EP3197181B1 - Method for reducing latency of a filter bank for filtering an audio signal and method for low latency operation of a hearing system - Google Patents

Description

The invention relates to a method for reducing the latency of a filter bank for filtering an audio signal, wherein a plurality of signal blocks in the time domain is formed from the audio signal, wherein in each case a filter function is specified for at least a plurality of the signal blocks, the signal Block is filtered with the given filter function transformed into the frequency domain, and thereby a transformed signal block is formed, and signal components of the transformed signal block are output for further processing. The invention further relates to a method for low-latency operation of a hearing system, wherein a first audio signal is generated from a sound signal by a first input transducer, wherein the first audio signal is filtered in a signal processing unit by means of a first filter bank, wherein signal components of the filtered first audio signal in the signal processing unit further processed and for generating an output signal, and wherein an output sound signal is generated from the output signal by an output transducer.

In a hearing aid, an audio signal generated by a microphone is usually transformed from the time domain into the frequency domain after digitization, ie after the digitization the audio signal initially exists in the form of time-resolved samples which, if necessary, become individual signal blocks (so-called "signal blocks"). Frames "), are decomposed by a Fourier transform such as FFT into individual spectral signal components of the generated audio signal. This has the advantage that frequency-selective algorithms such as noise reduction, directional microphone or dynamic compression can be applied. However, the mentioned transformation has the disadvantage that an audio signal converted back into the time domain after a corresponding frequency-selective processing has a delay with respect to the input signal, which is typically of the order of magnitude of several ms. This delay, also known as latency, is greater the higher the resolution in the frequency domain is chosen.

Many hard of hearing suffer primarily from a loss of hearing at high frequencies, such as a noticeably weakened perception from 5-10 kHz, while for low frequencies they show little deviation compared to a normal hearing person. In these cases, mainly high frequencies are significantly increased.

Moreover, an open adaptation of the hearing device is often chosen, in which the sound output from a speaker of the hearing aid output signal via a sound tube with Schirmchen or a listener with Schirmchen in the ear canal is passed to the eardrum. At the eardrum itself thus comes a mixture of a frequency selective attenuated direct sound of the environment as well as the output sound signal generated by the hearing aid. Depending on the hearing loss and the type of adaptation, which in turn influences the attenuation of the direct sound from the environment to the hearing in a frequency-dependent manner, different mixing ratios are therefore found as a function of the frequency.

In the superimposition of correlated signals with time offset, as in the case just described on the eardrum by the direct sound of the environment and the output sound signal of the hearing aid, often occur comb filter effects. These generate characteristic amplitude notches ("notches") at the same distance above the frequency at which an almost complete extinction of the signal component of the corresponding frequency takes place. The greater the time interval between the two superimposed signals, the lower is the frequency domain distance of these amplitude minima. This distorts the signal resulting from the overlay, causing a roaring sound. Especially in the case of binaural audio signal processing, as it is used in binaural hearing systems, the latency is particularly large and therefore the susceptibility to comb filter effects particularly large.

So to avoid these comb filter effects as possible, so it makes sense to reduce the overall latency in the binaural hearing system. However, the described problems with comb filter effects are not tied to a binaural hearing system, but can also occur in a monaural hearing system with only one hearing aid, in which a direct sound of the environment and an output sound signal of a hearing device overlaid with temporal displacement reach the eardrum of the user ,

The temporal offset is primarily due to the internal latency of the hearing system for signal processing and in particular in the filtering.

In the DE 10 2014 204 557 A1 It is described how, in particular for use in a binaural hearing aid in an input signal, wind noise is reduced on the basis of the typical frequency spectrum of the wind noise. For the lowest possible latency, it is proposed in this case to divide the input signal into two sub-signals, and to filter the sub-signals in each case with different frequency resolution and thus latency. In the higher-resolution signal branch filter parameters are now determined, which are applied to the filtered with less latency part signal.

In the DE 693 32 975 T2 A method is described for filtering an input signal by means of a desired impulse response, in which the impulse response in the time domain is decomposed into individual segments which are transformed into the frequency domain and from which respective coefficient blocks are formed for the filtering of the individual time-delayed frames in the frequency domain , The frames thus filtered with the coefficient blocks are summed with their corresponding time delay, and from this a signal in the time domain is generated by inverse transformation, of which in a predetermined time Way still individual signal components are discarded to obtain the final, filtered output signal.

