where an output signal (y) is generated based on the first input signal (x1) and the external input signal (xe) of the hearing instrument (1), - wherein, based on the output signal (y) in an adaptive first filter (18), a compensation signal (c) is generated to reduce echo and / or acoustic feedback (h), - where an error signal (e) is generated based on the first input signal (x1) and the compensation signal (c), - wherein a control variable (K) is generated based on filter coefficients of the first filter (18) and / or based on a comparison of the error signal (e) with the compensation signal (c) and / or with the first input signal (x1), - wherein, depending on the control variable (K), a second filter (40) is applied to suppress a residual echo and / or residual feedback on an intermediate signal (z) derived from the input signal (x1), thereby generating a transmission signal (t), and - wherein the transmission signal (t) is transmitted to the external communication device (4) via the communication device (24).

Description

[0001] The invention relates to a method for reducing echo in a hearing instrument, wherein a first input signal is generated from ambient sound by an electroacoustic first input transducer of the hearing instrument, wherein an external input signal is received from an external communication device by means of a communication device of the hearing instrument, wherein an output signal is generated based on the first input signal and the external input signal of the hearing instrument, wherein a transmission signal is generated based on an intermediate signal derived from the input signal, the transmission signal is transmitted to the external communication device by means of the communication device.

[0002] A hearing instrument is generally understood to be any device that is designed to generate a sound signal from an electrical signal - which can also be provided by an internal signal of the device - and to feed it to the ear of a wearer of this device, i.e. in particular headphones (e.g. as an "earplug"), a headset, data glasses with a loudspeaker, etc. Hearing instruments of this type are increasingly also having one or more microphones in order to control individual functions using voice commands, such as selecting a track from a music library. Hearing instruments of this type are, however, also increasingly being used in conjunction with mobile communication devices such as smartphones, tablet PCs or smartwatches, etc., whereby individual functions of the communication device can be controlled using voice commands using the microphone(s) of the hearing instrument (e.g. via a Bluetooth connection or similar).between the hearing instrument and the communication device).

[0003] Such hearing instruments are increasingly being used in conjunction with a communication device of the type mentioned to supplement the telecommunications functions of the communication device, in particular for calls via a mobile network or VoIP calls. The microphone(s) of the hearing instrument record the speech signal of a wearer of the hearing instrument (who is also the user of the communication device) and transmit it to a conversation partner via the communication device. This usually occurs without any further contributions from an audio signal recorded in the communication device itself by separate microphones there. An audio signal with the conversation partner's speech contributions, which was recorded by a communication device (e.g. a telephone or similar) of the conversation partner, is transmitted to the communication device in question, and from there to the hearing instrument.There, the said audio signal with the speech contributions of the conversation partner is routed to an output converter (e.g. a loudspeaker) of the hearing instrument, so that the wearer can hear the speech contributions through the reproduction of the audio signal.

[0004] However, this can result in acoustic feedback of the conversation partner's speech into the hearing aid's microphone, thus incorporating said speech into the signal being transmitted to the conversation partner. The conversation partner will then hear their own speech as an echo with a delay of several 10 ms to several 100 ms (depending on the type of connection), which is irritating.

[0005] EP 1 367 856 A2 proposes an amplifier device that, in addition to an amplifier device having an acoustic and an inductive input, has two separate compensation paths. In the first compensation path, a first filter device acts to compensate for acoustic feedback, and in the second compensation path, a second filter device acts to compensate for inductive feedback. This division allows the two very different feedbacks to be compensated individually without major effort.

[0006] EP 1 020 068 B1 discloses an echo cancellation method for a telecommunications device. Said method provides for generating at least one compensation signal as a function of an output signal of the telecommunications device by simulating transmission characteristics of a transmission path, taking nonlinear distortions of the transmission path into account, and for compensating echo components in an input signal of the telecommunications device as a function of the at least one compensation signal in order to obtain an echo-compensated result signal. The method is further developed in that nonlinear distortions of the transmission path are taken into account using an adaptive method, and that the adaptation in this method is performed on the basis of the echo-compensated result signal.The invention provides a method for echo cancellation that is particularly effective and flexible even in the case of strong non-linear distortions and requires little or no adjustment effort.

[0007] US 2020 / 0 322 738 A1 describes a method for operating a hearing aid with a main digital filter that is adapted to selectively represent a specific combination of serially and parallel-coupled virtual digital filters.

[0008] The invention is therefore based on the object of specifying a method for operating a hearing instrument in which the said type of echo is suppressed as effectively as possible and which can be implemented as efficiently as possible in the hearing instrument.

