EP2981099B1 - Method and device for suppressing feedback - Google Patents

Description

The invention relates to a method for feedback suppression and a device for carrying out the method. In the method according to the invention, a feedback transfer function is estimated, coefficients of an adaptive filter are adjusted to suppress feedback and the adaptive filter is applied to a signal which is derived from an acoustic input signal of the acousto-electric converter.

Hearing aids are portable hearing devices that are used to care for the hard of hearing. In order to meet the numerous individual needs, different types of hearing aids such as behind-the-ear hearing aids (BTE), hearing aids with an external receiver (RIC: receiver in the canal) and in-the-ear hearing aids (ITE), e.g Concha hearing aids or canal hearing aids (ITE, CIC) provided. The hearing aids listed as examples are worn on the outer ear or in the auditory canal. Bone conduction hearing aids, implantable or vibrotactile hearing aids are also available on the market. The damaged hearing is stimulated either mechanically or electrically.

In principle, hearing aids have an input converter, an amplifier and an output converter as essential components. The input transducer is typically an acousto-electric transducer, e.g. B. a microphone, and / or an electromagnetic receiver, z. B. an induction coil. The output transducer is usually an electroacoustic transducer, e.g. B. miniature speakers, or as an electromechanical transducer, z. B. bone conduction headphones realized. The amplifier is usually integrated into a signal processing device. The energy is usually supplied by a battery or a chargeable accumulator.

Because of the close spatial proximity between the microphone and the electroacoustic output transducer, there is always a risk that an acoustic signal will be transmitted as sound through the air, whether through a ventilation opening, a gap between the wall of the auditory canal and the hearing aid device or an earpiece of the hearing aid device or is transmitted inside the hearing aid device or as structure-borne noise via the hearing aid device itself. If the overall gain of a feedback loop, which results from the signal processing in the hearing aid device and the damping on the feedback path between the output transducer and the microphone, is greater than 1, then with a suitable phase shift of a signal, especially if the phase shift is 0 or an integer multiple of 2 * Pi, oscillation can occur along this feedback loop, resulting in an uncomfortable howling sound for the wearer.

Various measures are known from the prior art for suppressing feedback noises in hearing aid devices. One possibility is to provide an adaptive filter in the hearing aid device, the coefficients of which are derived from a response function of the feedback path that is determined in different ways. In this case, the respective change in coefficients of the adaptive filter is determined by means of a mathematical method according to a normalized minimum deviation of the squared error (normalized least mean square, NMLS). The speed with which the adaptive filter can adapt is influenced by an increment μ. If the step size is large, the adaptive filter can follow quickly; if the step size is small, the filter maps the input function better with small changes.

From the publication Antweiler C, Schiffer A and Dörbecker M, "Accoustic Echo Control with Variable Individual Step Size", Proc. IWAENC, pages 15 to 18, Norway, 1995 it is known, for example, that the step size μ is used for coefficients that have a larger time delay are assigned to be weighted with an exponential decay depending on the time delay. This is derived from the general knowledge that excitation of a damped oscillation decays exponentially over time. Since real impulse responses are made up of a large number of different damped oscillations with different decay times, there are deviations.

From the pamphlet US 6,876,751 B1 a method for reducing feedback in a hearing aid device is known. An adaptive filter that approximates the feedback path is updated using the so-called Normalized Least Mean Square (NLMS) algorithm. The step size of the NLMS algorithm is controlled based on the combined signal power of the error signal and the input signal of the adaptive filter.

From the pamphlet Benesti, Sondhi, Huang, Handbook of Speech Processing, Chapter 6.6.4, page 114, Springer Verlag, 2008 it is known to weight a coefficient by a factor proportional to a previous value of the same coefficient. However, if the feedback path and thus the impulse response changes, the adaptive filter slowly converges for coefficients with previously small values.

The object of the present invention is therefore to provide a method and a device in which feedback suppression is improved.

According to the invention, this object is achieved by a method according to claim 1 and by a device according to claim 3.

The method according to the invention relates to a method for reducing feedback in a hearing aid device. The hearing aid device has an acousto-electrical converter, a signal processing device, a feedback suppression device and an electro-acoustic converter.