The US Pat. No. 7,251,271 B1 refers to a method to avoid so-called aliasing effects when filtering a discrete input signal with a discrete impulse response. These can occur during the transformation of the individual frames of the input signal from the time domain into the frequency domain and the inverse transformation of the product of impulse response and frequency spectrum of the input signal into the time domain. To avoid the aliasing effects, individual frames are extended before the respective transformation by adding zeros in order to correspond to the respective filter length.

The invention is therefore based on the object of specifying a method for as low-latency spectral filtering of an audio signal with the highest possible spectral resolution. The invention is further based on the object of specifying a method for low-latency operation of a hearing system.

The first object is achieved according to the invention by a method for reducing the latency of a filter bank for filtering an audio signal, wherein a plurality of signal blocks in the time domain is formed from the audio signal. In this case, it is provided that for at least a plurality of the signal blocks in each case a filter function is specified, at least a subinterval of the signal block is given as a prediction period, signal components of the signal block are estimated in the at least one subinterval for the prediction period, and from the for the prediction period of estimated signal components and the signal components of the signal block outside the prediction period, a predicted signal block is generated. It is further provided that the predicted signal block is filtered with the predetermined filter function transformed into the frequency domain, and thereby a transformed signal block is formed, and signal components of the transformed signal block are output for further processing.

The second object is achieved by a method for low-latency operation of a hearing system, wherein from a sound signal through a first input transducer, a first audio signal is generated, wherein the first audio signal is transmitted directly to a signal processing unit, and in the signal processing unit directly by means of a first filter bank according to the above-described method for reducing the latency of a filter bank for filtering an audio signal is filtered, wherein signal portions of the filtered first audio signal in the signal processing unit further processed and used to generate an output signal, and wherein from the output signal directly by an output transducer, an output sound signal is generated. Advantageous and partly inventive in themselves embodiments are set forth in the subclaims and in the description below.

Preferably, a signal block ("frame") in the time domain is formed from the audio signal by the audio signal is converted by time and amplitude discretization in a plurality of each successive time points associated amplitude ("samples"), and a plurality of consecutive samples is combined into a signal block. The further processing of the signal components of the transformed signal block comprises in particular a frequency band-dependent amplification, a frequency band-dependent directional characteristic, a frequency band-dependent noise suppression and a backward transformation of frequency band-dependent treated signal components in the time domain.

The estimation of the signal components for the prediction period of a respective signal block preferably takes place via a prediction algorithm, such as by a linear prediction filter. In particular, an adaptive adaptation of time-correlated coefficients used for the estimation is possible such that an estimation coefficient, which is assigned as a coordinate in the signal block in each case to a sample with a specific time delay, depending on the error between an estimated sample and a real from the Audio signal obtained sample is corrected, whereby the corrector is renewed at periodic intervals. In particular, a signal component estimated for a signal block becomes is also used for a signal block which follows later, if the period corresponding to the signal component still falls within the prediction period of the signal block following later. The prediction period preferably comprises the respectively first and / or the last sample of a signal block. In particular, the period lying outside the prediction period in each case forms a coherent interval in a signal block. In particular, the prediction period comprises the first n samples and / or the last m samples, where n and m natural numbers are smaller than the number of samples in the respective signal block.

An input transducer or an output transducer of the hearing system includes any shape of an acousto-electric or an electro-acoustic transducer, for example a microphone or a loudspeaker. Direct transmission of the first audio signal to the signal processing unit is to be understood as meaning that the transmission of the first audio signal takes place immediately after its generation, that is to say in particular without another, via signal preprocessing, such as signal processing. A / D conversion and / or data compression beyond time delay takes place, as z. B. by a long-term physical storage, which is not based on the FIFO principle ("first-in-first-out"), would occur. In this case, the transmission takes place in particular locally within a hearing device, for example on the signal path predetermined by the signal lines. In particular, however, the transmission also takes place wirelessly, for example from a first hearing device of a binaural hearing system to a second hearing device of the binaural hearing system.