[0009] The stated object is achieved according to the invention by a method for reducing echo in a hearing instrument, wherein a first input signal is generated from ambient sound by an electroacoustic first input transducer of the hearing instrument, wherein an external input signal is received from an external communication device by means of a communication device of the hearing instrument, wherein an output signal is generated on the basis of the first input signal and the external input signal, in particular in a signal processing device of the hearing instrument, wherein a compensation signal for reducing echo and / or acoustic feedback is generated on the basis of the output signal in an adaptive first filter, and wherein an error signal is generated on the basis of the first input signal and the compensation signal, which error signal is fed in particular to the adaptive first filter as an input variable.

[0010] According to the method, a control variable is generated based on filter coefficients of the first filter and / or based on a comparison of the error signal with the compensation signal and / or with the first input signal; a second filter is applied, depending on the control variable, to suppress a residual echo or residual feedback on an intermediate signal derived from the input signal and in particular from the error signal, thereby generating a transmission signal; and the transmission signal is transmitted to the external communication device by means of the communication device. Advantageous and, in part, inventive embodiments are the subject of the dependent claims and the following description.

[0011] A hearing instrument generally includes a device of the type described above, i.e. in particular a pair of headphones (e.g. as an “earplug”), a headset, data glasses with a loudspeaker, etc. 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 an input signal generated from an ambient signal by means of a microphone is processed into an output signal and is amplified in particular in a frequency band-dependent manner, and an output sound signal generated from the output signal (or from a playback signal which is derived from the output signal) by means of a loudspeaker or similar is suitable for at least partially compensating for the wearer’s hearing impairment, in particular in a user-specific manner.

[0012] The hearing instrument preferably has a communication mode in which a voice call with an external communication device is possible using the communication device and a communication device associated with the hearing instrument via a corresponding connection. The external communication device is preferably a landline telephone, mobile phone, tablet PC, or similar device of a conversation partner of a wearer of the hearing instrument. The communication device associated with the hearing instrument is a mobile phone of the wearer, which can be connected to the hearing instrument via Bluetooth or similar, and is preferably also connected for the communication mode of the hearing instrument.

[0013] However, the hearing instrument can also be configured, in particular, to establish a connection to the external communication device via WLAN and a corresponding Internet connection in communication mode in order to receive the external input signal from the external communication device and to send the transmission signal to it.

[0014] The communication device is preferably provided as an antenna or as another appropriately suitable and configured transmitting / receiving device in the hearing instrument.

[0015] The output signal is preferably generated such that, in said communication mode, during a voice call conducted by the wearer via the hearing instrument, the external input signal, which preferably includes the voice contributions of a conversation partner in the voice call, is received together with the first input signal (which in particular includes the wearer's own voice contributions). The two input signals mentioned can be subjected to further processing steps during the generation of the output signal, and in particular, further signals can be added and / or subtracted. The output signal can then preferably be reproduced as a playback signal by an output transducer of the hearing instrument (e.g., a loudspeaker).However, the playback signal can also be given by the resulting transmission signal, in particular in the case where the second filter is applied to the output signal of the signal processing.

[0016] In general, here and in the following, generating a resulting signal from a preceding signal means that the signal components of the preceding signal are incorporated into the resulting signal in at least a number of frequency bands. Furthermore, the signal components of other signals can also be included, or corresponding signal components of other signals can be subtracted from the preceding signal or a signal derived therefrom. In particular, frequency band-wise amplification and frequency band-wise compression are also possible when processing the preceding signal to produce the resulting signal. The resulting signal is derived from the preceding signal if it is generated from the preceding signal (in the sense described here).

[0017] The adaptive first filter for generating the compensation signal preferably receives the error signal as an input variable and preferably models a corresponding acoustic feedback path from an output transducer of the hearing instrument, through which the output signal (or a playback signal derived therefrom) is reproduced, to the first input transducer, so that the application of the first filter to the output signal provides an estimate of the acoustic feedback. The error signal is preferably generated by subtracting, and particularly preferably by subtracting, the compensation signal from the first input signal.

[0018] The echo then arises, in particular, from the external input signal (or signal components thereof) reproduced via said output transducer being coupled into the first input transducer in the form of acoustic feedback, and thus into the intermediate signal generated from the input signal. If the intermediate signal is transmitted to the external communication device, a conversation partner of the wearer will hear their speech contributions—which are fed back to them as described—as an echo when the transmitted intermediate signal is reproduced. Consequently, the intermediate signal is further processed into the transmission signal by means of the second filter in order to suppress this echo, which is caused, in particular, by the acoustic feedback of the external input signal. The first filter can achieve the suppression of the acoustic feedback or echo, in particular the echo of the described type, in a manner known in the prior art.

[0019] The error signal is then compared with the compensation signal. The comparison can be performed, for example, by simply comparing the magnitudes, squares of the magnitudes, or signal levels. For this purpose, the compensation signal and / or the error signal can be temporally smoothed beforehand, e.g., using recursive averaging or similar.