In a step of the method according to the invention, a first feedback transfer function is used at a first point in time determined. The feedback transfer function forms feedback paths from the signal processing device via the electro-acoustic converter, an acoustic signal path from the electro-acoustic converter to the acousto-electric converter and via the acousto-electric converter back to the signal processing facility. The acoustic signal path depends on the environment of the head and changes, for example, when the wearer moves. The determination can include, for example, measuring different feedback transfer functions in a laboratory or also estimation using approximation methods such as NLMS when the hearing aid aid is operated on the wearer's ear.

In one step of the method according to the invention, a weighted mean value function is determined as a function of amplitude values of the first feedback transfer function. For this purpose, an envelope function is formed for the amounts of the amplitudes or a function of the squares of the amplitudes smoothed using a low-pass filter or band-pass filter, which reflects an energy of the impulse response over a time delay in relation to the impulse excitation.

In particular, the impulse response parameters are resolved over a time delay with respect to the impulse excitation, i.e. different impulse response parameters are determined for different values for the time delay. In this case, individual impulse response parameters are preferably determined as a function of different function values of the enveloping function for the amounts of the amplitudes or the function of the squared amplitudes smoothed by means of a low-pass filter or band-pass filter. In particular, the impulse response parameters are determined from the weighted average function, which depends on the first feedback transfer function. In this case, the weighted mean value function forms a weighted mean value over the first feedback transfer function and further feedback transfer functions, with the mean value being preferably formed point by point for the individual time delays after which the feedback functions are resolved.

The impulse response parameters preferably have a direct dependence on the impulse response of a feedback path on which is mapped by the first feedback transfer function or by a weighted average function of several feedback transfer functions. In this case, the impulse response of a feedback path is given in particular by a time-resolved amplitude which has a signal excited in the feedback path by a test pulse.

In another step of the method, a second feedback transfer function is estimated using an adaptive filter. The estimation preferably takes place at a second, different point in time. In this case, coefficients of the adaptive filter for suppressing a feedback signal are updated in a determined manner as a function of the weighted mean value function.

For example, in an estimation method, a current estimation function is formed from estimated values from the past and an estimate of the deviation of the estimated values from the past from the actual values. In order to estimate an impulse response, it is possible, for example, to take account of components with different delays in different coefficients. The weighting of the change in the different coefficients can in turn be weighted as a function of empirical values that result from mean value functions of exemplary or past impulse responses.

By definition, the adaptation speed of the adaptive filter is the speed at which the adaptive filter reacts to changes in the feedback transfer function to be estimated and thus "adapts" it to the changes. With a high adaptation speed, the adaptive filter reacts quickly to changes in the image to be mapped by the feedback transfer function Feedback path, whereby suggestions caused by the changes can be quickly suppressed. With a low adaptation speed, however, the adaptive filter is more stable, so that due to the higher inertia in an output signal, audible artefacts can be better avoided by feedback suppression. By updating the coefficients of the adaptive filter in such a way that the adaptation speed is formed by a function of the impulse response parameters, the adaptation behavior can be controlled via the impulse response parameters.

In particular, the function of the impulse response parameters for the adaptation speed is such that for time delays with regard to an impulse excitation, in which there is a comparatively strong impulse response of a feedback path on which the impulse response parameters are based, the adaptive filter quickly adapts to changes in the feedback path, while the adaptive Filter adapts more slowly to changes in the feedback path in the case of time delays with regard to an impulse excitation in which there is no significant impulse response of a feedback path on which the impulse response parameters are based. This is achieved, for example, by using a monotonic function of the time-smoothed amplitude amounts of the impulse response in the underlying feedback path as the impulse response parameter, and the adaptation speeds at different time delays with regard to an impulse excitation are each formed by the same monotonic function of the corresponding impulse response parameter.

The result of this is that the adaptive filter, which estimates the second feedback transfer function using its coefficients, makes changes to the estimated feedback path particularly quickly where it has a high impulse response. In that the impulse response parameters are not derived from the second feedback transfer function itself, but rather from the first feedback transfer function or a weighted mean value function, which should preferably be selected as a typical representative of a feedback transfer function that is possible in the given hearing situation with a corresponding feedback path, incorrect adaptation, for example due to tonal excitations in the feedback path, can be avoided, since the updating of the coefficients is no longer just from the error-prone estimate, but now also depends on an external reference.

In another step of the method according to the invention, the adaptive filter is applied to a signal which is derived from an acoustic input signal of the acousto-electric converter. For example, it is conceivable to use the adaptive filter to filter out or suppress a feedback component from the acoustic signal in that the adaptive filter mixes the audio signal with a signal that is approximately identical to the feedback component and has an inverse sign.