Direct filtering of the first audio signal in the signal processing unit analogously means that the filtering process for the audio signal takes place immediately after its input in the signal processing unit, ie in particular without a further, beyond the direct signal transmission time delay as z. B. by a long-term storage, which is not based on the FIFO principle ("first-in-first-out"), would occur. Likewise, an immediate generation of the output sound signal from the output signal means that immediately after the generation of the output signal is passed through the further processing, the output signal to the output transducer for output, ie in particular without a further, beyond the direct signal transmission outgoing time delay, z. By long-term storage.

In hearing systems, an important portion of the latency falls on the filter banks, which are used to transform the audio signals generated by the input transducers into the frequency domain (analysis filter banks), and the filter banks for the inverse transformation of the frequency resolved, further processed audio signals into the time domain (Synthetic filter banks), the former usually have a larger proportion. Furthermore, in a binaural hearing system, the transmission of an audio signal from one hearing aid to another for the generation of a binaural output signal is associated with a certain delay. However, the latter is difficult to reduce given the restrictions on encoding for transmission. Thus, even in the case of a binaural hearing system, it is advantageous to reduce the latency for the frequency band-wise filtering of the audio signal, that is strictly speaking the analysis filter for the transformation into the frequency domain, for as low-latency operation of the hearing system as possible.

In order to reduce the latency of the analysis filter, it would now be possible to choose the individual signal blocks, which are each used for a filtering process, shorter, ie to process fewer samples in a signal block, since for the processing a signal block should preferably always be present only all the required samples of the signal block. However, since the reduction of the samples in a signal block means a reduction of the information about the signal components available overall in the signal block, this also leads to a reduced frequency resolution in the transformed signal block without the implementation of corrective measures. However, this is undesirable since many signal processing algorithms used in hearing aids require a particularly frequency-selective application for a satisfactory sound character in the end result.

By now estimating the signal components for the prediction period of a signal block instead of using the corresponding signal components actually generated from the audio signal, the effective length of the signal block can be reduced with an appropriate choice of the prediction period, without This affects the frequency resolution of the filter bank. The frequency resolution of the filter bank depends on the temporal information content of the signal blocks to be used for the filter process, ie of their length. Since the signal components are now estimated in a signal block for a period of time, the latency of the filter bank can be reduced by the duration corresponding to the associated prediction period.

Preferably two temporally successive signal blocks overlap each other in part. The definition of the temporal sequence is preferably carried out via a reference sample for the respective signal block, e.g. the first sample. The consequence of the described overlap is that the relevant, successive signal blocks have several, preferably successive, samples in common. On the one hand, this improves the temporal resolution in the frequency domain, since this allows frequent frequency band information to be updated frequently, and on the other hand, the cost of estimating the signal components can be reduced because already estimated signal components are available for a subsequent block without a new estimation process stand.

Expediently, signal portions of the transformed signal block are output separately according to different frequency bands for further processing. For such a transfer, the latency of the filter bank, which is reduced by estimating the signal components of the prediction periods, is particularly advantageous given a constant high frequency resolution.

In each case, the filter function in the prediction period preferably has an average lower transmission amplitude than outside the prediction period. This is to mean that the value of the transmission amplitude of the filter function averaged over the entire prediction period is lower than that over the remaining one Period of the signal block outside the prediction period averaged value of the transmission amplitude of the filter function. In this case, it can be assumed that with a corresponding filtering in the frequency domain by means of the filter function errors that can occur for the prediction by deviating the estimate of the signal components of the real signal components, largely suppressed due to the average lower transmission amplitude of the filter function and thus do not enter significantly into the transformed signal block.

In an advantageous embodiment, the transmission amplitude of the filter function is formed in each case by a logarithmically concave function, wherein the prediction period spans the maximum of the transmission amplitude of the filter function. A logarithmically concave function is defined as a function whose logarithm is concave in the domain of definition - which is given here by the individual samples of the respective signal block. Such a function may for example be given by an approximation of a Gaussian bell curve over a finite, discretized domain of definition. The advantage of the logarithmic concave behavior of the transmission amplitude is that it has a maximum of two inflection points in the domain of definition, and thus is not subject to any oscillations. This results in an advantageous filter behavior, since thus no relevant signal components with a minimum value of an oscillation of the filter function are filtered.

