EP1453355B1 - Signal processing in a hearing aid - Google Patents

Description

The invention relates to a device and a method for signal processing in a hearing aid according to the preambles of the independent claims. The invention is particularly suitable for improving speech intelligibility by suppressing noise in hearing aids or hearing aids.

STATE OF THE ART

A generic method is for example from the EP 1 067 821 A1 known. Therein a hearing aid is described in which there is a suppression of noise in an input signal in a main signal path, which has neither a transformation in the frequency domain nor a division into subband signals, but only a suppression filter. A transfer function of the suppression filter is periodically re-determined due to attenuation factors which are determined in a signal analysis path parallel to the main signal path. The attenuation factors are used to attenuate signal components in frequency bands with a significant amount of noise. The suppression filter is implemented as a transversal filter, whose impulse response is periodically expressed as the weighted sum of the impulse responses of transversal filters Bandpass filters is recalculated. In this way, a processing with a low signal delay is possible.

PRESENTATION OF THE INVENTION

It is an object of the invention to provide a device and a method for signal processing in a hearing aid of the type mentioned, which realize a higher quality and intelligibility of the processed signal.

This object is achieved by a device and a method for signal processing in a hearing device having the features of patent claims 1 and 10 and a hearing device having the features of patent claim 20.

• werden Koeffizienten einer Kompressionsverstärkung, welche eine frequenzabhängige Anpassung des Eingangssignals nach Massgabe von frequenzabhängigen Signalpegeln des Eingangssignals beschreiben, bestimmt,
• werden Koeffizienten einer Geräuschunterdrückung, welche eine frequenzabhängige Anpassung des Eingangssignals nach Massgabe von im Eingangssignal detektierten Störgeräuschen beschreiben, bestimmt, und
• werden Koeffizienten eines Filters zur Filterung des Eingangssignals aus den Koeffizienten der Kompressionsverstärkung und den Koeffizienten der Geräuschunterdrückung berechnet..

In the inventive method for signal processing in a hearing aid

• determine coefficients of a compression gain which describe a frequency-dependent adaptation of the input signal in accordance with frequency-dependent signal levels of the input signal,
• are coefficients of noise suppression, which describe a frequency-dependent adjustment of the input signal in accordance with detected in the input signal noise detected, determined, and
• For example, coefficients of a filter for filtering the input signal are calculated from the coefficients of the compression gain and the coefficients of the noise suppression.

In summary, the term "adaptation of a signal" means both amplification and attenuation.

The invention makes it possible, the amplitude response of the filter to changing voice and interference signals and to the needs of a to match a person with the same name, whereby a delay time for the filtering of the input signal is kept small.

Another advantage is that the compression gain allows different gain values for different frequency ranges of the input signal.

Another advantage is that only a single controllable filter is used both for compression enhancement and for noise suppression.

In a preferred embodiment of the invention, the determination of the coefficients of the compression gain occurs in a first set of frequency ranges F _n with n = 1..N of the input signal on the basis of signal levels or amplitude components. A signal level is determined from a sub-signal of the input signal, which is formed by filtering the input signal and dividing into sub-signals with signal components in only one frequency range at a time. The signal levels are determined iteratively as instantaneous rms values of a signal power in the respective frequency ranges of the input signal. This makes it possible to track the compression gain with a temporal resolution corresponding to a sampling rate of the input signal.

In a preferred embodiment of the invention, the determination of the coefficients a _{m of} the noise suppression in a second set of frequency ranges Φ _m with m = 1..M of the input signal by determining modulation depths d _m and by determining the coefficients a _m for each of the frequency ranges Φ _m according to the corresponding modulation depth d _m . In this case, the modulation depths d _{m are determined} from a temporal sequence of maximum and minimum values of a signal level p _m in the respective frequency range Φ _m . This makes it possible to selectively filter out weakly modulated, ie monotone, noise. Time constants for the adaptation of the Noise suppression is preferably in the range of around 50 milliseconds or less.

In a preferred embodiment of the invention, the frequency ranges Φ _m for the noise suppression are small in comparison with the frequency ranges F _n for the compression gain. Thus, at least one frequency range F _n comprises two or more frequency ranges Φ _m . Accordingly, filters for determining proportions of the input signal in the frequency ranges Φ _{m have} a greater signal propagation time or delay than filters for the frequency ranges F _n . This allows a sharp division of the frequency range to suppress interference and at the same time a rapid adjustment of the compression gain to a changing voice signal. A maximum tolerable delay for the adaptation of coefficients of the compression gain is 5 milliseconds, values below 2.5 milliseconds are preferred. Values below one millisecond can be achieved according to the invention.