[0020] By comparing the compensation signal with the error signal, a control variable is generated, which the second filter operates based on. For example, the control variable can consist solely of a binary value determined from the sign of a magnitude comparison of the amplitudes or signal levels (or squared magnitudes, or generally any monotonic function of the magnitudes) of the two signals. Depending on this binary value, the second filter is then "switched on" or off with predefined parameter values.

[0021] However, the control variable can also consist of a continuously monotonic function of the amplitudes or signal levels (or similar) of the error signal and the compensation signal, whereby the second filter is applied or "bypassed" (i.e., deactivation of the second filter) depending on the sign of the said function. If the second filter is applied, the strength of the application depends on the value of the control variable, thus, in particular, varies with the amplitudes of the two aforementioned signals. In particular, the control variable can also be temporally smoothed (alternatively or in addition to the compensation and / or error signal).

[0022] Alternatively, or additionally, in an embodiment according to the invention, the control variable is determined based on filter coefficients of the first filter. This can be done, in particular, based on dead-time coefficients. Due to the acoustic propagation time between an output transducer of the hearing instrument and the first input transducer, a feedback path modeled by the first filter will necessarily have a number of coefficients, all of which are identically zero. These are the coefficients that model the transfer function of the feedback path at a time before the arrival of the actual feedback output sound.

[0023] On the one hand, these coefficients can be set to zero by definition, and the first filter can be "shifted" in time accordingly. On the other hand, these coefficients can also be adapted. The distance of these coefficients from zero then serves as an indicator of how well the first filter actually maps the real acoustic feedback path. This makes it possible to exploit the fact that errors in the adaptation are "on average" distributed evenly across the individual coefficients of a filter. By taking the sum of squares of the relevant coefficients, which should be zero according to the propagation delay, normalized over the total energy of all coefficients, a quantitative measure of the precision of the adaptation can be obtained. Using this measure, the control variable can then be determined (e.g. via a functional dependency, or via a lookup table depending on the above-mentioned normalized sum of squares of the so-called“Dead time coefficients”).

[0024] The second filter is designed to suppress any residual feedback remaining in the error signal after the adaptive first filter has been applied to suppress the feedback. The second filter is applied to an intermediate signal derived from the error signal, particularly including the trivial case of derivation where the intermediate signal is determined by the error signal itself. In this case, the playback signal, which is converted into an output sound signal by an output transducer of the hearing instrument, is determined by the output signal.

[0025] However, the second filter can also be applied to the output signal, which is derived from the error signal by means of a (particularly user-specific and frequency band-wise) amplification and / or compression and, if necessary,

[0026] Signal processing (including noise reduction and / or directional microphones using a second input signal from the hearing instrument) was generated.

[0027] The second filter can be implemented, for example, by applying an amplification or attenuation factor to the intermediate signal and / or by compression (or by changing the compression parameters such as the compression ratio or knee point of the characteristic curve). While the second filter can, in principle, also operate in the time domain, it is preferably implemented in the time-frequency domain and can be used there frequency band by frequency (e.g., by using frequency-band-dependent amplification factors to attenuate the intermediate signal). A frequency-band-by-frequency adaptive filter is also possible, which preferably has multiple filter coefficients for individual frequency bands in the time-frequency domain and is implemented by appropriate convolution with the intermediate signal in the respective frequency band.

[0028] The intermediate signal filtered by the second filter is now transmitted as the transmission signal to the external communication device.

[0029] The invention exploits the fact that the compensation signal, due to the appropriately designed first filter (which is intended to suppress feedback), contains more components of the feedback-generated external input signal and thus the speech contributions of the conversation partner, while the error signal, after the feedback has been eliminated, essentially contains the speech contributions of the hearing instrument wearer. The suppression of the feedback by the compensation signal generated by the first filter is not complete, which is why the residual feedback must still be suppressed by the second filter. However, the temporal information regarding the speech contributions of the wearer and their conversation partner, which is already contained in the compensation signal and accordingly also in the error signal, can be used for the application of the second filter: If the signal level or signal amplitude in the compensation signal is higher than in the error signal, especially over a period of several samples (such as one frame), it can be assumed that the call partner is active during this period, and their speech contributions are therefore included in the external input signal while the wearer is not speaking. Conversely, the signal level or amplitude in the error signal is higher when the wearer is speaking but not the call partner.

[0030] By comparing the signals mentioned above and using the resulting control variable, the conversation partner's speech contributions can now be identified and suppressed in the intermediate signal by the second filter before the resulting transmission signal is transmitted to the conversation partner's external communication device. This prevents an echo from the conversation partner, which could occur if their own speech contributions were played back.

[0031] A major advantage here is the use of the error signal and the compensation signal to control the suppression of the residual feedback, since these two signals are already provided by the adaptive first filter, and thus no additional computing power (or corresponding hard-wired components for signal processing) is required.

