EP1052881B1 - Hearing aid with oscillation detector and method for detecting oscillations in a hearing aid - Google Patents

Description

The invention relates to a hearing aid with an oscillation detector and a method for detecting oscillation in a hearing aid.

In principle, methods for the detection of oscillations and thus of acoustic feedback phenomena are known as state of the art. As a rule, these methods are based on relatively complex algorithms or complex frequency analyzes (eg by Fourier transformation). Such a method for oscillation detection, as for example from the EP 0 656 737 A1 is known, circuitry is relatively complex to implement and is therefore only conditionally suitable for use in a hearing aid.

From the WO 96/35314 A1 a feedback suppression is known in a hearing aid, in which the occurrence of feedback is determined. If feedback is detected, the gain is reduced in at least one frequency band and left unchanged in all other frequency bands. Finally, to set the hearing aid, a set of appropriate parameters is generated and stored. A disadvantage of this feedback suppression is that while the occurrence of feedback only during the adaptation of the hearing aid is detected and feedback during normal operation of the hearing aid not recognized and eliminated.

From the DE 37 33 983 A1 A method of attenuating noise against speech in hearing aids is known, which also reduces the tendency for feedback. In this Method, the spectral distribution of detected sound signals by Fourier analysis in a plurality of frequency windows is determined and compared with predetermined limits. The feedback suppression is based on a frequency transposition of the microphone signal so that feedback does not even occur. However, this method leads to a distortion of the input signal and thus affects the sound of the hearing aid.

From the DE 39 27 765 C2 is a hearing aid with a signal processing unit for improved separation of speech signals to noise signals known. This low-frequency noise signals are detected and attenuated. Acoustic feedbacks are not detected and suppressed.

In Maxwell / Zurek "Reducing Acoustic Feedback in Hearing Aids" (IEEE TRANSACTION ON SPEECH AND AUDIO PROCESSING, VOL. 3, NO 4, July 1995 ) uses an adaptive filter whose frequency response always corresponds to the inverted signal frequency response. If the input signal has strong tonal components, a kind of blocking filter is formed whose blocking band lies on the largest frequency component of the microphone signal. When the energy of the microphone signal in the stopband exceeds a threshold, the output is the output of the notch filter. Otherwise, the microphone signal will remain unchanged. There is thus no reliable feedback detection. This signal processing not only reduces any feedback but generally all the dominant spectral components of the input signal. Therefore, voice signals having a pronounced frequency structure are also attacked. As reported in Maxwell / Zurek, this leads to a significant deterioration of the sound quality.

Finally, out of the DE 38 02 903 C2 a hearing aid with a parallel circuit of several Frequenzselektierkanälen known. For suppression of interfering signals in such areas, in which these interfering signals are not obscured by spectral components of the speech, there are threshold switches in each of these channels with a control signal generator. The sum of the output signals of the individual threshold value switches is supplied to the acoustic output transducer. This circuit is not suitable for suppressing feedback.

The invention has for its object to provide a hearing aid with an oscillation detector and a method for detecting oscillations, which can be realized with little circuit complexity in a hearing aid.

The object is achieved for the hearing aid by the features of the characterizing part of claim 1 in conjunction with the features of the preamble. Advantageous embodiments of the hearing aid are described in the further claims 2-13. For the method, the object is solved by the features of the characterizing part of patent claim 14 in conjunction with the features of the preamble. Advantageous process variants can be found in the further claims 15 - 23rd

In the following, a separate hearing aid device, e.g. BTE or ITE device, or an implantable hearing aid understood.

The hearing aid according to the invention is characterized by a particularly simple design of the oscillation detector, which requires only small space and can be realized with a low circuit complexity.

The oscillation detector of the hearing aid according to the invention comprises a period measuring element for determining the respective number of digitized samples of successive periods of an input signal of the microphone of the hearing aid.

In this case, the input signal is digitized by an A / D converter, which can be connected upstream of the oscillation detector or integrated into it.

The output values of the period measuring element undergo downstream averaging elements to determine a long-term average and a short-term average, which are respectively related to a longer and a shorter time period.

If it is determined in a comparison element that the corresponding long-term and short-term average values are substantially identical, the presence of an oscillation and thus of an acoustic feedback is detected.

The hearing aid according to the invention thus takes into account that oscillations to be detected usually occur in the form of sinusoidal input signals and that in such sinusoidal input signals of the long-term and short-term average of the number of digitized samples in successive signal periods is substantially identical. As a result, sinusoidal signals can be distinguished from non-sinusoidal input signals.