In a further preferred embodiment of the invention, the filter is not tracked exactly to the newly calculated coefficients in each sampling interval. Instead, it is tracked only according to one or more changed coefficients. This allows adaptation with low computational effort and correspondingly low energy consumption. The adaptation preferably takes place only for the one or more coefficients whose change exceeds a predetermined threshold or which is comparatively large or largest. Also possible is a periodic change of one or a few coefficients or a pseudorandom traversal and adaptation of all coefficients.

In a further preferred embodiment of the invention, an influence of the noise suppression is taken into account in the determination of the coefficients for the compression gain. To this end, a means to Determining Noise Suppression Coefficients A compression gain coefficient determining means Corresponding to a signal attenuation caused by the noise suppression.

The inventive device has the features of claim 10. A hearing aid according to the invention has means for carrying out the method according to the invention.

Further preferred embodiments emerge from the dependent claims. Characteristics of the method claims are analogously combined with the device claims and vice versa.

BRIEF DESCRIPTION OF THE DRAWINGS

In the following, the subject invention will be explained in more detail with reference to preferred embodiments, which are illustrated in the accompanying drawings.

Figur 1

schematisch eine Struktur der Signalverarbeitung;

Figur 2

ein Blockdiagramm einer Berechnung von Verstärkungswerten; und

Figur 3

ein Blockdiagramm einer Berechnung von Abschwächungswerten und Korrekturgrössen gemäss der Erfindung.

Show it:

FIG. 1

schematically a structure of the signal processing;

FIG. 2

a block diagram of a calculation of gain values; and

FIG. 3

a block diagram of a calculation of attenuation values and correction quantities according to the invention.

The reference numerals used in the drawings and their meaning are listed in the list of reference numerals. Basically, the same parts are provided with the same reference numerals in the figures.

WAYS FOR CARRYING OUT THE INVENTION

The FIG. 1 schematically shows a structure of the signal processing in a hearing aid according to the invention. An input signal X is fed to a controllable filter 6, to a means for determining a compression gain 7 and to a means for determining a noise suppression 8. The controllable filter 6 is designed to form an output signal Y in accordance with filter coefficients c ₁ ... C _M.

On the average for determining the compression gain 7, the input signal X is fed to a first filter unit 1. The first filter unit 1 is designed to determine signal components x ₁ .x _{N of} the input signal X in a first set of frequency ranges F _n with n = 1..N. In a signal processing for the compression gain 3, the signal components x ₁ .x _{N are used to} calculate parameters or coefficients or adaptation values of the compression gain g ₁ ..g _M. These coefficients are also referred to as gain values with regard to the amplification function of the hearing aid. However, other coefficients are also referred to as gain values.

On the average for determining the noise suppression 8, the input signal X is fed to a second filter unit 2. The second filter unit 2 is designed to determine signal components y ₁ .y _{M of} the input signal X in a second set of frequency ranges Φ _m with m = 1..M. In a signal processing for noise suppression 4, the signal components y ₁ .y _{M are used to} calculate parameters or coefficients or adaptation values of the noise suppression a ₁ .a _M. These coefficients are also referred to as attenuation values in view of the noise suppression achieved thereby.

The combining unit 5 combines the coefficients of the compression gain g ₁ ..g _M with the coefficients of noise suppression a ₁ ..a _M and calculates therefrom combined logarithmic amplification values c ₁ ... c _M as filter coefficients of the controllable filter 6. Preferably, the mentioned coefficients g _i , a _i and c _{i are} scaled logarithmically and in the combination unit 5 is essentially a subtraction c _m = g _m -a _{m performed} with m = 1..M.

In a preferred embodiment of the invention, the noise suppression signal processing 4 transmits correction values r ₁ ..r _N corresponding to a respective signal attenuation caused by the noise suppression in the frequency ranges F ₁ ..F _n .

In a further preferred embodiment of the invention, the first filter unit 1 and the second filter unit 2 are not implemented as separate units, but as a combined filter unit. For example, filtering with broad frequency bands is carried out sequentially to determine the signal components x ₁ ..x _N , and these filtered signals are further filtered to determine the signal components y ₁ ..y _M.

The invention in the embodiment shown works in summary as follows: The input signal is split into three signal paths, a main signal path with a controllable filter, a first parallel signal analysis path for the compression gain, and a second parallel signal analysis path for noise suppression.