By using the inventive method for determining the coefficients experiences from a past feedback transfer function in the form of weighting to determine a current set of coefficients, it advantageously allows a faster and more accurate estimation of the current feedback transfer function and thus a more effective and more accurate suppression of feedback while reducing Artifacts caused by the feedback suppression. The coefficients of the adaptive filter are advantageously adapted in such a way that rapid adaptation is ensured in those areas of the feedback impulse response which contain a great deal of energy, whereas areas with low energy are only subject to slow adaptation. Low-energy areas do not contribute to the risk of feedback-related howling, so it is important in these areas to ensure the greatest possible freedom from artefacts through slow adaptation. Using an enveloping function ensures that that regions near zero crossings in the feedback impulse response do not falsely result in slow adaptation. Averaging over time ensures that short-term fluctuations do not lead to incorrect adaptations.

The hearing aid device according to the invention for carrying out the method shares the advantages of the method according to the invention.

Further advantageous developments of the invention are specified in the dependent claims.

In a conceivable embodiment of the method according to the invention, a multiplicity of feedback transfer functions are determined at different points in time and the weighted mean value function is determined as a function of the multiplicity of feedback transfer functions.

It is thus advantageously conceivable for the feedback suppression device to form an average value function from feedback transfer functions over a longer period of time or, in particular, to take into account feedback transfer functions with very different properties.

In one possible embodiment of the method according to the invention, the first and second feedback transfer functions are determined by estimating the feedback transfer functions in the hearing aid device.

The hearing aid device can thus advantageously adapt to the wearer's environment during operation and offer him better functionality with less feedback and artifacts.

In a conceivable embodiment of the method, the first feedback transfer function is determined by measuring the feedback transfer functions.

Measurement advantageously makes it possible to detect specific hearing situations more precisely and also to provide an averaging function for the hearing aid device even before the wearer uses it for the first time, so that the wearer can use it without a training phase.

In a conceivable embodiment of the method, the feedback suppression device is implemented as part of the signal processing device, so that the signal processing device carries out the steps of the method.

In this way, the number of components in the hearing aid device can advantageously be reduced and synergies can be used when determining the coefficients, for example by accessing common data.

In a conceivable embodiment of the method according to the invention, this is carried out in a plurality of disjunctive or partially overlapping frequency ranges.

This enables the hearing aid device to react to different feedback conditions at different frequencies and to adapt the method to them. For example, because of higher damping of an excited oscillation at high frequencies, shorter filter lengths are conceivable, or a lower sampling rate at lower frequencies.

In one possible embodiment of the method, after the step of applying the adaptive filter, the step continues with the step of determining a weighted averaging function, the second feedback transfer function being used together with the first feedback transfer function to form the weighted averaging function, and a new second feedback transfer function being estimated.

In this way, the adaptive filter and the increment can be continuously updated in an advantageous manner, so that fast convergence with few artefacts can be achieved even with changing feedback conditions.

In a preferred embodiment of the method, the impulse response parameters are determined by a smoothing function of the amplitude amounts as a function of the first feedback transfer function. The first feedback transfer function and a weighted mean value function of various feedback transfer functions are included as a function of the first feedback transfer function. In particular, the feedback transfer function or the weighted mean value function is designed as an impulse response function, so that a smoothing function of the amplitude amounts represents a preferred temporal smoothing of the amounts of the impulse responses of the feedback path corresponding to the feedback transfer function at different time delays in relation to the impulse excitation. The smoothing function is preferably in the form of an envelope of the amplitude values. The envelope is preferably normalized with respect to a reference value dependent on the adaptive filter or with respect to a maximum value for the amplitude magnitudes. By preferably smoothing a function of the amplitude magnitudes over time, on which the impulse response parameters are based, it can be achieved that an impulse response parameter is not influenced by a zero crossing of an oscillating amplitude with strong absolute values in the corresponding range that coincidentally falls on the corresponding time delay an adaptation speed would not erroneously be chosen too low for the corresponding time delay.

In a further expedient embodiment, the adaptation speed of the adaptive filter is reduced in this range for a monotonous decrease in the amplitude values in the argument of the smoothing function via the impulse response parameters.

The first feedback transfer function or the weighted mean value function, on which the impulse response parameters are based, preferably represents a typical representative of a feedback transfer function that is possible in the given hearing situation with a corresponding feedback path from, this means that such a feedback path usually supplies correspondingly decreasing contributions to the feedback in this range. Accordingly, the adaptation speed in the estimation of the second feedback transfer function is also reduced for these areas.