Specifically, the averaging elements of the hearing aid device according to the invention can be designed as low-pass filters. In order to reduce the circuit complexity, first-order low-pass filters are used, a low-pass filter having a longer time constant for determining the long-term average value and a further low-pass filter having a shorter time constant for determining the short-time average value.

The oscillation detector has an amount element to cause a sign correction of the difference between the long-term and the short-term average, so that all the evaluation values with positive signs are present. This allows a comparison with a likewise positive threshold, without having to monitor a threshold range with both negative and positive values.

Finally, the oscillation detector has another averaging element, e.g. another low-pass filter to even out and smooth out the difference between the long-term and the short-term mean value as determined in the predicate.

When comparing the long-term mean value with the short-time mean value, the comparison element of the oscillation detector used for this purpose advantageously has an adjustable threshold value in order to be able to regulate the sensitivity of the oscillation detector. The threshold value can either be set manually or adjusted automatically depending on detected ambient or noise situations.

An object of the method according to the invention for detecting oscillation is to detect substantially sinusoidal input signals of the microphone, since in the presence of such signals usually oscillation and thus present feedback is assumed.

First, the number of digitized samples in successive periods of the input signal of the microphone is determined. Now it is a question of determining whether the number of sampled values obtained in successive periods changes or is substantially identical. For this purpose, a long-term average value N _L and a short-term average value N _{K of} the number of sampled values determined are formed.

In the following, a difference of N _L and N _{K is} formed, the sign of the difference by magnitude formation optionally corrected and finally the difference value of N _L and N _K still smoothed.

In a subsequent comparison of the long-term mean value with the short-term mean value by subtraction, it can be determined whether the two mean values are substantially identical. If this is the case, it can be assumed that the periods of successive signal periods of the input signal are substantially identical and thus a substantially sinusoidal signal is present and thus the presence of oscillation is to be determined.

Overall, the method according to the invention makes it possible to detect the presence of substantially sinusoidal signals and thus of oscillation with a small number of process steps which are simple to implement in terms of circuitry.

When oscillation is detected, a filter element is set to the appropriate notch frequency and activated to reduce frequency-specific (narrow-band) gain and to suppress the feedback effect. This ensures that a gain reduction is set only when in fact a feedback effect to be suppressed occurs. Furthermore, the interference suppression by the activated filter element concentrates only on that frequency or that frequency range in which the feedback actually occurs. The remaining frequency ranges of the signal to be processed are not reduced in their gain.

By the oscillation detector feedback effects are detected, which are characterized by sinusoidal signals.

With the oscillation detector, it is also possible for the hearing aid device according to the invention to detect fluctuating feedback effects over time and to determine the respective frequency suitable for suppression and via activation of the filter element. This also allows changing feedback situations to be detected and remedied.

In an advantageous embodiment, the hearing aid device can have a plurality of filter elements (notch filters, for example notch filters). If it is determined upon activation of a first filter element that the feedback effects persist, thus further filter elements can be activated in order to achieve a gain reduction at the respective further frequencies.

A general adjustability of the filter characteristic of the filter elements (eg position and width of the rejection frequency range, degree of gain reduction) makes it possible, after a determination of the respective feedback effect by the oscillation detector not only the frequency range of the gain reduction, but also other suitable countermeasures (eg "width" and " Depth "of the notch effect of the notch filter). With a particularly detailed typification of the feedback effects exactly matched countermeasures can be taken by appropriate generation of the filter elements, whereby a particularly targeted and adapted feedback suppression is achieved without this being perceived by the user of the hearing aid, since only individual frequency ranges of the microphone signal to be processed are affected ,

The reliability of the oscillation detector is increased when the level of the output signal is detected by a level measuring unit and feedback is detected only when an adjustable level threshold is exceeded.

Via a control unit, the oscillation detector, the filter elements and / or level measuring unit connected, can be controlled and activated. Thus, it can be achieved by the control unit that upon detection of feedback conditions activation of the filter element takes place by the oscillation detector only when a signal level detected by the level measuring unit and above an adjustable threshold value is present.

The control unit may also be programmable and take on further functions of the signal analysis of the signal to be processed, so that e.g. due to the signal analysis, a suitable threshold value is set in the level measuring unit and this is changed if necessary.

In the case of a detected feedback and a subsequent appropriate activation of the filter elements, the desired frequency-specific gain reduction is realized in the signal processing path. Then, when the goal of feedback suppression is achieved, no feedback effect can be detected.