FIG. 2 FIG. 12 shows a block diagram of a calculation of gain values in compression amplification signal processing 3. For the compression gain, signal levels in N are computed in relatively few frequency ranges. FIG. 2 shows the calculation for one of these N frequency ranges, for the remaining frequency ranges the same structure is used. By a signal component x _n in this frequency range, a signal power is formed in a block 21, for example, as a running sum of squared signal values. In a block 22, a signal level p _{n is} formed by logarithmization. The term signal level here refers to the effective value of the instantaneous signal power in the frequency range F _n expressed in a logarithmic number range, eg in dB. From the signal level p _n , a modified signal level p _n 'is calculated by subtracting 23 a correction value r _n . The determination of correction values r _n will be discussed separately below. Each frequency range F _{n of} the compression gain is assigned at least one frequency range Φ _{m of} the noise suppression. For each of these associated frequency ranges Φ _m (in the FIG. 2 if these are three, corresponding to blocks 24, 24 ', 24 "), a separate function f _{m is} predefined, which calculates a gain value g _m from the modified signal level p _n ', ie $${\mathrm{G}}_{\mathrm{m}}\mathrm{=}{\mathrm{f}}_{\mathrm{m}}\left({\mathrm{p}}_{\mathrm{n}}\mathrm{\text{'}}\right)$$

These functions f _m into account an individual hearing loss and audiological experience. Parameters, amplification values or hearing correction values contained in the functions f _m are preferably user-specific and stored, for example, in an EPROM of the hearing device. The total number of these functions f _m and the gain values g _m , ie over all N frequency ranges F _{n of} the compression gain, is equal to the number M of the frequency ranges Φ _{m of} the noise suppression.

If one aims to amplify quiet phonemes, ie consonants, in a speech signal more than loud phonemes, ie vowels, so that as far as possible all phonemes in continuously spoken speech become audible for a hearing impaired person, then the signal levels p _{n must} be determined such that Differences between quiet and loud consecutive phonemes are well captured. In addition, the continuously determined amplification values g _m must be applied in a timely manner to those signal sections in which the associated phonemes are located, ie the amplification values must act synchronously on the audio signal X. So fast, in the rhythm of successive phonemes acting, synchronous compression gain gives only good results when the number of separate frequency ranges is chosen small, for example, N ≤ 5, preferably N ≤ 3. Otherwise, for the different phonemes characteristic spectral differences between the frequency ranges are reduced too much and thus impaired speech intelligibility. The compression gain with a few, relatively wide frequency ranges is possible with a low processing delay of the order of 1 millisecond, which comes close to the desire for an ideally lag-free signal processing. In a preferred embodiment of the invention, the compression gain is performed jointly for only a single frequency band, ie for the entire frequency range of the audio signal. In another embodiment of the invention, two frequency bands are used for this, ie N = 2.

berechnet, wobei 0 < ε << 1 gewählt wird.The signal analysis for determining signal levels in frequency ranges f _n for the compression gain is preferably carried out iteratively, wherein current signal levels are determined for each new value of the input signal. For this purpose, recursive signal analysis methods are preferably used. For example, iteratively, the root-mean-squared value of the signal x [k] at the k-th sampling time is called $$\mathrm{s}\left[\mathrm{k}\right]\mathrm{=}\mathrm{s}\left[\mathrm{k}\mathrm{-}\mathrm{1}\right]\mathrm{+}\mathrm{\epsilon }\mathrm{\cdot }\left({\mathrm{x}}^{\mathrm{2}}\left[\mathrm{k}\right]\mathrm{-}\mathrm{s}\left[\mathrm{k}\mathrm{-}\mathrm{1}\right]\right)\mathrm{.}$$

calculated, where 0 <ε << 1 is selected.

A corresponding signal level value, eg in dB, then results in $$\mathrm{p}\left[\mathrm{k}\right]\mathrm{=}\mathrm{10}\mathrm{*}\log \mathrm{10}\left(\mathrm{s}\left[\mathrm{k}\right]\right)$$

The noise suppression is about attenuating sub-signals in frequency ranges of the audio signal, which are mainly only monotone noise. For this purpose, first differences in M between separate frequency ranges Φ _m between temporally successive maximum and minimum values of the signal levels p _m , so-called modulation depths d _m , are determined, where m = 1,..., M.

For noise suppression, an iterative determination of the signal levels in the bar of the sampling rate of the input signal is not necessary. To save on arithmetic operations, it is therefore preferable to work with reduced sampling rates. In this case, the signal level p _{m is formed in} segments for segments of a length of about 20-30 ms as the instantaneous effective value of the signal power in the corresponding frequency range Φ _m . Thus, the noise suppression can be tracked with a temporal resolution p _m, for example under 50 ms.