Advantageously, this can result in the adaptation speed not being increased unnecessarily in these areas as a result of incorrect adaptation, for example due to a tonal excitation in the input signal, which could lead to unwanted artefacts in an output signal.

In a further advantageous embodiment variant, the coefficients of the adaptive filter are updated using an NLMS algorithm, the entries of a vector-valued step size of the NLMS algorithm for updating the coefficients of the adaptive filter being formed using the impulse response parameters, and the impulse response parameters using a smoothing function of the magnitudes of the amplitudes as a function of the first feedback transfer function.

An NLMS ("Normalized Least Mean Squares") algorithm is a filter which is used particularly often to suppress feedback and which updates existing coefficients of the filter as a function of an output signal and an error signal via a step size. The individual coefficients of the filter are then listed with their corresponding time order - i.e. the time delay with regard to an impulse excitation - applied to a signal derived from the input signal. By forming the step size for updating the coefficients as a vector based on the impulse response parameters, the step size with which each coefficient is updated for an adaptation to a change can be selected as a function of the impulse response in the feedback path, so that on the one hand the adaptation fast enough to capture sudden changes caused by excitations in the input signal, while avoiding artifacts.

The device according to the invention shares the advantages of the method according to the invention.

The properties, features and advantages of this invention described above, and the manner in which they are achieved, will become clearer and more clearly understood in connection with the following description of the exemplary embodiments, which are explained in more detail in connection with the drawings.

Fig. 1

eine beispielhafte schematische Darstellung eines erfindungsgemäßen Hörhilfegeräts in Funktionsblöcken;

Fig. 2

ein schematisches Ablaufdiagramm eines erfindungsgemäßen Verfahrens

Fig. 3

ein Diagramm mit beispielhaften Impulsantworten von Rückkopplungspfaden;

Fig. 4

ein Diagramm mit beispielhaften Mittelwertsfunktionen zu den Impulsantworten;

Fig. 5.

ein Diagramm mit beispielhaften Gewichtungskoeffizienten;

Fig. 6

zwei Diagramme die vergleichsweise Reaktionsfähigkeit einer Adaptation mit vektorwertigen Schrittweiten; und

Fig. 7

ein Diagramm die vergleichsweise Stabilität einer Adaptation mit vektorwertigen Schrittweiten.

Show it:

1

an exemplary schematic representation of a hearing aid device according to the invention in functional blocks;

2

a schematic flowchart of a method according to the invention

3

a diagram with exemplary impulse responses of feedback paths;

4

a diagram with exemplary mean value functions for the impulse responses;

figure 5

a diagram with exemplary weighting coefficients;

6

two diagrams of the comparative responsiveness of an adaptation with vector-valued step sizes; and

7

a diagram of the comparative stability of an adaptation with vector-valued step sizes.

1 shows a hearing aid device 100 according to the invention as a schematic representation in functional blocks. The hearing aid device according to the invention has an acousto-electrical converter 2, which converts a mechanical vibration, usually recorded as airborne sound d(k), into an electrical signal m(k). The acousto-electric converter 2 is usually one or more microphones, usually capacitive and sometimes also micromechanically designed as a MEMS microphone made of silicon. It is conceivable that the signals from a plurality of microphones are interconnected as a microphone with a directional characteristic. In this case, the signal m(k) is preferably a signal with a directional characteristic.

The hearing aid device 100 also has a signal processing device 3, which is designed to amplify an incoming signal e(k), preferably in a frequency-dependent manner, so that a hearing impairment of a wearer can be compensated for and soft sounds below the wearer's hearing threshold are raised to a range above the wearer's hearing threshold . For this purpose, the signal processing device 3 can have a filter bank, for example.

Conceivable additional functions of the signal processing device 3 are dynamic compression, classification of hearing situations, noise suppression, control of directional characteristics of the microphone, binaural signal processing if the hearing aid device 100 has a signal connection with a second hearing aid device 100 via a communication interface (not shown).