It proves to be particularly expedient if in each case the prediction period contains only convex regions of the transmission amplitude of the filter function. A logarithmic concave function can be represented as a function reciprocal to a particular logarithmic convex function. A logarithmically convex function is in turn convex. This means that the reciprocal, logarithmically concave function due to the reciprocity property has a maximum of two inflection points.

With a suitable choice of the filter function, for example an approximation of a Gaussian bell curve, the maximum of the transmission amplitude lies in a convex region, so that beyond the inflection points the transmission amplitude runs concavely. In these two ranges, the transmission amplitude usually already has sufficiently low values, so that the choice of the prediction time period in at least one of the two ranges can ensure that errors which can occur due to deviations in the estimation of the signal components from the real signal components, be largely suppressed due to the sufficiently lower transmission amplitude of the filter function, and thus not significantly enter into the transformed signal block.

It proves to be further advantageous if an empty signal is respectively estimated as signal components for the prediction period of at least one signal block. An empty signal is in this case the signal which has no amplitude for the period in question. The estimation of an empty signal is carried out in particular in the event that the signal components of the audio signal which are used for the estimation method of the signal components of the prediction period do not allow a qualitatively sufficiently high-quality estimate of the signal components as a result of insufficient correlations. This can occur, for example, if there is a high proportion of white noise in the audio signal, which reduces the correlation of successive samples and thus makes prediction more difficult.

In particular, by means of a prediction, estimated signal components, which are different from the empty signal, with respect to the quality of the estimate, are to be compared with the corresponding real signal components of the audio signal in order to be able to evaluate the quality of the prediction. In the case of excessive deviation - defined by a deviation measure such. B. an averaged over several samples difference and an associated upper bound for the deviation measure - is determined instead of the predicted signal components, an empty signal as estimated for the prediction signal component. It is also possible to check the signal components of the audio signal for correlations even before the prediction, and to set an empty signal as signal component for the prediction period if the correlation is too low.

In a further advantageous embodiment of the method for low-latency operation of a hearing system, a second audio signal is generated from the sound signal by a second input transducer spatially separated from the first input transducer, the second audio signal is transmitted directly to the signal processing unit and filtered by a second filter bank, and wherein signal components of filtered second audio signal in the signal processing unit further processed and used to generate the output signal.

In particular, the filtering of the second audio signal by means of the second filter bank in accordance with the above-described method for reducing the latency of a filter bank for filtering an audio signal. Immediate transmission of the second audio signal to the signal processing unit is understood to mean that the transmission of the second audio signal without another, via a signal preprocessing such. A / D conversion and / or data compression and the direct signal transmission beyond time delay takes place, as z. B. by a long-term physical storage, which would not based on the FIFO principle ("first-in-first-out") would occur.

This named embodiment makes possible, in particular, a low-latency operation of a binaural hearing system, taking into account the special features occurring in such a hearing system as a result of the signal transmission from one hearing aid to another occurring for the generation of the binaural auditory sensation. Since, in the case of a binaural hearing system for compression, the real information content of signal portions of the audio signal received by the respective other hearing device for the generation of the binaural hearing sensation is often reduced for better transmission, for example by data compression, this is possible by estimating the signal components in the prediction period reduced errors in its meaning. In the case of this audio signal, there is already a loss of information due to the transmission, so that the deviations by the estimation for the prediction period do not represent an additional cumulative source of error but only an alternative source of error to be considered as an alternative. In short, it makes little difference whether an error is made statistically by the data compression or the estimation.

A further advantage of using the method for low-latency operation of a binaural hearing system is that a certain latency of several ms is already introduced into the hearing system through the described transmission of the audio signals. The reduction of further possible latencies, e.g. in the present case through the filter banks, here helps to minimize the losses of the sound quality by comb filter effects as low as possible.

The invention further provides a hearing aid, comprising at least one input transducer for generating an audio signal, an output transducer for generating an output sound signal, and a local signal processing unit having a first filter bank, which for carrying out the above-described method for reducing the latency of a filter bank for filtering an audio signal is set up. The advantages stated for the method and its developments can be transferred analogously to the hearing aid.