Separate estimated value functions are tracked to determine maximum and minimum values. For this purpose, a stored maximum value is either reduced by a small increment linearly or according to an exponential function in each sampling interval, or the current level value is accepted if it exceeds this reduced maximum value. Analogously, the minimum value is increased by a small increment in each sampling interval, or the current level value is adopted if it falls below the raised minimum value. The modulation depth is the difference between these two estimates. A small modulation depth thus arises at constant signal energy. In order to avoid sudden changes in the modulation depth, the difference values thus determined are preferably subjected to a smoothing process. By appropriate selection of the mentioned increments, the extremes with time constants in the range of a few seconds sound.

For speech in a quiet acoustic environment, the modulation depth assumes values of 30 dB and more. In traffic noise, the low frequency range up to about 500 Hz is often dominated by monotone noise, so that even in the presence of speech signals, the modulation depth in this frequency range drops to near 0 dB. Other noise in turn covers the speech signal in higher frequency ranges. Preferably, partial signals in frequency ranges Φ _{m are} attenuated, in which the modulation depth d _m below a critical value of eg 15 dB precipitated, the amount of attenuation a monotonic and _m, for example, increases linearly with decreasing modulation depth.

For the most accurate detection and separation of frequency ranges with different modulation depth, a large number of separate frequency ranges are advantageous, e.g. M = 20. For signal processing in so many narrow frequency ranges, there is inevitably a long time delay of the order of 10 ms, which, however, is well tolerated with a gradual weakening and occasional raising of the sub-signals in these frequency ranges.

The gain values g _{m of} the compression gain 3 and the attenuation values a _{m of} the noise suppression 4 are combined per frequency range and fed as control quantities c _{m to} the controllable filter 6 in the main signal path. If required, the transfer function of the controllable filter is tracked in each sampling interval of the input signal in a frequency-specific manner in one or a few frequency ranges and left unchanged in all other frequency ranges.

For the combined use of compression enhancement and noise suppression, it is possible to perform signal analysis in a relatively large number of frequency ranges Φ _m , as appropriate for noise suppression, and then to appropriately summarize the results with respect to the few frequency ranges F _n relevant to the compression gain. The disadvantage of such a sequential procedure is that a long signal delay of the order of 10 ms results for the entire signal processing. From a mathematical point of view, the fast Fourier transformation and the inverse Fast Fourier transformation in particular seem attractive for such an implementation. This will be the audio signal successively in individual segments with a duration of about 10 ms in the frequency domain transformed, analyzed and modified, and then transformed back into the time domain. The use of the segment-wise signal processing, however, results in the following further disadvantages: The signal levels p _n are calculated as mean values in a segment, whereby a pronounced signal rise at a specific time is detected only with the temporal resolution of a processing segment. Also, the determination of the individual gain values and thus the entire transfer function is carried out only in time with the successive segments.

Preferably, therefore, the filtering of the input signal X is performed on the basis of a separate and parallel signal analysis for the noise suppression as well as for the compression amplification. In this case, the coefficients a _m for noise suppression which have been delayed obtained are combined with faster obtained coefficients g _m for compression amplification, and several of the coefficients g _m with different functions f _{m are} calculated from the same, optionally modified, signal level p _n '= p _n -r _n Frequency range F _{n determined} for compression gain.

The combined and parallel processing takes place in detail as follows: In the lowest signal path, the audio signal passes through a controllable filter 6, which performs the required frequency-dependent signal modifications. The two upper signal paths each contain a filter unit which divides the audio signal into sub-signals of separate frequency ranges. The first filter unit 1 causes a signal division into only a few, N wide frequency ranges F _n , which can be carried out with little signal delay. The second filter unit 2 causes a signal division into many, M narrow frequency ranges Φ _m , which entails a long delay time. In this case, the frequency ranges are preferably selected such that each frequency range Φ _{m is a} subrange of a frequency range F _n . The frequency ranges to compression gain F _n coincide preferably the same frequency range as the frequency ranges for noise suppression Φ _m . A frequency range for compression amplification covers several frequency ranges for noise suppression. Ratios between the widths of frequency ranges and between the distribution of frequency ranges are preferably at least approximately logarithmic.