Furthermore, the hearing aid has an electro-acoustic converter 4, which is designed as a loudspeaker or earpiece. In the case of a behind-the-ear hearing aid device 100, the electro-acoustic converter 4 can be arranged in a housing behind the ear and the sound can be transmitted via a sound tube to an earpiece in the wearer's auditory canal. It is also conceivable with a behind-the-ear hearing aid that the electro-acoustic converter 4 is arranged in the ear canal of the wearer and receives a signal to be output via an electrical signal connection. Finally, the hearing aid device 100 can also be an in-the-ear or CiC (complete in channel) hearing aid device, so that all components of the hearing aid device are arranged on or in the wearer's auditory canal.

There is always a feedback path g(k) between the electro-acoustic converter 4 and the acousto-electric converter 2 via which acoustic energy can be transmitted back to the acousto-electric converter 2 . The feedback path can be formed through the air, for example through a gap between the auditory canal and a seal of the auditory canal (e.g. an ear shell or an "ear dome") or as structure-borne noise transmission through a housing of the hearing aid device 100. A combination of both is also conceivable Ways. The properties of the feedback path are also dependent on the surroundings of the wearer's head, for example a reflection on a wall or a car window or a telephone receiver close to the ear. Attenuation of the feedback path is strongly frequency-dependent. If the overall amplification via the electro-acoustic converter 4, the feedback path g(k), the acousto-electric converter 2 and the signal processing 3 is greater than 1, taking the phase into account, feedback whistling occurs.

In order to prevent or at least reduce such feedback whistling, the hearing aid device 100 has a feedback suppression device 6, which is shown in FIG Embodiment has an adaptive filter 7 and a mixer 8. The adaptive filter 7 receives the input signal e(k) fed to the signal processing device 3 via a first signal line 11 and the signal x(k) output by the signal processing device via a second signal line 9 . Furthermore, the adaptive filter 7 is connected to the signal processing device 3 via a third signal line 10 in order to detect its effect on the processing of the input signal e(k). This can be done, for example, by transmitting processing parameters.

The adaptive filter 7 processes the supplied signals to form a compensation signal c(k), which is mixed with the electrical signal m(k) via a mixer 8 in order to reduce feedback. More information on the type of generation of the compensation signal c(k) is provided below Fig.2 explained in more detail.

It should be noted that in particular the division of functionalities in the 1 is only exemplary. It is also conceivable that the feedback suppression unit 6 is not, as in 1 shown, is designed as separate function blocks 7 and 8, but only as program-controlled functions in the signal processing device 3, or as hardware-implemented circuits therein. It is also conceivable that the adaptive filter 7 does not filter by generating a compensation signal c(k) and mixing it with the electrical signal m(k) in order to reduce a feedback signal through destructive interference, but rather as a subtractive filter itself in the signal path m (k) is provided. The signals x(k) and e(k) can also be taken from the signal flow at different points without departing from the principle of the invention. It is conceivable, for example, for the adaptive filter 3 to determine the influence of the signal processing device 3 itself by comparing the signals e(k) and x(k). However, it is also conceivable that the adaptive filter 7 transmits all information about the function of the signal processing system 3 the signal connection 10 receives, but only one of the signals e(k) or x(k).

2 shows an exemplary sequence of a method according to the invention on a hearing aid device 1 .

In a step S10, a first feedback transfer function at a first point in time on a feedback path from the signal processing device 3 via the electro-acoustic converter 4, an acoustic signal path g(k) from the electro-acoustic converter 4 to the acousto-electric converter 2 and via the acousto-electric converter 2 back to the signal processing device 3 is detected.

It is conceivable for the feedback transfer function to be measured in a measuring box by a hearing aid acoustician or in a laboratory by measuring the wearer or an artificial head. In these embodiments, the feedback transfer function can be measured more accurately since the input and output signals can be externally detected and processed together. It is conceivable to represent typical listening environments, such as talking on a mobile phone or sitting in a car with your ear close to a window.

Preferably, multiple feedback transfer functions are measured for typical environments.

However, it is also conceivable for the feedback transfer functions to be estimated in the hearing aid device itself when it is being worn, i.e. to be detected by the approximation functions explained in step S30 or S30'. The feedback transfer functions recorded in this way advantageously have no influence from the measurement environment and can correspond to everyday situations of the wearer.

3 shows two exemplary impulse responses as a possible form of representation of a feedback transfer function equivalent to each other in the sense that one can be clearly derived from the other using mathematical methods. The x-axis shows the time in multiples of a sampling cycle, and the y-axis shows a normalized amplitude. The x-axis indicates a time delay compared to an excitation pulse.