The invention also mentions a binaural hearing system with two hearing aids as described above, which is set up to carry out the method for low-latency operation of a hearing system with at least two input transducers. The advantages specified for the method and its developments can be transferred analogously to the binaural hearing system.

Fig. 1

in einem Blockdiagram ein binaurales Hörsystem mit zwei Hörgeräten, und

Fig. 2

in einer Zeitdarstellung ein von einem Hörgerät nach Fig. 1 erzeugtes Audiosignal und in einer Ausschnittdarstellung ein Signal-Block des Audiosignals mit einer Filterfunktion und einem Prädiktionszeitraum.

An embodiment of the invention will be explained in more detail with reference to a drawing. Here are shown schematically in each case:

Fig. 1

in a block diagram a binaural hearing system with two hearing aids, and

Fig. 2

in a time representation of a hearing aid Fig. 1 generated audio signal and in a detail representation of a signal block of the audio signal with a filter function and a prediction period.

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

In Fig. 1 1 is a schematic block diagram of a binaural hearing system 1. The binaural hearing system 1 is in this case formed by a first hearing device 2 and a second hearing device 4. The first hearing device 2 has a first input transducer 8 configured as a microphone 6, which generates a first audio signal 10 from a sound signal 9. The second hearing device 4 has a second input transducer 14 configured as a microphone 12, which generates a second audio signal 16 from the sound signal 9. The first audio signal 10 and the second audio signal 16 are respectively prepared in the respective hearing device 2, 4 by a local signal preprocessing 18, 20, which in each case in particular comprises an A / D conversion, for the further signal processing processes. In this case, the local signal preprocessing 18, 20 comprises in particular only runtime processes, ie those processes which do not involve any further delay, in particular no longer-term storage and charging processes of the signal components, over the duration of the signal processing taking place.

The first audio signal 10 is first transmitted immediately after the local signal pre-processing 18 in a binaural transmission process 22 from the first hearing aid 2 to the second hearing aid 4, where it is filtered in a signal processing unit 24 in a first filter bank 26 in a manner to be described. The binaural transmission process 22 takes place immediately after the local signal preprocessing 18, that is, in particular without further delay, in particular without longer-term storage and recharging the relevant signal components on a FIFO memory addition. The filtered first audio signal 28 will now be subjected to frequency bandwise signal processing algorithms 30, e.g. Noise suppression, directional microphone or dynamic compression applied.

The second audio signal 16 is supplied immediately after the local signal pre-processing 20 of the signal processing unit 24, where it is first filtered in a second filter bank 32 in a manner to be described, wherein as a filtered second audio signal 34, the respective signal components are transmitted separately in individual frequency bands. In the filtered second audio signal 34 resulting from the second filter bank 32, the respective signal components are output separately in individual frequency bands. Also on the filtered second audio signal 34 frequency bandwise signal processing algorithms 28 such as noise reduction, directional microphone or dynamic compression are now applied. From the filtered first audio signal 26 and the filtered second audio signal 34, an output signal 36 is generated after the frequency band-wise signal processing 28, which locally reflects the binaural hearing at the location of the second hearing device 4.

The output signal 36 is converted directly, ie in particular without further long-term storage and recharges of the signal components, from an output transducer 40 designed as a loudspeaker 38 into an output sound signal 42.

In Fig. 2 is against a time axis t, the first audio signal 10 after Fig. 1 which is split into individual partially overlapping signal blocks 50a-f. The individual signal blocks 50a-f are in this case formed from a large number of successive samples of the first audio signal 10, individual samples occurring as a result of the overlap of the successive signal blocks 50a-f in at least two signal blocks. The individual signal blocks 50a-f are now each transformed in a manner to be described, the frequency domain. Due to the short time interval of each of two consecutive signal blocks 50a-f, the spectral signal components of the first audio signal 10 can thus be updated in short intervals in the frequency domain. Due to the relatively high number of individual samples and thus the high time-resolved information content per signal block 50a-f, there is also a high spectral resolution of the first audio signal 10 after transformation into the frequency domain. In order to reduce the high latency occurring in the filtering process and the transformation into the frequency domain at a high temporal resolution, the individual signal blocks 50a-f are determined Signal portions, which is shown for the signal block 50c on the basis of a detail representation.