A typical frequency range for the input signal is: 0 to 10 kHz. For example, this is divided into the following frequency ranges for compression gain and noise cancellation: Compression gain (Hz) Noise suppression (Hz) 0 to 1250 0 to 312.5 312.5 to 625 625 to 937.5 937.5 to 1250 1250 to 2500 1250 to 1562.5 1562.5 to 1875 1875 to 2187.5 2187.5 to 2500 2500 to 10,000 2500 to 3125 3125 to 3750 3750 to 4375 4375 to 5000 50000 to 6250 6250 to 7500 7500 to 10000

The sampling rate is for example 20 kHz and accordingly the useful bandwidth is half, that is 10 kHz. In another embodiment of the invention, these values are 16 kHz and 8 kHz, respectively.

In the signal analysis for noise suppression for each of the M frequency ranges Φ _{m is} a determination of the associated signal level p _m , the modulation depth d _m and the attenuation value a _m , the latter being advantageously expressed in a logarithmic number range. The determination of the modulation depth d _m is carried out as described above in accordance with ie as a function of the time profile of the corresponding signal level p _m , and the determination of the coefficients a _{m in} accordance with the corresponding modulation depths d _m . The second filter unit 2 and a part of the noise suppression signal processing 4 thus form a means for determining these quantities p _m , d _m and a _m in a second set of frequency ranges of the input signal X.

In compression amplification signal analysis, in each of the N frequency ranges F _n, the signal level p _{n is} determined such that each signal value of the sub-signal x _{n [} k] contributes to an update of the signal level, resulting in a higher temporal resolution than the mere Determination of a segment-wise mean value.

bestimmt, wobei jeder modifizierte Signalpegel p_n', also die um die Korrekturwerte r₁..r_N verringerten Pegel, zur Bestimmung der Verstärkungswerte in all jenen Frequenzbereichen Φ_m benützt wird, die zusammengefügt den Frequenzbereich F_n ergeben. Die Korrekturwerte r_n berücksichtigen eine etwaige Abschwächung der Signalleistungen infolge der Geräuschunterdrückung.The first filter unit 1 and part of the signal compression for compression gain 3 thus form a means for determining signal levels in a first set of frequency ranges of the input signal X. Thereafter, for all M frequency ranges Φ _m gain values $${\mathrm{G}}_{\mathrm{m}}\mathrm{=}{\mathrm{f}}_{\mathrm{m}}\left({\mathrm{p}}_{\mathrm{n}}\mathrm{\text{'}}\right)$$

where each modified signal level p _n ', ie the level reduced by the correction values r ₁ ..r _N , is used to determine the gain values in all those frequency ranges Φ _m which, combined, give the frequency range F _n . The correction values r _n take into account a possible attenuation of the signal power due to the noise suppression.

Each of the gain values g _m with m = 1..M is thus assigned to a frequency range Φ _m . With the determination of M different gain values for the narrow frequency ranges Φ _m , the compression gain in the signal processing combined according to the invention is also realizable with a much more flexible transfer function, ie with M instead of only N functions f _m , than if only one gain value for each broad frequency range F _n would be set. The gain values g _m are again preferably expressed in a logarithmic number range. The functions f _m specify frequency-specific as a function of the signal level, a desired frequency-specific amplification according to audiological principles.

The M gain and attenuation values arrive at the combination 5 of gains and attenuations, where they are combined separately in each frequency range Φ _m , which is done by simple subtraction when using a logarithmic number range:
c _m = g _m - a _m .

The M combined logarithmic gain values c _m arrive at the controllable filter 6, where they are transformed into linear gain values γ _m . The controllable filter 6 with transfer function H (z) can be composed of M parallel filters whose transfer functions H _m (z) have a blocking characteristic only in the frequency range Φ _m and a blocking characteristic in all other frequency ranges and to achieve the desired frequency-dependent modification of Audio signals X are each multiplied by the linear gain value γ _m $$\mathrm{H}\left(\mathrm{z}\right)\mathrm{=}{\mathrm{<figure-callout\; id="1"\; label="\gamma "\; filenames="imgf0001.png,imgf0002.png"\; state="\{\{state\}\}">\gamma </figure-callout>}}_{\mathrm{1}}\mathrm{\cdot }{\mathrm{H}}_{\mathrm{1}}\left(\mathrm{z}\right)\mathrm{+}{\mathrm{\gamma }}_{\mathrm{2}}\mathrm{\cdot }{\mathrm{H}}_{\mathrm{2}}\left(\mathrm{z}\right)\mathrm{+}\mathrm{.....}\mathrm{+}{\mathrm{\gamma }}_{\mathrm{M}}\mathrm{\cdot }{\mathrm{H}}_{\mathrm{M}}\left(\mathrm{z}\right)\mathrm{,}$$