In a step S20, a weighted mean value function is determined from the detected first feedback transfer function as a function of amplitude values of the first feedback transfer function. In a step S20', a plurality of impulse response parameters are determined as a function of amplitude values of the first feedback transfer function. If step S20' takes place as an alternative to step S20, then the impulse response parameters are determined directly from the feedback transfer function detected in step S10. If step S20′ takes place directly after step S20, then the impulse response parameters are determined from a weighted mean value function of a plurality of feedback transfer functions, which include the first feedback transfer function detected in step S10.

4 shows a function for each impulse response that is generated by normalizing a function depending on the amplitude values. The functions therefore only have a positive sign. For the large amplitudes at the beginning, the function value is set equal to 1 in the sense of a limitation.

A mean value can take place in the sense of a temporal smoothing of the feedback transfer function, for example by forming an envelope of the positive amplitudes. A low-pass or band-pass using a function of the square of the amplitude is also conceivable.

A mean can also be used in the sense of an arithmetic mean or other averaging, for example by adding formed from a plurality of function values of different feedback transfer functions and parts by the number of functions detected, provided that a plurality of feedback transfer functions were detected. This can be done, for example, by measuring or by iterating the method over a plurality of the feedback transfer functions. However, other forms are also conceivable, such as weighting a function when averaging as a function of the age of the corresponding feedback transfer function.

If the detected feedback transfer function of step S10 is a measured function, then the mean value function can already be calculated outside of the hearing aid device 100 in a measuring device and transmitted to the hearing aid device 100 . If, on the other hand, it is a feedback transfer function estimated in the hearing aid device 100, then the weighted mean value function is preferably determined in the hearing aid device 100, e.g. by the feedback suppression device 6.

In a step S30 or S30' of the method according to the invention, a second feedback transfer function is estimated.

The adaptive filter 7 preferably models the time-dependent feedback transfer function as a time-dependent impulse response g(k) of the feedback path.

An example of an estimation method is updating coefficients of an adaptive filter using the NLMS algorithm. From a value at a time k, the value at a time k+1 is estimated using the following formula: $$\mathrm{H}\left(\mathrm{k}+1\right)=\mathrm{H}\left(\mathrm{k}\right)+\mathrm{\mu }\left[\left(\begin{array}{cc}\mathrm{e}*\left(\mathrm{k}\right)& \mathrm{x}\left(\mathrm{k}\right)\end{array}\right)/\left(\begin{array}{cc}\mathrm{x}*\left(\mathrm{k}\right)& \mathrm{x}\left(\mathrm{k}\right)\end{array}\right)\right]$$

In this case, k indicates a discrete time scale, x is the input value of the feedback suppression device, e=m−c is the error signal given as a difference of the microphone signal m and the compensation signal c, µ is a step size that controls an adaptation speed of the filter, and * denotes the complex conjugate of a value. Here h, x and µ are vectors in a space whose dimensionality is given by the length of the filter or the number of coefficients: h(k) = [h0(k), h1(k), h2(k), . .., hN(k)], where N is the number of coefficients in the model of the estimated function.

• S. Haykin, Adaptive Filter Theory. Englewood Cliffs, NJ: Prentice-Hall, 1996 .
• Toon van Waterschoot and Marc Moonen, "Fifty years of acoustic feedback control: state of the art and future challenges", Proc. IEEE, vol. 99, no. 2, Feb. 2011, pp. 288-327

See also:

• S. Haykin, Adaptive Filter Theory. Englewood Cliffs, NJ: Prentice-Hall, 1996 .
• Toon van Waterschoot and Marc Moonen, "Fifty years of acoustic feedback control: state of the art and future challenges", Proc. IEEE, vol. 99, no. 2, Feb. 2011, pp. 288-327

• LMS - Least mean squares
• RLS - Recursive least squares
• Affine Projection

Other conceivable methods for estimating a feedback transfer function are:

• LMS - Least Mean Squares
• RLS - Recursive least squares
• Affine projection

The coefficients of the adaptive filter for suppressing a feedback signal are adapted to the second feedback transfer function or, in other words, the feedback transfer function is modeled by the coefficients, with a change in the coefficients depending on the mean value function or the impulse response parameter being weighted. To adapt the coefficients, a correction value is weighted with a weighting factor or an increment. In the embodiment shown, this weighting takes place via the increment μ, which, as shown above, is included in the estimation of the feedback transfer function modeled by the coefficients. The weighting factor is derived from the averaging function over the impulse response parameters. In the simplest case, it could be the value of an in 4 shown mean function itself. The value of a weighting factor µ(k) is then, for example, a function value of an in 4 shown function for the value k in the x-axis.

