Method for determining a maximum gain in a hearing device as well as a hearing device

The present invention is related to a method for determining a maximum gain in a hearing device according to the pre-amble of claim 1 as well as to a hearing device according to the pre-amble of claim 8.

The feedback stability of a hearing device is a crucial variable in the fitting of the hearing device to the users hearing loss and hearing preferences. The feedback stability depends on a couple of factors, e.g. the acoustic path from the receiver back to the microphone, including ear geometry, type of ear shell, vent size, tubing, etc., the mechanical stability of the hearing device housing, especially the mechanical coupling between receiver and microphones, and electromagnetic couplings.

Hence, the fitting software offers (or requires) to test the feedback stability with a feedback test. Different methods are known to perform this task:

A first method is called "Direct Method" and is implemented by increasing the gain in every frequency band until the system becomes unstable. The gain at which just no feedback occurs is then used as the maximum gain in the corresponding frequency band.

A second method is called "Negative-Slope Method" and is described in US-7 010 135 B2. The known technique uses a gain curve (i.e. gain vs. input level) having a negative slope in each band. Because of a high gain at a low input signal, feedback builds up and the input level increases. As a result thereof, the gain decreases until a stable state is reached. The gain at this stable point is the maximum stable gain, i.e. the feedback threshold.

A third method is called "Open-Loop Identification" and is, for example, described in the publication entitled "Adaptive Filter Theory" by S. Haykin (Prentice Hall, 1996) . The loop consisting of a signal processing unit in the hearing device and the feedback path is opened. While applying a probe signal at the output of the signal processing unit of the hearing device, the response at the input of the signal processing unit is measured. By way of correlation (or adaptive filtering) the feedback threshold can be determined. Robustness against environment sound can be achieved through longer averaging times and pseudo-noise techniques .

A fourth method is called "Closed-Loop Identification": While the hearing device is in normal operation, a probe signal is preferably injected at the output of the hearing device. The identification techniques are the same as for the third method described above. The accuracy is though lower, because of the closed-loop operation. Consequently, the signal level of a necessary probe signal has to be

rather high so that it is often perceived as uncomfortably loud.

The known techniques have at least one of the following drawbacks:

- An interruption of normal operation of the hearing device is necessary.

- An explicit measurement is necessary, i.e. an action is required by the fitter.

- Because of high signal levels that occur during the feedback test, the hearing device user is discomforted during the feedback test.

- An adaptation to slowly changing feedback conditions due to a reduced fit of the ear shell, due to dirt in the vent, at the receiver sound outlet or on the microphone cover element, etc. is not possible.

- The addition of the level of noise and gain is less than the maximum power output. - A principal susceptibility to an environment noise level must be accepted.

It is therefore one object of the present invention to provide a method for determining a maximum gain in a hearing device, which method does at least not have one of the above-mentioned disadvantages.

This and other objects are accomplished by the measures specified in the characterizing part of claim 1. Additional

- A -

embodiments of the present invention as well as a hearing device are specified in further claims.

The present invention is firstly directed to a method for determining a maximum gain that is applicable in a forward signal path starting at a microphone and ending at a receiver of a hearing device. The maximum gain is a gain value at which just no feedback occurs. The method according to the present invention comprises the steps of: - estimating an estimated feedback transfer function characterizing a feedback signal path starting at a receiver and ending at a microphone of the hearing device, while the hearing device is inserted into an ear of a hearing device user, and - adapting the estimated feedback transfer function as a result of a changing feedback signal path by applying an adaptive algorithm.

The method according to the present invention is characterized by determining the maximum gain from the estimated feedback transfer function.

An embodiment of the present invention is characterized in that the step of determining the maximum gain comprises:

- extracting coefficients of the estimated feedback transfer function, and

- determining the maximum gain from the extracted coefficients .

A further embodiment of the present invention is characterized by further comprising the step of averaging the extracted coefficients.

A further embodiment of the present invention is characterized by adjusting a gain in the forward signal path of the hearing device in accordance with the determined maximum gain.

A further embodiment of the present invention is characterized by performing the step of adjusting the gain in the forward signal path during regular operation of the hearing device.

A still further embodiment of the present invention is characterized by determining the maximum gain during regular operation of the hearing device. The invention allows tracking the long-term changes of the feedback stability.

A further embodiment of the present invention is characterized by controlling the step of determining the maximum gain by one of several of the following criterions:

- a preset gain in the forward signal path; - input level at the microphone of the hearing device;

- variance of the coefficients;

- histogram of the coefficients;

- predefined point in time.