[0032] Advantageously, the second filter is applied to the intermediate signal depending on the sign of the logarithm of the control variable. In particular, this means that no filter is applied for a positive sign if it is applied for a negative sign. The control variable thus serves as a "switch" for the second filter via the sign of the logarithm—or preferably a comparable, especially equivalent, mathematical implementation.

[0033] It proves advantageous if the control variable is calculated using a quotient of the amplitudes, absolute values, squared values, and / or signal levels of the error signal and the compensation signal. In particular, when implementing the method for different frequency bands, different of the aforementioned characteristic variables of the two signals can be used to calculate the respective quotient in the frequency band. In particular, temporal smoothing of the respective characteristic variables can also be performed before the quotient is calculated, for example, by means of recursive averaging. Using a quotient, it is particularly easy to estimate whether the signal contributions predominate in the compensation signal or in the error signal.

[0034] Preferably, the second filter exhibits a continuous, monotonic dependence on the control variable. Particularly preferably, the second filter is applied to the intermediate signal as a function of the sign of the logarithm of the control variable (or a mathematically equivalent function). For example, a gain factor for attenuating the intermediate signal can be applied to the intermediate signal as a monotonic function of the control variable if the sign of the logarithm of the control variable indicates that the signal contributions predominate in the compensation signal.

[0035] In an advantageous alternative, the second filter is operated with a predefined parameter value that is, in particular, independent of the magnitude of the control variable. This means, in particular, that the control variable is used merely as the aforementioned "switch" for the second filter, but the second filter exhibits no further dependence on the magnitude of the control variable (assuming the logarithm remains unchanged). Operating with predefined parameters, such as gain factors for a fixed attenuation, saves further computational resources.

[0036] In advantageous embodiments, the second filter applies an amplification factor and / or a compression to the intermediate signal to attenuate it, wherein in the case of compression, in particular a compression parameter, e.g. a compression ratio and / or a knee point of a compression characteristic curve, can be set in (in particular continuously monotonic) dependence on the control variable.

[0037] Preferably, the second filter is applied to a number of frequency bands of the intermediate signal in the time-frequency domain. The second filter can then apply individual frequency-band-wise gain factors and / or compression to at least some frequency bands of the intermediate signal.

[0038] In particular, the second filter can alternatively or additionally be applied to the intermediate signal in a number of frequency bands as a second adaptive filter, each with more than one filter coefficient. This means that the feedback suppression is divided between the first filter and, in individual frequency bands in which the conversation partner's speech contributions are returned to them as an echo due to the residual feedback, into the aforementioned adaptive second filter. The second filter can then be implemented, for example, as an FIR or IIR filter by convolving at least two samples of the intermediate signal in the frequency band, whereby filter coefficients can be determined, for example, using an NLMS algorithm.

[0039] To generate a playback signal, non-linear processing, particularly frequency distortion such as frequency shifting, is advantageously applied to the intermediate signal or to a signal derived from the intermediate signal from the hearing instrument. The playback signal is preferably reproduced by an electroacoustic output transducer of the hearing instrument. This allows the playback signal to be additionally decorrelated from the input signals in order to interrupt the feedback loop formed by the acoustic feedback path around the hearing instrument and its internal signal amplification.

[0040] In an advantageous embodiment, the compensation signal is generated based on the output signal or the transmission signal to which the first filter is applied, whereby the first filter is adapted based on the error signal. The error signal serves as a measure of the quality of the adaptation for the first filter (e.g., in an NLMS algorithm).

[0041] Preferably, the error signal or the output signal is used directly as the intermediate signal. In the case of the output signal, the error signal then passes through the hearing instrument's internal signal processing, which is then incorporated into the transmission signal. If the transmission signal is not to be influenced by the hearing instrument's internal signal processing, the error signal is used as the intermediate signal, from which the transmission signal is generated by applying the second filter.

[0042] The invention further mentions a hearing instrument comprising an electroacoustic first input transducer for generating a first input signal from an ambient sound, a communication device for receiving an external input signal from an external communication device and for sending a transmission signal of the hearing instrument to the external communication device, and a signal processing device which is configured to generate an output signal based on the first input signal and the external input signal.Here, an adaptive first filter is implemented in the hearing instrument, which is configured to generate a compensation signal for reducing echo and / or acoustic feedback based on the output signal. The signal processing device is further configured to generate an error signal based on the first input signal and the compensation signal, and to generate a control variable based on a comparison of the error signal with the compensation signal. Furthermore, a second filter is implemented in the hearing instrument, which is configured to suppress a residual echo or residual feedback in an intermediate signal derived from the input signal and in particular from the error signal, and thereby generate the transmission signal.