For an update of the controllable filter 6 in time with the sampling rate of the audio signal X, this elementary relationship is not suitable because the computational effort and the associated power consumption of an integrated circuit would be much too large. It is only suitable for one segment Tracking, but this is not optimal because of the reduced temporal resolution in the embodiment shown here by way of example.

wobei die Grösse δH(z)[k] für die exakte Aktualisierung des steuerbaren Filters 6 in einem oder allenfalls einigen wenigen Frequenzbereichen Φ_m steht. Im Falle der Aktualisierung in einem einzigen Frequenzbereich Φ_m gilt folglich $$\mathrm{\delta H}\left(\mathrm{z}\right)\left[\mathrm{k}\right]\mathrm{=}\left({\mathrm{\gamma }}_{\mathrm{m}}\left[\mathrm{k}\right]\mathrm{-}{\mathrm{\gamma }}_{\mathrm{m}}\left[{\mathrm{\kappa }}_{\mathrm{m}}\right]\right)\mathrm{\cdot }{\mathrm{H}}_{\mathrm{m}}\left(\mathrm{z}\right)\mathrm{,}$$

wobei κ_m jenes Abtastintervall bezeichnet, in dem der Frequenzbereich Φ_m das letzte Mal aktualisiert wurde. Es werden also in den vorgegebenen regelmässigen Abtastrespektive Zeitintervallen, vorzugsweise mit der Abtastrate des Eingangssignals, nicht alle, sondern nur ausgewählte Koeffizienten angepasst, vorzugsweise genau einer.In order to achieve a better temporal resolution, the transfer function H (z) of the controllable filter 6 is preferably updated in each sampling interval k according to $$\mathrm{H}\left(\mathrm{z}\right)\left[\mathrm{k}\right]\mathrm{=}\mathrm{K}\left(\mathrm{z}\right)\left[\mathrm{k}\mathrm{-}\mathrm{1}\right]\mathrm{+}\mathrm{delta\; H}\left(\mathrm{z}\right)\left[\mathrm{k}\right]\mathrm{.}$$

wherein the variable δH (z) [k] for the exact update of the controllable filter 6 in one or at most a few frequency ranges Φ _m . In the case of updating in a single frequency range Φ _m , therefore $$\mathrm{delta\; H}\left(\mathrm{z}\right)\left[\mathrm{k}\right]\mathrm{=}\left({\mathrm{\gamma }}_{\mathrm{m}}\left[\mathrm{k}\right]\mathrm{-}{\mathrm{\gamma }}_{\mathrm{m}}\left[{\mathrm{\kappa }}_{\mathrm{m}}\right]\right)\mathrm{\cdot }{\mathrm{H}}_{\mathrm{m}}\left(\mathrm{z}\right)\mathrm{.}$$

where κ _m denotes the sampling interval in which the frequency range Φ _{m was} last updated. Thus, not all, but only selected coefficients, preferably exactly one, are adapted in the predefined regular sampling perspective, time intervals, preferably with the sampling rate of the input signal.

In principle, various options are available for the choice of the frequency range Φ _m to be updated in a specific sampling interval. For example, it is possible to update in each case that frequency range Φ _m for which | c _m [k] -c _m [κ _m ] | is maximum or those frequency ranges Φ _m , in which these quantities exceed a certain threshold, for example 1 dB. Yet another possibility is that m just keeps going through all values from 1 to M systematically or pseudo-randomly over and over again.

In a preferred embodiment of the invention, the following fact is taken into account by means of the correction values r ₁ ..r _n : The noise suppression determines attenuation values which depend only on the modulation depths but not on the signal levels themselves, as is correct for normal hearing persons. Hearing impaired people, whose subjective perception of loudness, however, in In general, in a non-linear manner with the signal level, a signal attenuation by a fixed value a _m will be perceived differently depending on the signal level. In a serial processing, ie in a noise suppression with subsequent compression gain, this effect would be corrected automatically. However, since there is a parallel processing, the correction values r ₁ ..r _{n are} transmitted from the noise suppression to the compression gain to make this correction. Thus, in the noise suppression signal analysis, attenuation-related correction values r _{n are determined} for the N signal levels of the compression gain, and the gain values are calculated with signal levels reduced by these correction values. Thus, the compression gain is corrected according to the noise suppression. This ensures that the signals optimally processed by means of noise suppression for the normal hearing person are individually correctly imaged in the hearing range of each hearing impaired person.