Secondly, a hearing device with a forward signal path comprising a microphone, a gain unit and a receiver, and with an adaptive filter comprising an estimated feedback path unit, an LMS unit and a subtraction unit, the LMS unit being operationally connected to the estimated feedback path unit, an output signal of the gain unit being operationally connected to the LMS unit, to the estimated feedback path unit as well as to the receiver. An output of the estimated feedback path unit is operationally connected to a first input signal of the subtraction unit, while an output signal of the microphone is operationally connected to a second input signal of the subtraction unit in order that the output signal of the estimated feedback path unit is subtracted from the output signal of the microphone. Furthermore, an output of the subtraction unit is operationally connected to the gain unit. The hearing device is characterized in that the estimated feedback path unit is operationally connected to an extraction unit, by which a maximum gain is determined at which just no feedback occurs.

An embodiment of the hearing device according to the present invention is characterized in that the extraction unit is a means for extracting coefficients of an estimated feedback transfer function estimated in the feedback path unit .

A further embodiment of the hearing device according to the present invention is characterized in that the extraction unit comprises a preprocessing unit and a conversion unit.

A further embodiment of the hearing device according to the present invention is characterized in that the extraction unit comprises a preprocessing unit for extracting coefficients of the estimated feedback transfer function and a conversion unit for determining the maximum gain from the extracted coefficients.

A further embodiment of the hearing device according to the present invention is characterized in that the extraction unit is operationally connected to the gain unit.

A further embodiment of the hearing device according to the present invention is characterized by further comprising a control unit that is operationally connected to the extraction unit.

The present invention is further explained in more detail by referring to drawings illustrating exemplified embodiments of the present invention.

Fig. 1 schematically shows a block diagram of a known hearing device.

Fig. 2 schematically shows a block diagram of a hearing device with an extraction unit according to the present invention.

Fig. 3 shows a block diagram of a first embodiment of the extraction unit according to Fig. 2.

Fig. 4 shows a block diagram of a second embodiment of the extraction unit according to Fig. 2.

Fig. 5 shows a block diagram of a third embodiment of the extraction unit according to Fig. 2.

Fig. 6 shows the block diagram of the first embodiment of the extraction unit being partially realized outside a hearing device.

Fig. 7 shows a simplified block diagram of an inventive hearing device that does not automatically adjust a gain in a main signal path of the hearing device.

Fig. 8 shows a simplified block diagram of an inventive hearing device that automatically adjusts the gain in the main signal path of the hearing device.

Fig. 1 shows a block diagram of a known hearing device with a forward signal path comprising a microphone 1, a time-to- frequency domain transfer unit 2, a gain unit 3, a decorrelation unit 4, a frequency-to-time domain transfer unit 5, a limiting unit 6 and a loudspeaker 11, often called receiver in the technical field of hearing devices. In addition to the forward signal path, a feedback signal path 15 is indicated comprising a feedback transfer unit 12 with a feedback transfer function F(z). The feedback signal

_ Q _

path 15 represents an acoustic path that starts at the receiver 11 and ends at the microphone 1. The feedback transfer function F(z) is not known a priori and depends on the ear geometry, type of ear shell, vent size, tubing, mechanical couplings, etc. As has been pointed out in the introductory part, the feedback transfer function F(z) has a direct influence on the maximum gain that can be adjusted in the hearing device, i.e. in the gain unit 3.

For determining a maximum gain that is adjusted in the gain unit 3, the feedback transfer function F(z) is estimated using an adaptive filter. In fact, the feedback transfer function F(z) in the feedback signal path 15 is estimated to obtain an estimated feedback transfer function F ¹ (j, k) . Thereto, the very well known technique for adaptive feedback canceller is applied using a LMS- (Least Mean Square) -algorithm for minimizing the error of the adaptation.

The LMS-algorithm is implemented in a LMS unit 10, to which the delayed output signal U(j, k) of the limiting unit 6 is fed via a further frequency-to-time domain transfer unit 8. In addition, a difference signal E(j, k) , which is fed to the gain unit 3 as well as to a multiplication unit 13, is also inputted to the LMS unit 10. From the two input signals, coefficients for the estimated feedback path transfer function F' (j, k) can be calculated. As long as the output signal E(j, k) contains a portion of the feedback signal of the feedback signal path 15, the

estimation of the estimated feedback path transfer function F' (j, k) can be further improved.

The estimated feedback path transfer function F' (j, k) or F _j [k] mimics the external feedback path 15 - i.e. its transfer function F(z) - and can therefore be described by its coefficients, which are called FC coefficients hereinafter. It is pointed out that the adaptive filter can be implemented in the frequency or in the time domain.

The FC coefficients are updated with a fast tracking speed with the adaptive filter algorithm. The movement of the FC coefficients follows each change in the feedback path and also possesses natural fluctuations. In addition, the adaptive filter algorithm is not perfect such that temporarily misadjusted FC coefficients may follow. This is especially true if the loop gain is low, which will be further explained in more detail below.