[0043] The hearing instrument according to the invention shares the advantages of the method according to the invention. The advantages stated for the method and its further developments can be applied analogously to the hearing instrument. Advantageous embodiments of the method are preferably implemented by corresponding configurations of the hearing system.

[0044] An embodiment of the invention is explained in more detail below with reference to the accompanying drawings, each of which shows schematically: Fig. 1 in a block diagram a hearing instrument by means of which speech contributions of the wearer are sent to an external communication device in a long-distance call, Fig. 2 in a block diagram the hearing instrument according to Fig. 1, by means of which speech contributions of the conversation partner in the long-distance call are received, which experience a feedback, Fig. 3 in a block diagram a design of the hearing instrument according to Fig. 1, through which the feedback to Fig. 2 can be suppressed, and Fig. 4 shows in a time diagram a control variable for the suppression of the feedback in the embodiment according to Fig. 3.

[0045] Corresponding parts and sizes are provided with the same reference numerals in all figures.

[0046] In Fig. 1 schematically shows a block diagram of a hearing instrument 1, which in this case is a hearing aid 2 (in the narrower sense). The hearing instrument 1 is designed so that a wearer (not shown) of the hearing instrument 1 can use it to conduct a long-distance call, e.g. a telephone or VoIP call, with a conversation partner (not shown), who in turn uses a communication device such as a mobile phone, which in the Fig. 1 is shown as an external communication device 4.

[0047] The hearing instrument 1 has an electroacoustic first input transducer 6, which in this case is provided by a microphone, and which is configured to generate a first input signal x1 from an ambient sound 8 of the hearing instrument 1. In a manner to be described below, the first input signal x1 is fed to a signal processing device 10, in which an output signal y is generated based on signal components of the first input signal x1. Said output signal y is converted, as a playback signal w, by an electroacoustic output transducer 12 of the hearing instrument 1 into an output sound 14. The output transducer 12 is provided in this case by a loudspeaker. From the output transducer 12, portions of the output sound 14 reach the first input transducer 6 via an acoustic feedback path 16, thereby resulting in acoustic feedback h of the output signal Y.The hearing instrument may also have a second input transducer (not shown) which accordingly generates a second input signal which is processed together with the first input signal, in particular by means of directional microphones.

[0048] To suppress feedback h, an adaptive first filter 18 is implemented in the hearing instrument 1, which is applied to the output signal y and generates a compensation signal c therefrom. The compensation signal c is subtracted from the first input signal x1 at a first node 20, so that an error signal e is generated therefrom. The error signal e is fed to the adaptive first filter 18 in order to estimate the quality of the adaptation based on the error signal e. The first filter 18 and the first node 20 can be physically implemented in the signal processing device 10, which further processes the error signal e to the output signal y by means of signal processing that is individually tailored to the audiological needs of the wearer by means of frequency band-dependent amplification and / or compression. For reasons of clarity, the signal processing device 10 is shown in Fig. 1, however, should be understood merely as the blog of signal processing in which the said “actual” processing of the error signal e to the output signal y takes place.

[0049] In the aforementioned long-distance call, speech contributions 22 of the wearer of the hearing instrument 1 are recorded in the first input signal x1 by the first input transducer 6, reduced by the compensation signal c at the node 20 to correct the acoustic feedback h, and processed in the signal processing device 10 to form the output signal y (dashed line). Further noise suppression and / or speech enhancement algorithms can also be implemented here. The output signal y is fed as a transmission signal t to a communication device 24 of the hearing instrument 1, which can be provided, for example, by an antenna for Bluetooth or WLAN. The transmission signal t, which contains the speech contributions 22 of the wearer, is then transmitted to the external communication device 4 by means of the communication device 24.There, an external playback sound 28 is generated from the transmission signal t by means of a loudspeaker 26, so that the conversation partner can hear the speech contributions 22 accordingly.

[0050] The transmission of the transmission signal t to the external communication device 4 can be carried out in particular using a local communication device (not shown) of the wearer of the hearing instrument 1, which is connected to the hearing instrument 1 via the communication device 24 and forwards the transmission signal t to the external communication device 4. The local communication device can in particular be a smartphone or similar device. However, the hearing instrument 1 can also have an internet connection directly via the communication device 24 via WLAN, so that the long-distance call with the external communication device 4 is conducted as a VoIP call.

[0051] In Fig. 2 is a schematic block diagram of the hearing instrument 1 according to Fig. 1 illustrates the case where the other party is speaking during a long-distance call. A microphone 32 of the external communication device 4 generates an external input signal xe containing speech contributions 30 (dashed line) from the other party. The external input signal xe containing the speech contributions 30 is transmitted to the hearing instrument 1, where it is received by the communication device 24. The transmission takes place in the manner just described.