Preferably, however, as in figure 5 an increment from the mean function of the 4 derived. In figure 5 instead of a linear, normalized scale to 1, a scale according to the logarithm of 10 log ₁₀ is plotted. In this way, the dynamic range of the step size is much larger, so that for large values of the impulse response in 3 Fast convergence is achieved, while small values result in high adjustment accuracy and therefore low artefacts.

However, it is also conceivable that in a method according to the invention the estimation of the second feedback transfer function takes place separately from a weighting of the coefficients one after the other.

Finally, in a step S40, the adaptive filter is applied to a signal derived from an acoustic input signal of the acousto-electric converter. Derived is understood to mean any signal processing that is conceivable in a hearing aid device, such as A/D conversion, amplification, also frequency-dependent, formation of a directivity or other functions that are possible in the signal processing 3 . In the 1 the application of the filter is represented by the compensation signal c(k), which represents an estimated feedback signal and is added to the signal m(k) of the microphone with the opposite sign, so that ideally the signal of the adapted filter and the feedback component of the microphone signal m( k) cancel.

In a preferred embodiment of the method, this is continued after step S40 with step S20, wherein the second feedback transfer function is used together with the first feedback transfer function to form the averaging function and a new second feedback transfer function is estimated in step S30.

In a preferred embodiment of the method according to the invention, steps S10 to S40 are each carried out in separate or only partially overlapping frequency bands, so that different feedback conditions in different frequencies can be optimally suppressed in each case. For this purpose, for example, a filter bank can be provided in the feedback suppression device 6 or a filter bank in the signal processing device 3 can also be used.

When suppressing feedback using an adaptive filter, an excitation in the form of a tonal input signal can lead to incorrect adaptation. In the example of the cited NLMS algorithm, the adaptive filter supplies the feedback transfer function of the respective feedback path as a solution, to which an error term that depends on the autocorrelation of the input signal is added. As a result of the comparatively high autocorrelation of a tonal input signal, in this case it is usually not possible to adequately suppress incorrect adaptation to the excitation in the form of the tonal input signal using conventional means.

The method described now provides a satisfactory solution for this. Misadaptations as a result of excitations in the input signal are significantly suppressed, while the adaptation speed for normal changes in the feedback path is nevertheless sufficiently high. A high stability of the feedback suppression is thus achieved, which has an improved sound quality, without impairing the responsiveness with regard to changes in the feedback path. It is thus no longer necessary to make a balancing trade-off between sound quality and the ability to adapt to changes in the feedback path.

The behavior or the ability to react to changes in the feedback path, which the method allows, is shown in two diagrams in 6 shown. The diagrams each show the system distance, which is defined as ∥ g ( k ) - h ( k )∥/∥ g ( k )∥ , plotted against a time axis scaled in seconds. The system distance is a measure of the extent to which the coefficients h(k) of the adaptive filter correspond to the actual impulse response g(k) in the feedback path. A good match is characterized by values close to zero for the system distance. The excitation underlying the feedback path is white noise. For the top graph, a uniform step size μ was used in each case when updating the coefficients h(k) of the adaptive filter. For the lower graph, the step width μ was adapted to the impulse response of a typical feedback path over the individual coefficients in the manner described when updating the coefficients.

After 2.5 seconds there is an instantaneous change in the feedback path. From the respective diagram it can be read that the responsiveness to this change in the feedback path is not impaired by the use of individual step sizes for the different coefficients h(k) of the adaptive filter, although this significantly reduces the step size for a large part of the coefficients. This is because said reduction in step size - and hence reduction in filter responsiveness - occurs for coefficients which represent areas of low impulse response in a typical feedback path and therefore contribute only insignificantly to the overall performance of the feedback path.

The improvement in the stability of the feedback suppression, i.e. in particular the reduction in incorrect adaptations, by updating the coefficients h(k) of the adaptive filter using individual step sizes is illustrated using the diagram in 7 clear: Again, the system instance is plotted against a time axis scaled in seconds, with the three scenarios presented being given by: the classic NLMS algorithm and updating the coefficients with a constant increment (upper line 18), updating the coefficients using individual, but not time-dependent step sizes (middle line 19), and an update of the coefficients by individual, time-dependent step sizes depending on a feedback path "learned" by weighted averaging (lower line 20).