The decorrelation unit 4 is used to prevent a correlation between the error signal E(j, k) and the signal U(j, k) and is, for example, a frequency shifter as known, for example, from the paper entitled "Adaptive feedback cancellation with frequency compression for hearing aids" by Harry Alfonso L. Joson et al. (J. Accoust. Soc. Am. 94 (6),

December 1993, pp. 3248-3254) . Other implementations for the decorrelation unit 4 use phase shifting or non-linear frequency processing.

The use of the decorrelation unit 4 further stabilizes the operation of the adaptive filter such that the resulting FC coefficients are less erroneous.

Fig. 2 shows a block diagram of a hearing device according to the present invention. In addition to the blocks shown in Fig. 1, an extraction unit 16 is provided, in which the maximum gain is determined that can be applied in the gain unit 3.

In Fig. 3, the extraction unit 16 of Fig. 2 is depicted in more detail. The FC coefficients of the extraction unit 16 (Fig. 2) are fed to a preprocessing unit 17 that is connected to a conversion unit 18. The conversion unit 18 determines a feedback threshold or maximum gain that is adjusted in the gain unit 3 (Fig. 2) . In addition, a control unit 19 is provided, which is fed by further parameters P taken into account while determining the feedback threshold and maximum gain, respectively. The further parameters P are also used to optimize the determination of the feedback threshold and maximum gain, respectively.

The further parameters P might be one or several of the following:

- input level at the microphone 1 of the hearing device;

- momentary gain value of the gain unit (3) ;

- predefined gain value of the gain unit (3);

- pre-calculated maximum gain based on standard values providing information about type of hearing device, used ear shell, etc.;

- current value of the estimated frequency transfer function F ¹ (j, k) .

In a first embodiment of the present invention, the preprocessing unit 17 is used to smooth FC coefficient fluctuations, i.e. an averaging of the FC coefficients is performed in the preprocessing unit 17 in order to get rid of fast changing FC coefficients.

The averaged FC coefficients F _j [k], wherein j is the frame number and k is the frequency bin, can be seen as a frequency-domain representation, or the averaged FC coefficients f _t [n], wherein t is the time and n is the filter time lag, can be seen as a time-domain representation of the estimated frequency transfer function F' (j, k) (Fig. 1) . The maximum stable gain the hearing device can achieve for an estimated frequency transfer function F' (j, k) is determined by the conversion unit 18.

In a further embodiment of the present invention, the maximum gains may have to be known in terms of specific frequency bands (e.g. on the Bark scale). The conversion from frequency-domain or time-domain FC coefficients to frequency bands is also done by the conversion unit 18.

A possible processing performed in the conversion unit 18 can be performed using the following formulas:

In time domain:

In frequency domain:

where MSG is an acronym for Maximum Stable Gain and F2B denotes the conversion from frequency bins to Bark bands, wherein a so-called Bark band comprises a collection of adjacent frequency bins. The operation performed on a collection of frequency bins is, for example, an operation to obtain a maximum of the values of specified frequency bins in a Bark band according to the following formula, for example:

F2B{F[k]}=km=ka..x.h _h F[k]

where l _b and h _b are the lower and upper border of Bark band b. Other operations, such as mean or median, are also possible .

Although Fig. 2 shows an embodiment implemented in the frequency domain, the present invention is not restricted to the frequency domain but can readily be implemented in the time domain or together with a time domain feedback canceller, respectively.

The control unit 19 steers the preprocessing of the FC coefficients in the preprocessing unit 17. The means of steering comprises a possible freezing of the running preprocessing, an adjustment of the time constants of the preprocessing as well as a (time-dependent) weighting of the FC coefficients prior to the preprocessing.

The preprocessing may be frozen if the (variable) gain in the gain unit 3 is too low, if the difference to the theoretical feedback threshold is too high or if the hearing device is not in operation, which may be detected by an automatic detection unit (not shown in Fig. 3) .

As has been mentioned above, an averaging of the FC coefficients is performed in the preprocessing unit 17 in one embodiment. In a further embodiment, the preprocessing unit 17 comprises a decision unit that decides, when the preprocessing, for example the averaging of FC coefficients, is to be activated and when frozen, or how the time-constant for the preprocessing is adapted. In another embodiment, a dependency on the prescribed gain is taken into account during the preprocessing step. In still another embodiment, the input level of the acoustic signal is consulted for controlling the preprocessing. In a further embodiment, the variance or the histogram of the FC coefficients are analyzed as well as certain measures derived thereof, as for example percentiles by using dual- slope averaging.