[0052] The external input signal xe is now processed together with the error signal e in the signal processing device 10 to form the output signal y. The error signal e, which according to Fig. 1, which contains the wearer's own speech contributions 22 of the hearing instrument 1, is included in the output signal y (thus in the playback signal w), since otherwise it would be unpleasant for the wearer not to hear his own voice in the output sound 14 when he speaks.

[0053] In Fig. 2, however, only the speech contributions 30 of the conversation partner are represented in the output signal y and correspond to the output sound 14. Via the acoustic feedback path 16 from the output transducer 12 to the first input transducer 6, the speech contributions 30 of the conversation partner are input into the first input signal x1 (and possibly into a second input signal of a second microphone of the hearing instrument), and thus also into the error signal e. Accordingly, these speech contributions are converted back into the output signal y after processing in the signal processing device 10 (and amplification during this process). The speech contributions 30 have thus passed through a closed feedback loop in the hearing instrument 1 once.Since the output signal y is now transmitted as a transmission signal t to the external communication device 4, from which the external playback sound 28 is generated there by the loudspeaker 26, the speech contributions 30 of the conversation partner in the external playback sound 28 become audible to the conversation partner as an echo.

[0054] Even if the feedback is suppressed by the adaptive first filter 18, this suppression is usually 15 dB to 25 dB. However, it is recommended (among others by the ITU-T) that in long-distance conversations using hearing instruments, the re-coupling of the "telecommunication signal" (i.e. the received signal) by acoustic feedback be suppressed by at least 35 dB. The residual feedback remaining after suppression by the first filter 18 can therefore be described as highly relevant. The speech contributions 30 therefore remain in the error signal e (dotted line) even "after" correction of the acoustic feedback h by the compensation signal c, albeit weaker, and are accordingly transmitted to the external communication device 4, where they are output in the external playback sound 28.

[0055] In Fig. 3 is a schematic block diagram of a further development of the hearing instrument 1 according to Fig. 1 and Fig. 2, which enables the suppression of said residual feedback. For the sake of clarity, only the signal flow for the speech contributions 30 of the conversation partner is shown, who communicates at the location of the external communication device 4 and directly through it. At the same time, or in rapid alternation, speech contributions 22 of the wearer of the hearing instrument 1 also occur, as shown in Fig. 1. For these, the signal flow is as shown in Fig. 1 shown.

[0056] As shown in the Fig. 2 shown long-distance call, is also shown in Fig. 3, the external communication unit 4 generates the external input signal xe through the microphone 32, which contains the speech contributions 30 of the conversation partner. The signal flow for the external input signal xe and thus for the said speech contributions 30 (dashed and dotted lines) corresponds up to the first node 20 to the signal flow according to Fig. 2, with the exception of a second node 42 newly added to the signal flow and the resulting consequences described below.

[0057] However, in the present exemplary embodiment according to the invention, the hearing instrument 1 now has a second filter 40, which is provided and configured to suppress the residual feedback (after compensation by the compensation signal c). This second filter 40 is generally applied to an intermediate signal z, which is derived from the input signal e (in which the speech contributions 30 propagated via the acoustic feedback path 16 are input) and in particular from the error signal e (which has already been cleaned of feedback by the compensation signal c). In the present case, the intermediate signal z is given by the output signal y resulting from the signal processing unit 10, to which, as described below, individual gain factors gj are applied frequency band by frequency band by the second filter 40 at the second node 42. The intermediate signal z thus corresponds to the output signal y after Fig. 1 and Fig. 2. However, in another embodiment not shown, the intermediate signal z can also be given by the error signal e.

[0058] The second filter 40 can also apply compression to the intermediate signal z in individual frequency bands and / or operate as a "true" adaptive filter, whereby individual filter coefficients can be determined, for example, using an NLMS algorithm. The second filter 40 and the control system explained below are preferably to be physically implemented in the signal processing device 10, but for reasons of clarity, are not shown in the figure. Fig. 3 separated from the “actual” signal processing (e.g. frequency band-wise amplification, especially depending on the wearer’s audiological requirements, etc.).

[0059] For the operation of the second filter 40, a quotient Q is formed from the error signal e and the compensation signal c for each frequency band. This quotient Q forms a control variable K for the second filter 40. First, the temporal signal components in the first input signal x1 or in the error signal e in which speech contributions 30 from the conversation partner are present are determined by the sign of the logarithm of the quotient Q (or equivalently, by a > / < comparison of the quotient with 1). For these, the second filter 40 is generally applied to the output signal y, whereby the frequency band-wise amplification factors gj are also dependent on the value of the quotient Q. However, if there are no speech contributions 30 from the conversation partner, it is also not necessary to suppress residual feedback, since the conversation partner will not perceive any echo.