It can therefore no longer be determined when the e.g. due to certain environmental conditions only temporarily occurred feedback effect is no longer present. Advantageously, therefore, the filter elements are activated only for an adjustable period of time and then turned off again. Thereupon it can be determined by the oscillation detector whether the initially determined feedback is still present and, if necessary, a new filter activation takes place. By a time-limited filter activation is avoided that the filters remain turned on, although due to changing environmental conditions, the causes of the feedback effect are eliminated.

A time-controlled activation of the filter elements can take place via a time switching element, which can also be integrated in the control.

In a method according to the invention for operating a hearing aid, it is first determined for feedback detection, whether an oscillating microphone signal is present and then - if this is the case - set a corresponding frequency-specific gain reduction of the processed microphone signal. Occurring feedback effects are suppressed frequency-specific and narrow-band, without other frequency ranges of the microphone signal to be processed are touched. The method according to the invention enables a targeted feedback suppression, which is particularly convenient for the user, since the usual amplification potential is retained in all frequency ranges which are not affected.

According to an advantageous variant of the method, it is continuously determined whether a feedback occurs and then respectively suitable measures for frequency-specific gain reduction are taken, which adapt continuously in time to changes in the determined feedback effects. By constantly monitoring the sound situation, changes in feedback effects are also detected, and it is possible to react accordingly by means of suitable dynamic gain reduction measures, so that precisely the respective detected feedback is always suppressed and changes and an omission of the feedback effects are taken into account.

In a further method step, it can be determined whether the output signal exceeds a certain level threshold value. This takes into account the large increase in signal levels occurring in a feedback situation.

Advantageously, in a variant of the method it is additionally queried whether the signal level of the sinusoidal signal is above a certain threshold value. Only when an oscillating signal and an increased level are cumulative, a feedback is actually detected.

If the respective level threshold is not exceeded, a sinusoidal largely monofrequent input signal can also be present without a feedback situation.

By taking the level criterion into consideration, it is thus possible to reliably prevent misdetections.

In the frequency-specific gain reduction, the respective measure for the suppression of the feedback can be adapted in an optimized manner and, by activating a suitable filter characteristic, an adaptation of e.g. the position and the size of the notch frequency range and the degree of gain reduction, a further improved feedback suppression can be achieved.

According to a further advantageous method step, the gain reduction takes place only for a certain period of time and then is canceled on its own, in order to take back and switch off the originally set gain reductions with changes in the sound situation and an omission of the originally detected feedback effect. Upon renewed detection of feedback effects, a renewed activation of filter measures for gain reduction can again be set.

FIG 1

eine Ausführungsform des Oszillationsdetektors sowie

FIG 2

eine schematische Darstellung des Hörhilfsgeräts mit dem Oszillationsdetektor.

Further details and variants of the method according to the invention are explained in connection with the description of the embodiment shown in the drawing figures. Show it:

FIG. 1

an embodiment of the oscillation detector as well

FIG. 2

a schematic representation of the hearing aid with the Oszillationsdetektor.

The operation of the oscillation detector 4 follows from the block diagram FIG. 1 out. An input signal 11 from the microphone 1 (see. FIG. 2 ) Is first digitized in the A / D converter 10 at a sampling rate f _S. The A / D converter 10 may be integrated in or connected to the oscillation detector 4. The digitized samples of the A / D converter 10 are supplied to the period measuring element 5 to determine there the number of samples in the respective successive periods of the input signal 11. With an assumed sampling rate f _S = 20 kHz of the A / D converter 10, it could thus be ascertained by the period measuring element 5 that the successive periods contain, for example, 4 to 6 samples.

Such a period measurement in the period measuring element 5 may e.g. by detecting and analyzing the zero crossings of the digitized samples. Thus, a change of sign and also the direction of the sign change (from + to - or vice versa) can be detected. Overall, therefore, the beginning and the end of a period of the input signal 11 (period) can be determined and the number of digital samples between the beginning and the end of the period can be determined.

Averaging elements downstream of the period measuring element 5, namely the low-pass filters 6, 7, are used to determine the long-term average value N _L and the short-term average value N _K of the number of sampling values determined by the period measuring element in the respective signal periods of the input signal 11.