In the first case, as can be seen from the system instance represented by the top line 18, significant misadaptations occur throughout the period. The mean for the system distance is 0.98. The individual increments used for the second case (middle line 19) significantly reduced the incorrect adaptations, the mean system distance has a value of 0.40. By adapting the individual increments to the "learned" feedback path, as is done in the third scenario (lower line 20), the incorrect adaptations could be further reduced, with the mean value for the system instance now being only 0.14. The only serious misadaptation, caused by a drastic change in the feedback path, is found here at a point in time of about 4.3 seconds. However, the system distance, which represents the misadaptations, can be adjusted to this point in time for the other two scenarios for scaling reasons using the diagram of 7 cannot be reproduced at all. It is thus clear that the proposed method significantly improves the stability when suppressing the feedback.

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 therefrom by those skilled in the art without departing from the scope of the invention.

Claims (5)

1. A method for reducing feedback in a hearing aid device (100), the hearing aid device (100) comprising an acousto-electric transducer (2), a signal processing device (3), a feedback suppression device (6) and an electro-acoustic transducer (4), the method comprising the steps of:

   (S10) determining an impulse response function of a first feedback transfer function on a feedback path from the signal processing device (3) via the electro-acoustic transducer (4), an acoustic signal path from the electro-acoustic transducer (4) to the acousto-electric transducer (2) and via the acousto-electric transducer (2) back to the signal processing device (3);

   (S20) determining a weighted average function in dependence on amplitude absolute values of the impulse response function of the first feedback transfer function, wherein as the weighted average function, an envelope function for the absolute values of the amplitudes is formed, or a low-pass or band-pass smoothed function of the amplitude squares of the impulse response function reflecting an energy of an impulse response over a time delay with respect to an impulse excitation;

   (S30) estimating a second feedback transfer function of the feedback path by means of an adaptive filter (7), wherein vector-valued coefficients h(k)= [h0(k), h1(k), h2(k), ..., hN(k)] of the adaptive filter (7) of length N are updated depending on the weighted average function, wherein, for a time k for an adaptation of the vector-valued coefficients h(k) → h(k+1) of the adaptive filter, a respective correction value is weighted with a step size µ, wherein the step size µ is given by a vector of length N which depends on the weighted average value function;

   (S40) applying the adaptive filter (7) to a signal derived from an acoustic input signal of the acousto-electric transducer (2) in order to reduce a feedback.
2. The method according to claim 1, wherein the method is carried out in a plurality of disjoint or only partially overlapping frequency ranges.
3. A hearing aid device, the hearing aid device (100) comprising an acousto-electric transducer (2), a signal processing device (3), a feedback suppression device (6) and an electro-acoustic transducer (4), wherein the hearing aid device (100) is adapted to:

   determine an impulse response function of a first feedback transfer function on a feedback path from the signal processing device (3) via the electro-acoustic transducer (4), an acoustic signal path from the electro-acoustic transducer (4) to the acousto-electric transducer (2) and via the acousto-electric transducer (2) back to the signal processing device (3);

   determine a weighted average function as a function of amplitude absolute values of the impulse response function of the first feedback transfer function, wherein as the weighted average function, an envelope function for the absolute values of the amplitudes is formed, or a low-pass or band-pass smoothed function of the amplitude squares of the impulse response function reflecting an energy of an impulse response over a time delay with respect to an impulse excitation;

   estimate a second feedback transfer function of the feedback path by means of an adaptive filter (7), determine vector-valued coefficients h(k)= [h0(k), h1(k), h2(k), ... hN(k)] of the adaptive filter (7) of length N as a function of the weighted average function, wherein for a time k for an adaptation of the vector-valued coefficients h(k) → h(k+1) of the adaptive filter, a respective correction value is weighted with a step size µ, wherein the step size µ is given by a vector of length N which depends on the weighted average function;

   apply the adaptive filter (7) to a signal derived from an acoustic input signal of the acousto-electric transducer (2) in order to reduce a feedback.
4. The hearing aid device according to claim 3, wherein the feedback suppression device (6) is part of the signal processing device (3).
5. The hearing aid device according to any one of claims 3 or 4, wherein the hearing aid apparatus is configured to determine another weighted average function after applying the adaptive filter, wherein the second feedback transfer function is used together with the first feedback transfer function to form the weighted average function, and the hearing aid device is adapted to estimate a new second feedback transfer function.

Images