[0060] The dependence of the application of the second filter 40 on the quotient Q as the control variable K in the respective frequency band is shown by Fig. 4 briefly outlined. Fig. 4 shows for an unspecified frequency band the said quotient Q in dB from the error signal e and the compensation signal c according to Fig. 3, plotted against a time axis Tj.

[0061] In a first time window T1, the value of the quotient Q is below 0 dB, meaning that the compensation signal c predominates. Since the error signal contains the speech contributions 22 of the wearer of the hearing instrument 1, while the compensation signal c, as an image of the feedback output signal y, can contain both speech contributions 22, 30 (depending on who is speaking), if the compensation signal c predominates over the error signal e, it can be assumed that essentially only speech contributions 30 of the conversation partner are present. In this case - i.e., for the first time window T1 - the second filter 40 is Fig. 3 is applied to the output signal y.

[0062] In a second time window T2, the value of the quotient Q is above 0 dB, meaning the error signal e predominates. The second filter 40 thus remains switched off (for the period between the first and second time windows T1, T2, the value of the quotient is exactly 0 dB; for this, it is assumed that neither the wearer nor the call partner is speaking; even for Q = 0 dB, the second filter 40 remains switched off).

[0063] In a first segment T3a of a third time window T3, the value of the quotient Q is initially below 0 dB; in a subsequent second segment T3b of the third time window T3, the quotient fluctuates strongly around 0 dB, before then assuming a stable value above 0 dB in a third segment T3c of the third time window T3. For the first segment T3a, as in the first time window T1, speech contributions 30 are assumed only from the conversation partner, and the second filter 40 is applied accordingly. For the third segment T3c, as in the second time window T2, speech contributions 22 are assumed only from the wearer of the hearing instrument 1, and the second filter 40 is switched off accordingly.

[0064] For the second segment T3b, the dominance of error signal e and compensation signal c alternates rapidly, both speech contributions 22, 30 coexist simultaneously (the sign of Q / [dB] then follows rather random fluctuations of both speech contributions 22, 30): The wearer and the conversation partner speak simultaneously or interrupt each other. Here, too, the second filter 40 can be operated in the time-frequency domain depending on the sign of Q / [dB]: In different frequency bands, frequency contributions of the two speakers will usually be found separated from each other in time (i.e., due to the frequencies of speech, it can be assumed that the wearer and his conversation partner rarely occupy a frequency bin of the time-frequency domain simultaneously). Fig. 4 is for the application of the second filter 40 according to Fig. 3 an upper limit Gr for the quotient Q is shown, above which the application is suspended.

[0065] The quotient Q as control variable K therefore provides in the hearing instrument 1 after Fig. 3 a switch for the application of the second filter 40. The gain factors gj can be, for example, continuously monotonic in Q in each frequency band (for example by directly applying the respective quotient Q as gain factor gj, or by scaling using an exponential function of Q).

[0066] In an alternative embodiment (not shown), the minimum value for the quotient Q (in dB) can also be determined as a reference value using a slow-reacting minimum tracker. From this reference value, a distance to a desired suppression is then determined, for example, at a determined minimum of Q min= -20 dB and a desired suppression of -35 dB, a distance of -15 dB corresponds to this. Consequently, with a negative value of the quotient (in dB), even with variations around this value, the constant distance value (see above), possibly provided with an additional safety buffer of, for example, another -5 dB, can be determined as the constant gain factor gj for the second filter 20.

[0067] The amplification factors gj are then applied, as described, to the intermediate signal z given by the output signal y at the second node 42, depending on the quotient Q = e / c as the control variable K. The intermediate signal z is branched off in a separate signal path for this purpose. By applying the amplification factors gj to the intermediate signal z, the transmission signal t is generated. The playback signal w is generated directly as the output signal y. In the transmission signal t, which is transmitted to the external communication device 4 and reproduced there by the loudspeaker 26, there are now no more speech contributions 30 (or only to a negligible extent; break in the dotted line) originating from the conversation partner. The conversation partner will no longer hear his speech contributions 30 via the loudspeaker 26 as an "echo", since these have now been essentially completely suppressed at the second node 42. Only the (in Fig. 1, but not in Fig. The speech contributions 22 of the wearer of hearing instrument 1 (shown in Figure 3) can still be heard by him.