Specifically, this means that, in addition to the already existing signal power s [k], a signal power u [k] reduced as a result of the frequency-specific noise suppression is also calculated for each frequency range Φ _m . For the frequency ranges Φ _m contained in a frequency range F _n , the s [k] and the u [k] are added separately. From the logarithmic ratio of the two sums with respect to F is obtained _n valid logarithmic correction value r _n.

FIG. 3 shows a block diagram for a corresponding signal processing, as in the signal processing for noise suppression 4 takes place to determine the correction quantities r _n . There is shown a case where three frequency ranges φ _{m of} the noise suppression are included in a frequency range of the compression gain. In a block 31, a signal power s [k] is determined in a known manner on signal path 38 and from this a signal level in block 32, and from this in block 33 a modulation depth d _m and therefrom in Block 34 has an attenuation value a _m . In block 35, the logarithmic attenuation value a _{m is} linearly scaled, and the reduced signal power u [k] on signal path 36 is calculated by multiplication with the signal power s [k].

The reduced signal power u [k] is calculated in parallel for each of the three frequency ranges, that is to say for y _m , y _{m + 1} , y _{m + 2} , and summed up in node 37. Signal power s [k] of the three frequency ranges is summed in summation point 39. The sums are scaled logarithmically in blocks 40 and 41 respectively and in subtraction 41 the correction value r _{n is} formed as a difference.

The device according to the invention is preferably implemented at least partially as an analog circuit or microprocessor-based or using application-specific integrated circuits or with a combination of these techniques.

LIST OF REFERENCE NUMBERS

1

first filter unit

2

second filter unit

3

Signal processing for compression amplification

4

Signal processing for noise suppression

5

combination unit

6

controllable filter

7

Means for determining a compression gain

8th

Means for determining a noise suppression

X

input

Y

output

21

achievement Education

22

Level calculation, logarithmic scaling

23

subtraction

24, 24 ', 24 "

gain function

31

achievement Education

32, 40, 41

Level calculation, logarithmic scaling

33

Modulation depth determination

34

Attenuation value determination

35

linear scaling

36

reduced signal power u [k]

37.39

summation

38

Signal power s [k]

42

subtraction

Claims (20)

1. Device for the signal processing in a hearing aid, comprising a filter (6) for the frequency-dependent amplitude adaptation of an input signal (X) and means for the adaptation of coefficients of this filter (6) in accordance with the input signal (X),
   characterised in that the device comprises
   a means for determining coefficients of a compression amplification g_m, which coefficients describe a frequency-dependent adaptation of the input signal (X) in accordance with frequency-dependent signal levels X_n of the input signal (X),
   a means for determining coefficients of a noise suppression a_m, which coefficients describe a frequency-dependent adaptation of the input signal (X) in accordance with interference noises detected in the input signal (X),
   wherein the means for the adaptation of coefficients of the filter (6) establishes these coefficients from the coefficients of the compression amplification g_m and the coefficients of the noise suppression a_m.
2. Device in accordance with claim 1, wherein the means for determining coefficients of the compression amplification g_m comprises a means for determining signal levels p_n in a first number of frequency ranges F_n, with n=1..N, of the input signal (X) and a means for determining the coefficients g_m for the compression amplification for each one of a second number of frequency ranges Φ_m, with m=1..M, of the input signal (X) as function of an optionally modified signal level p_n assigned to the frequency range φ_m.
3. Device according to claim 2, wherein the means for determining signal levels p_n forms these iteratively as momentary effective values of a signal power in the corresponding frequency range F_n.
4. Device according to any preceding claim, wherein the means for determining coefficients of the noise suppression a_m comprises means for determining modulation depths d_m in a second number of frequency ranges φ_m, with m=1..M, of the input signal (X) and a means for determining the coefficients a_m for the noise suppression for each of the frequency ranges φ_m of the input signal (X) in accordance with the corresponding modulation depths d_m.
5. Device according to any of claims 2 to 4, wherein N<M applies and at least one of the frequency ranges F_n for the compression amplification comprises at least two of the frequency ranges φ_m for the noise suppression.
6. Device according to claim 5, wherein the signal processing for the compression amplification 3 is designed to determine each coefficient g_m for the compression amplification respectively as g_m=f_m(p_n) wherein p_n is the optionally modified signal level of that frequency range F_n for the compression amplification which comprises the frequency range φ_m for the noise suppression, and f_m is one of M functions, which in their totality determine a frequency-dependent compression amplification.
7. Device according to claim 6, wherein the coefficients a_m und g_m being combined with one another are logarithmically scaled and their combination by subtraction forms a combined logarithmic amplification value c_m=g_m-a_m.
8. Device according to any preceding claim, wherein the means for the adaptation of coefficients of the filter (6) is designed to adapt not all, but only selected coefficients at predefined time intervals.
9. Device according to any preceding claim, comprising means (23, 35, 36, 37, 38, 39, 40, 41, 42) for the correction of the compression amplification (3) by modification of the signal levels p_n in accordance with the noise suppression.
10. Method for the signal processing in a hearing aid, in which coefficients of a filter (6) for the frequency-dependent amplitude adaptation of an input signal (X) are adapted in accordance with this input signal (X), characterised in that the method comprises the following steps:

    - determining coefficients of a compression amplification g_m, which coefficients describe a frequency-dependent adaptation of the input signal (X) in accordance with frequency-dependent signal levels of the input signal (X),

    - determining coefficients of a noise suppression a_m, which coefficients describe a frequency-dependent adaptation of the input signal (X) in accordance with interfering noises detected in the input signal (X), and

    - calculation of the coefficients of the filter (6) from the coefficients of the compression amplification g_m and the coefficients a_m of the noise suppression.
11. Method according to claim 10, wherein for determining coefficients of the compression amplification g_m in a first number of frequency ranges F_n respectively assigned signal levels p_n, with n=1..N, of the input signal (X) are determined, and the coefficients of the compression amplification g_m for each one of a second number of frequency ranges φ_m, with m=1..M, of the input signal (X) are determined as function of a signal level p_n assigned to the frequency range φ_m.
12. Method according to claim 11, wherein a signal level p_n is iteratively calculated respectively as momentary effective value of a signal power in the corresponding frequency range F_n.
13. Method according to any of claims 10 to 11, wherein for determining coefficients of the noise suppression a_m in a second number of frequency ranges φ_m, with m=1..M, of the input signal (X) modulation depths d_m are determined and the coefficients a_m are determined for each one of the frequency ranges φ_m in accordance with the corresponding modulation depth d_m, wherein the modulation depths d_m are determined from a time-dependent sequence of maximum and minimum values of a signal level p_m in the respective frequency range φ_m, and the signal level p_m is formed in a frequency range φ_m as effective value of the signal power in the corresponding frequency range φ_m.
14. Method according to claim 13, wherein for every modulation depth d_m, which exceeds a predefined value, the assigned coefficient a_m is zero, and for values of the modulation depth d_m below the predefined value, the coefficient a_m increases monotonically with declining modulation depth d_m.
15. Method according to claims 10-14, wherein at least one of the frequency ranges F_n for the compression amplification comprises at least two of the frequency ranges φ_m for the noise suppression, and every coefficient g_m for the compression amplification is determined respectively as g_m=f_m(p_n), wherein p_n is the signal level of that frequency range F_n for the compression amplification, which comprises the frequency range φ_m for the noise suppression, and f_m is one of M functions, which in their totality determine a frequency-independent compression amplification, and wherein the coefficients a_m and g_m are logarithmically scaled and their combination by subtraction forms a combined logarithmic amplification value c_m=g_m-a_m.
16. Method according to claims 10-15, wherein the coefficients of the filter (6) are updated at regular time intervals, wherein, however, during each updating not all, but only a few of the coefficients are updated, in particular only those coefficients, the changes of which are the greatest or exceed a predefined value.
17. Method according to claim 16, wherein the combined coefficients of the filter (6) c_m in the filter (6) are transformed into linear values γ_m and an iterative, frequency-specific updating of a transfer function of the filter (6) in accordance with H(z)[k]=H(z)[k-1]+Σ_m(γ_m[k]-γ_m[k_m])·H_m(z) takes place, wherein H_m(z) only in the frequency range φ_m comprises a pass characteristic and otherwise a blocking characteristic, K_m designates a sampling interval, in which the transfer function for the frequency range φ_m has been updated the last time, and a Summation Σ_m in a sampling interval k respectively only comprises one or some few of the overall M frequency ranges.
18. Method according to claims 10-17, wherein the determination of coefficients of the compression amplification g_m takes into consideration the values of the coefficients of the noise suppression a_m.
19. Method according to claim 18, wherein the coefficients of the compression amplification are determined from modified signal levels p_n' instead of the signal levels p_n, wherein p_n'=p_n-r_n applies, and r_n are logarithmically scaled correction values, which correspond to a signal attenuation caused by the noise suppression.
20. A hearing aid, comprising means for the implementation of the method according to any of claims 10 to 19.

Images