In Fig. 4, a further embodiment of the extraction unit 16 (Fig. 2) is depicted. Instead of one preprocessing unit 17 as shown in Fig. 3, two preprocessing units 17a and 17b and corresponding conversion units 18a and 18b are used to independently determine feedback thresholds. The control unit 19 is connected to both conversion units 18a and 18b in order to control the behavior of the conversion units 18a and 18b. A possible application of such a dual extraction system might be used when one preprocessing unit 17a with the corresponding conversion unit 18a is used for a first hearing program, for example for a general acoustic situation, while the other preprocessing unit 17b with the corresponding conversion unit 18b is used for a second hearing program, for example for a telephone situation. Therefore, the control unit 19 is designed to switch between the two preprocessing units 17a, 17b and the corresponding conversion unit 18a, 18b in order to obtain the correct maximum gain and feedback threshold, respectively.

In a more general embodiment, more than two preprocessing units and corresponding conversion units are used to independently determine frequency feedback thresholds. Again, the control unit may then be connected to some or all conversion units to control its behavior. Such an embodiment can be used to readily adapt to different hearing programs, for example.

In a further embodiment of the present invention, the extraction unit 16 (Fig. 2) is implemented as depicted in

Fig. 5. In contrast to other mentioned embodiments, a histogram of the FC coefficients is computed in the preprocessing unit 17 as well as statistical parameters (e.g. percentiles, variance, number of peaks, etc.) derived from the histogram. Of these parameters, the variance may serve as indicator for the accuracy of the measurement, for example.

It is pointed out that a further embodiment of the present invention is that it can be used during regular operation of the hearing device. This is true for threshold level determination during regular operation as well as during a fitting session, wherein an audiologist adjusts a hearing device for a later regular use by the hearing device user. Also during the fitting session, no probe signals must be presented to the hearing device user. Normal surround signals (noise) are sufficient during a fitting session in order to adjust the hearing device, in particular the maximum gain and the threshold level, respectively. In both applications, the hearing device is continuously adjusted during regular operation using the acoustic signals that are presented to the hearing device user. There is no need to interrupt regular operation, nor is it necessary to present a certain probe signal.

Fig. 6 shows the embodiment of the extraction unit 16 as has been shown in Fig. 3. In addition to the block diagram shown in Fig. 3, a dashed line is inserted in Fig. 6 to indicate the possibility to implement the control unit 19 and the preprocessing unit 17 in the hearing device while

- li ¬

the conversion unit 18 is implemented in an external calculation unit, as for example in a personal computer, which is designed to read out the FC coefficients preprocessed in the preprocessing unit 17 in order to complete the calculation, namely the determination of the maximum gain and threshold level, respectively, in the calculation unit. In practice, the calculation unit will be the personal computer to which the hearing device is hooked up either via a wired or via a wireless connection to the fitter's personal computer.

Fig. 7 shows a simplified block diagram of a further embodiment of the present invention in that a hearing device is depicted comprising a microphone 1, a gain unit 3, an adaptive filter unit 30, a receiver 11 and a extraction unit 16. In contrast to the concept implemented in the embodiment according to Fig. 2, the input signal of the hearing device is not adapted by the adaptive filter unit 30. In fact, the adaptive filter unit 30 is run in the background in the embodiment according to Fig. 7. Although the input signal is not adapted by the adaptive filter unit 30, the adaptive filter unit 30 nevertheless adjusts the FC coefficients such that a feedback threshold can be extracted.

This embodiment can be used, for example, if feedback threshold estimation is desired before the feedback canceller is activated.

Fig. 8 shows an embodiment, in which the gain of the gain unit 3 is adjusted within the limits of maximum gain and threshold level, respectively, determined by the extraction unit 16. Thereto, a gain adjustment unit 31 is provided that is connected in-between the extraction unit 16 and the gain unit 3 in order to automatically adjust the gain such that the loop-gain is, for example, approximately at -5dB, where a desired accuracy of the estimation is achieved. The gain adjustment can thereby be performed stepwise or continuously.

The first use case is as replacement of state-of-the-art feedback threshold estimation methods. The feedback threshold is measured during the fitting session in order to set maximal gains such that the hearing instrument operates in a stable condition. Other optimizations depending on the measured feedback threshold are possible.

Described method can also be used during every-day operation of the hearing instrument. In this use case, the hearing instrument measures the hearing threshold continuously. This continuously measured feedback threshold may be readout by the fitting software in the following fitting session and used as information for the fitter, who can make adjustments based on the continuously measured feedback threshold.

The measured feedback threshold can also be used to adjust parameters of the hearing instrument online and

automatically, e.g. reduce the maximum gain if the feedback threshold has worsened.