For the signal block 50c, the individual real signal components 52a, 52b are shown against a time axis t '. The real signal components 52a, 52b are each given by the amplitude of the corresponding sample. Further, for the signal block 50c, the transmission amplitude 54c of the filter function 56c is shown, which in the present case is approximately given by a Gaussian bell curve.

The filter function 56c in this case represents a window function with which the edges of the signal block 50c for the transformation into the frequency domain are to be smoothed out. This is because without such a window function, the Fourier transform of the signal components of the signal block 50c is de facto a Fourier transform of the signal components of the first audio signal 10, which are multiplied by a rectangular function corresponding to the duration of the signal block. As a result of the convolution theorem, this multiplication in the time domain means a convolution of the frequency components of the first audio signal 10 with the Fourier transform of the rectangular function, which is given by a strongly oscillating sin (x) / x or Sinc function. In order to avoid such oscillations, the edges of the signal block 50c for the transformation into the frequency domain are "hidden" by means of a suitable filter function 56c. This is done by the transmission amplitude 54c of the filter function 56c converges to zero at the edges of the signal block 50c as free of oscillation as possible, ie in particular with as few turning points as possible. In particular, a function having such properties is given by a logarithmic concave function, such as e.g. the approximated Gaussian bell curve of the present case.

The described profile of the transmission amplitude 54c of the filter function 56c can now be exploited to reduce the latency of the first filter bank 24, without losing any of the resolution capability in the frequency domain. For this purpose, a sub-interval 58c at the end of the signal block 50c as a Prediction period 60c defined. The sub-interval 58c lies beyond the point of inflection 62c of the transmission amplitude 54c, ie in particular far away from the maximum 64c of the transmission amplitude 54c, so that in the sub-interval 58c, which defines the prediction time 60c, the transmission amplitude 54c has only low values. For the prediction period 60c, the signal components to be used for the transformation are now estimated there by means of a prediction algorithm, for example a linear prediction filter, instead of the real signal components 52b. The signal portions 66b estimated in the prediction period 60c and the signal portions 52a of the signal block 50c outside the prediction period 60c now form a predicted signal block 68c.

This predicted signal block 68c is now multiplied by the filter function 56c and transformed into the frequency domain by means of a fast Fourier transformation, so that there the frequency-resolved information of the transformed signal block 50c is available for further processing by means of frequency band-dependent signal processing algorithms stands. For the other signal blocks 50a, 50b, 50d-f, too, the procedure described is used to estimate signal components for a prediction period which is to be chosen favorably on the basis of the respective filter function to be used, in order thus to reduce the latency for the transformation into the frequency domain. since then the last samples of a signal block need not yet exist, so that the transformation can be started several ms earlier due to the estimation.

An important role in this case plays the course of the transmission amplitude 54c of the filter function 56c. A possible error which could result from the deviation of the signal portions 66b estimated for the prediction period 60c from the real signal portions 52b is suppressed by the fact that for the prediction period 60c the transmission amplitude 54c has only comparatively small values relative to its maximum 64c, and thus by the corresponding multiplication with the filter function 56c, the estimated signal components 66b make only a small contribution to the transformed signal block anyway. However, this contribution is important for the spectral resolution. In particular, tonal signal components can be estimated relatively well anyway by means of conventional prediction methods. Even with a white noise, which is unfavorable due to its static properties, due to the mentioned suppression of the errors due to possible deviations, the described method gives good results.

In the binaural hearing system 1 of Fig. 1 is the first audio signal 10 in the first filter bank 24 according to the basis of Fig. 2 described method filtered. The filtering of the second audio signal 16 in the second filter bank 32 can be done in the same way; however, a conventional filter method-that is to say without estimation of signal components for a respective prediction period of the individual signal blocks-can also be used for this purpose. The decision on this is made in particular as a function of the tolerable overall latency of the binaural hearing system 1 and the delay caused by the binaural transmission process.

Although the invention has been illustrated and described in detail by the preferred embodiment, the invention is not limited by this embodiment. Other variations can be deduced therefrom by those skilled in the art without departing from the scope of the invention.