To reduce the circuit complexity, the two low-pass filters 6, 7 are constructed as recursive systems of the first order, which determine the respective averaging values N _L and N _K according to the following processing instruction: $${\mathrm{y}}_{\mathrm{t}}\mathrm{=}{\mathrm{cy}}_{\mathrm{t}\mathrm{-}\mathrm{1}}\mathrm{+}\left(\mathrm{1}\mathrm{-}\mathrm{c}\right){\mathrm{x}}_{\mathrm{t}}$$

The recursion constant c is in the range of 0 to 1 and determines the time constant of the low passes 6, 7. If the long-term average value N _{L is} determined in the low-pass filter 6, a relatively high recursion constant (for example c = 0.99) enters its processing specification. When determining the short-term average value N _K in the low-pass filter 7, its recursion constant is relatively small (eg c = 0.5). x represents the input value (ie the number of sampled values determined per period) and y the respective output value (ie N _L or N _K ) of the processing instruction, in each case provided with the indices t for the respective sampling time interval.

In difference element 12, a difference between N _L and N _{K is} formed, which is advantageously sign-corrected in magnitude element 13, so that only positive values occur. Finally, another smoothing by another averaging element, here also for example a first order low-pass filter 14, also takes place. Thereafter, in the comparison element 8, the magnitude of the difference between N _L and N _{K is} judged. If it is below a certain adjustable threshold value, it is assumed that a substantially sinusoidal input signal 11 is present and thus the presence of oscillation can be assumed.

In the case of oscillation detected in this way, the oscillation frequency 16 (f _OSZ ) is now determined in the frequency determiner 15, specifically as the product of the sampling rate f _S with the reciprocal of the long-term mean value N _L.

If oscillation has been detected and a sampling rate f _S = 20 kHz is present and the long-term mean value N _L = 5 has been determined, this means that an oscillation frequency f _OSZ = 4 kHz is present. This oscillation frequency f _OSZ can be passed from the oscillation detector 4 to a control element 20, so that suitable filter elements 17, 18, 19 can be activated in order to suppress the detected oscillation, for example by filters with notch effect (cf. FIG. 2 ).

From the schematic diagram according to FIG. 2 is a hearing aid with at least one microphone 1, a signal processing device 2 with means for signal processing and a handset 3, wherein in the signal processing path between the microphone 1 and the handset 3 filter elements 17, 18, 19 are provided for frequency-specific gain reduction.

The filter elements 17, 18, 19 mentioned are set to a determined blocking frequency and activated if oscillating signals are detected by the oscillation detector 4 and thus a feedback is analyzed as present.

If this is the case, this is reported by the oscillation detector 4 to the control element 20 and there is the adjustment (setting the blocking frequency) of at least one of the filter elements 17, 18 or 19. If necessary, a plurality of filter elements 17, 18, 19 are activated. In principle, any number of filter elements 17, 18, 19 can be provided in order to achieve a required extensive feedback suppression.

About the Pegelmeßeinheit 21, the control element 20 is informed of the level of the output signal. Feedback is detected by the control element 20 only if the presence of an oscillating signal is registered by the oscillation detector 4 and at the same time a level of the output signal, ie the processed microphone signal, exceeding the determined threshold value of the level measuring element 21 is detected by the level measuring element 21.

In this way, the increased level often occurring in the case of feedback effects can also be taken into account and a detection of strong monofrequency input signals in feedback-free situations is avoided.

Via the control element 20, the filter elements 17, 18 and 19 are not only activated, but also adjusted according to their filter characteristics (e.g., position and magnitude of the notch frequency range, degree of gain reduction). For this, the control may be programmable and e.g. initially set the usual basic parameters of filter characteristics, then - if the feedback is not sufficiently suppressed - make a corresponding adjustment of the filter characteristics after a specified or self-learning program flow.

An analysis or typing of the feedback detected by the oscillation detector 4 can also take place via the control element 20 and can be used to determine or determine the filter characteristic of the filter elements 17, 18 and 19.

By means of a time switching element 22 integrated in the control element 20, it is achieved that, given an originally detected feedback which led to an activation of the filter elements 17, 18, 19, this activation lasts only for a limited period of time. Thereafter, the activation of the filter elements 17, 18, 19 is automatically withdrawn and it is again determined via the oscillation detector 4, whether the originally occurred feedback is still present or whether a changed feedback (eg frequency shift) has been found, the one but again with Activation of the filter elements 17, 18, 19 provided with a different type of filter characteristic may appear expedient.