[0068] Although the invention has been illustrated and described in detail by the preferred embodiment, the invention is not limited to the disclosed examples and other variations may be derived therefrom by those skilled in the art without departing from the scope of the invention. List of reference symbols 1 hearing instrument 2 hearing aids 4 external communication device 6 first input converter 8 Ambient noise 10 Signal processing device 12 output converters 14 Output sound 16 acoustic feedback path 18 first filter 20 first nodes 22 language contributions 24 Communication device 26 speakers 28 external playback sound 30 language contributions 32 microphone 40 second filter 42 second node c compensation signal e error signal gj gain factor Gr upper limit h Feedback K Control variable Q quotient t transmission signal T1-T3 first, second, third time window T3a-T3c first, second, third segment Tj Timeline w Playback signal x1 first input signal xe external input signal y output signal z intermediate signal

Claims

[1] Method for reducing echo in a hearing instrument (1), - wherein a first input signal (x1) is generated from an ambient sound (8) by an electroacoustic first input transducer (6) of the hearing instrument (1), - wherein an external input signal (xe) is received from an external communication device (4) by means of a communication device (24) of the hearing instrument (1), - wherein an output signal (y) is generated based on the first input signal (x1) and the external input signal (xe) of the hearing instrument (1), - wherein a compensation signal (c) for reducing echo and / or acoustic feedback (h) is generated in an adaptive first filter (18) on the basis of the output signal (y), - wherein an error signal (e) is generated based on the first input signal (x1) and the compensation signal (c), - wherein a control variable (K) is generated based on filter coefficients of the first filter (18) and / or based on a comparison of the error signal (e) with the compensation signal (c) and / or with the first input signal (x1), - wherein, depending on the control variable (K), a second filter (40) for suppressing a residual echo and / or a residual feedback is applied to an intermediate signal (z) derived from the input signal (x1), and thereby a transmission signal (t) is generated, and - wherein the transmission signal (t) is transmitted to the external communication device (4) by means of the communication device (24). [2] Method according to claim 1, wherein the second filter (40) is applied to the intermediate signal (z) as a function of a sign of the control variable (K) or a sign of the logarithm of the control variable (K). [3] Method according to claim 1 or claim 2, wherein the control variable (K) is formed on the basis of a quotient (Q) of the amplitudes, the absolute values ​​and / or the squares of the values ​​of the error signal (e) and the compensation signal (c). [4] Method according to claim 2 or claim 3, wherein the second filter (40) has a functional dependence on the control variable (K). [5] Method according to claim 4, wherein the second filter (40) has a continuously monotonic dependence on the control variable (K). [6] Method according to claim 4 or claim 5, wherein the second filter (40) is applied to the intermediate signal (z) as a function of the sign of the control variable (K) or the sign of the logarithm of the control variable (K). [7] A method according to claim 2 or claim 3, wherein the second filter (40) is operated with a predetermined parameter value. [8] Method according to claim 7, wherein the predetermined parameter value for the operation of the second filter (40) is independent of an amount of the control variable (K). [9] Method according to one of the preceding claims, wherein the second filter (40) applies a gain factor (gj) to the intermediate signal (z) to attenuate it. [10] Method according to one of the preceding claims, wherein compression is applied to the intermediate signal (z) by the second filter (40). [11] Method according to one of the preceding claims, wherein the second filter (40) is applied in the time-frequency domain to a number of frequency bands of the intermediate signal (z). [12] The method of claim 11, wherein the second filter (40) is applied to the intermediate signal (z) in a number of frequency bands as a second adaptive filter with more than one filter coefficient each. [13] Method according to one of the preceding claims, wherein non-linear processing is applied to the intermediate signal (z) or to a signal derived from the intermediate signal (z) to generate a playback signal (w). [14] A method according to claim 13, wherein frequency distortion is applied as non-linear processing. [15] Method according to one of the preceding claims, wherein the compensation signal (c) is generated from the output signal (y) or the transmission signal (t) to which the first filter (18) is applied, and wherein the first filter (18) is adapted based on the error signal (e). [16] Method according to one of the preceding claims, wherein the error signal (e) is generated by subtracting the compensation signal (c) from the first input signal (x1). [17] Method according to one of the preceding claims, wherein the error signal or the output signal (y) is used as the intermediate signal (z). [18] Hearing instrument (1), comprising - an electroacoustic first input transducer (6) for generating a first input signal (x1) from an ambient sound (8), - a communication device (24) for receiving an external input signal (xe) from an external communication device (4) and for sending a transmission signal (t) of the hearing instrument (1) to the external communication device (4), - a signal processing device (10) which is designed to generate an output signal (y) based on the first input signal (x1) and the external input signal (xe), wherein an adaptive first filter (18) is implemented in the hearing instrument (1), which is designed to generate a compensation signal (c) for reducing echo and / or acoustic feedback (h) based on the output signal (y), wherein the signal processing device (10) is further configured to - to generate an error signal (e) based on the first input signal (x1) and the compensation signal (c), and - to generate a control variable (K) based on filter coefficients of the first filter (18) and / or based on a comparison of the error signal (e) with the compensation signal (c) and / or with the first input signal (x1), and wherein a second filter (40) is implemented in the hearing instrument (1), which filter is designed to suppress a residual echo or a residual feedback (h) in an intermediate signal (z) derived from the input signal (x1), and thereby to generate the transmission signal (t).

Images