LIST OF REFERENCE NUMBERS

1

Binaural hearing system

2

first hearing aid

4

second hearing aid

6

microphone

8th

first input converter

9

sound signal

10

first audio signal

12

microphone

14

second input converter

16

second audio signal

18

local signal preprocessing

20

local signal preprocessing

22

binaural transmission process

24

Signal processing unit

26

first filter bank

28

filtered first audio signal

30

frequency bandwise signal processing

32

second filter bank

34

filtered second audio signal

36

output

38

speaker

40

output transducer

42

Output sound signal

50a-f

Signal block

52a, b

real signal components

54c

transmission amplitude

56c

filter function

58c

subinterval

60c

Prädiktionszeitraum

62c

turning point

64c

maximum

66b

estimated signal components

68c

predicted signal block

t, t '

timeline

Claims (11)

1. Method for reducing the latency period of a filter bank (26, 32) for filtering an audio signal (10, 16), wherein a large number of signal blocks (50a-f) in the time domain are formed from the audio signal (10, 16), wherein for at least a plurality of the signal blocks (50a-f) in each instance

   - a filter function (56c) is predetermined,

   - at least one partial interval (58c) of the signal block (50a-f) is predetermined as a prediction period (60c),

   - signal components (66b) of the signal block (50a-f) in the at least one partial interval (58c) are estimated for the prediction period (60c), and a predicted signal block (68c) is generated from the signal components (66b) estimated for the prediction period (60c) and from the signal components (52a) of the signal block (50a-f) outside the prediction period (60c), and

   - the predicted signal block (68c), filtered with the predetermined filter function (56c), is transformed into the frequency domain, and by this means a transformed signal block is formed, and

   - signal components of the transformed signal block are output for further processing.
2. Method according to Claim 1,
   wherein each two temporally consecutive signal blocks (50a-f) partially overlap.
3. Method according to Claim 1 or Claim 2,
   wherein in each instance signal components of the transformed signal block according to various frequency bands are output separately for further processing (30) .
4. Method according to one of the preceding claims,
   wherein in each instance the filter function (56c) exhibits a smaller - on average - transmission amplitude (54c) within the prediction period (60c) than outside the prediction period (60c).
5. Method according to Claim 4,
   wherein the transmission amplitude (54c) of the filter function (56c) is constituted in each instance by a logarithmically concave function, and
   wherein the prediction period (60c) omits the maximum (64c) of the transmission amplitude (54c) of the filter function (56c).
6. Method according to Claim 5,
   wherein in each instance the prediction period (60c) includes only convex regions of the transmission amplitude (54c) of the filter function (56c).
7. Method according to one of the preceding claims,
   wherein a blank signal is estimated in each instance as signal components (66b) for the prediction period (60c) of at least one signal block (50a-f).
8. Method for low-latency operation of a hearing system (1),
   wherein a first audio signal (10) is generated from an acoustic signal (9) by a first input transducer (8), wherein the first audio signal (10) is transmitted immediately to a signal-processing unit (24) and filtered immediately in the signal-processing unit (24) by means of a first filter bank (26) in accordance with a method according to one of the preceding claims, wherein signal components of the filtered first audio signal (28) are subjected to further processing (30) in the signal-processing unit (24) and used for generating an output signal (36), and
   wherein an output acoustic signal (42) is generated immediately from the output signal (36) by an output transducer (40).
9. Method according to Claim 8,
   wherein a second audio signal (16) is generated from the acoustic signal (9) by a second input transducer (14) spatially separated from the first input transducer (8),
   wherein the second audio signal (16) is transmitted immediately to the signal-processing unit (24) and filtered by means of a second filter bank (32), and wherein signal components of the filtered second audio signal (16) are subjected to further processing in the signal-processing unit (24) and used for generating the output signal (36).
10. Hearing aid (2, 4), comprising at least one input transducer (8, 14) for generating an audio signal (10, 16), an output transducer (40) for generating an output acoustic signal (42), and also a signal-processing unit (24) with a first filter bank (26), which hearing aid has been set up to implement the method according to Claim 8 or Claim 9.
11. Binaural hearing system (1) with two hearing aids (2, 4) according to Claim 10, which has been set up to implement the method according to Claim 9.

Images