Claims (23)

1. Hearing aid with a microphone, a signal-processing arrangement and a receiver,
   characterized in that provision is made for an oscillation detector (4) for detecting acoustic feedback and the oscillation detector (4) has

   - a period measurement element (5) for determining the respective number of digitized sampling values in successive periods in an input signal (11) of the microphone (1),

   - averaging elements (6, 7) for determining a long-term average N_L and a short-term average N_K of the number of sampling values determined by the period measurement element (5),

   - a difference element (12) for forming the difference between N_L and N_K,

   - an absolute-value element (13) for correcting the sign of the difference between N_L and N_K,

   - a further averaging element (14) for smoothing the difference value between N_L and N_K, and

   - a comparison element (8) for comparing N_L and N_K.
2. Hearing aid according to Claim 1,
   characterized in that the oscillation detector (4) has an A/D convertor (10) for determining digitized sampling values from the analogue input signal (11) from the microphone (1).
3. Hearing aid according to Claim 1 or 2,
   characterized in that the averaging elements are designed as low-pass filters (6, 7, 14).
4. Hearing aid according to Claim 3,
   characterized in that first-order low-pass filters are provided.
5. Hearing aid according to one of Claims 1-4,
   characterized in that the comparison element (8) has an adjustable threshold for comparison with the difference between N_L and N_K.
6. Hearing aid according to one of Claims 1-5,
   characterized in that provision is made for a filter element (17) with correspondingly differently adjustable blocking frequencies for suppressing the detected feedback.
7. Hearing aid according to Claim 6,
   characterized in that provision is made for a plurality of filter elements (17, 18, 19) with different blocking-frequency ranges and/or filter characteristics.
8. Hearing aid according to either of Claims 6 and 7,
   characterized in that different filter characteristics (e.g. position and width of the blocking frequency range, degree of the amplification reduction etc.) can be adjusted in the filter elements (17, 18, 19).
9. Hearing aid according to one of Claims 6-8,
   characterized in that provision is made for a control element (20) that is connected to the oscillation detector (4) and the filter elements (17, 18, 19).
10. Hearing aid according to one of Claims 6-9,
    characterized in that provision is made for a timer element (22) for setting the duration of the activation of the filter elements (17, 18, 19).
11. Hearing aid according to Claim 10,
    characterized in that the timer element (22) is integrated into the control element (20).
12. Hearing aid according to one of Claims 9-11,
    characterized in that provision is made for a level measurement unit (21) for detecting the level of the output signal.
13. Hearing aid according to Claim 12,
    characterized in that the level measurement unit (21) has an adjustable threshold for activating the control element (20).
14. Method for identifying oscillations in a hearing aid with a microphone, a signal-processing arrangement and a receiver, more particularly a hearing aid according to one of Claims 1-13,
    characterized by the following method steps:

    a) Determining the respective number of digitized sampling values in successive periods in an input signal from the microphone,

    b) Determining a long-term average N_L and a short-term average N_K of the number of sampling values determined according to a),

    c) Forming the difference between N_L and N_K,

    d) Forming the absolute value of the difference between N_L and N_K,

    e) Smoothing the difference value between N_L and N_K,

    f) Comparing N_L and N_K, and

    g) Detecting an oscillation if N_L and N_K are basically identical.
15. Method according to Claim 14,
    characterized in that the sign of the digitized sampling values is detected and analysed in the method step a) according to Claim 14.
16. Method according to Claim 14 or 15,
    characterized in that an adjustable threshold is provided in the method step g) according to Claim 14, an oscillation being detected if said threshold is exceeded.
17. Method according to one of Claims 14-16,
    characterized in that when oscillation according to method step g) according to Claim 14 is detected, the oscillation frequency f_osc is determined as the product of the sampling rate f_S, at which the input signal from the microphone was digitized, and the reciprocal of N_L.
18. Method according to one of Claims 14-17,
    characterized in that when oscillation is detected, there is a frequency-specific amplification reduction of the microphone signal to be processed in order to suppress feedback.
19. Method according to Claim 18,
    characterized in that there is continuous detection of whether feedback is present and a correspondingly changed frequency-specific amplification reduction is carried out.
20. Method according to one of Claims 18-19,
    characterized in that filter elements with correspondingly set blocking frequencies are activated for frequency-specific amplification reduction.
21. Method according to Claim 20,
    characterized in that the filter characteristics (e.g. position and width of the blocking frequency range, degree of the amplification reduction etc.) of the filter elements are changed for optimized feedback suppression.
22. Method according to one of Claims 18-21,
    characterized in that the amplification reduction is set for an adjustable duration and then is lifted again.
23. Method according to one of Claims 14-22,
    characterized in that feedback is detected if there is an oscillating microphone signal and the output signal exceeds an adjustable minimum level.

Images