EP4235664A1 - Method for suppressing an acoustic reverberation in an audio signal - Google Patents

Abstract

The invention relates to a method for suppressing acoustic reverberation in an audio signal (50), wherein an audio signal (50) is provided, wherein a first level measurement (p1) of the audio signal (50) is carried out, wherein during the first level measurement (p1) a second level measurement (p2) of the audio signal (50) is carried out, the first level measurement (p1) being carried out using a first adjustment parameter and a first decay parameter in such a way that the first level measurement (p1) has a first adjustment time (6) and has a first decay time (16), the second level measurement (p1) being carried out by means of a second control parameter and a second decay parameter in such a way that the second level measurement (p2) has a second control time (12 ) and a second decay time (18) which is greater than the first decay time (6), and wherein a difference (Δ) is formed from the first level measurement (p1) and the second level measurement (p2). It is provided that a reverberation noise level is estimated based on the difference (Δ) and based on the second level measurement (p2), and based on the first level measurement (p1) and based on the reverberation noise level (prn) an amplification parameter for the audio signal (50th ) is determined.

Description

The invention relates to a method for suppressing acoustic reverberation in an audio signal, with a first level measurement of the audio signal being carried out, with a second level measurement of the audio signal being carried out during the first level measurement, with the first level measurement having a first settling time and a first decay time, with the second level measurement has a second settling time identical to the first settling time and a second decay time which is greater than the first decay time, and a difference is formed between the first level measurement and the second level measurement,

Acoustic reverberation usually occurs in closed or at least partially closed rooms as a result of multiple reflections of a generating sound event on the walls of a room and on other objects that are present in the room. Depending on the geometry of the room and its walls as well as the type, number and geometry of the objects in the room, the decay time of the reverberation varies, which is also influenced by the nature of the surfaces in the room. In contrast to the echo, which can be perceived in isolation as a type of "repetition" of the generating sound event, the reverberation here forms an essentially continuous "ringing over" of the sound event.

While a minimum of reverberation is even desirable for a pleasant sound perception, especially of music, in order to counteract a too "dry", staccato-like sound, acoustic reverberation is for a Intelligibility of speech contributions is often disadvantageous, since the characteristic sound events for distinguishing the individual consonants in particular are of very short duration, and a corresponding superimposition with the reverberation can sometimes falsify the spectral information considerably. Depending on the decay time, this can even become a problem for distinguishing the formants for recognizing vowels.

In hearing instruments in which audio signals are reproduced for a wearer, it is particularly important for the wearer to be able to play back speech contributions by the wearer’s conversation partner in a way that is as easy to understand as possible, since a lack of acoustic understanding of a speech contribution and the associated loss of information that the wearer can see is particularly important can be perceived clearly and therefore particularly unpleasantly. This applies in particular to hearing aids "in the narrower sense", which are often used to compensate for a hearing loss of the wearer in question. For this reason, techniques for improving the intelligibility of speech contributions are often used in hearing instruments, above all in the hearing aids mentioned.

However, especially when dynamic compression is used frequently, acoustic reverberation can impair the so important speech intelligibility to a particular degree: dynamic compression is intended to contribute in particular to quiet sound events that are hardly or no longer perceived by the wearer (whether as a result of a hearing loss of the wearer, or as a result of the fundamentally low sound level of the sound event), up to a sufficient level of perceptibility, without applying the same amplification to sufficiently loud sound events, which can be perceived by the wearer without major problems, and thus another Amplification could lead to an uncomfortable volume.

However, this dynamic compression also "compresses" the acoustic reverberation, so that it experiences a correspondingly higher amplification than the sound event that generated it. As a result, on the one hand, the decay time in the present environment is perceived by the wearer as longer, and on the other hand, as a result of the relationships described above, speech contributions are impaired in their intelligibility.

In the DE 10 2018 210 143 A1 discloses suppressing reverberation in an audio signal by taking two level measurements of the reverberation with different time constants and performing an attenuation of the audio signal based on the difference in the level measurements.

It is the object of the present invention to improve the method described above for suppressing reverberation in an audio signal.

The stated object is achieved according to the invention by a method for suppressing acoustic reverberation in an audio signal, an audio signal being provided, a first level measurement of the audio signal being carried out, a second level measurement of the audio signal being carried out during the first level measurement, the first level measurement is carried out using a first adjustment parameter and a first decay parameter in such a way that the first level measurement has a first adjustment time and a first decay time, the second level measurement being carried out using a second adjustment parameter and a second decay parameter in such a way that the second level measurement has a second settling time identical to the first settling time and a second decay time which is greater than the first decay time, and wherein a difference is formed from the first level measurement and the second level measurement.

It is provided here that a reverberation noise level is estimated on the basis of the difference and on the basis of the second level measurement, and an amplification parameter for the audio signal is determined on the basis of the first level measurement and on the basis of the reverberation noise level. Advantageous and partly inventive configurations are the subject matter of the subclaims and the following description.

Suppression of an acoustic reverberation in an audio signal includes in particular a suppression of those signal contributions in the audio signal which arise from acoustic reverberation in the real acoustic situation depicted by the audio signal. In this case, the audio signal is provided in particular by means of one or more electroacoustic converters, which convert said real acoustic situation into one or more, in particular, electrical signals. In order to provide the audio signal based on the generated, in particular electrical signal(s), pre-processing can also be carried out, which can include, for example, digitization, amplification, dynamic compression or also noise suppression. Acoustic reverberation here includes, in particular, reflections of sound from a generating sound event on walls and/or objects, e.g. an at least partially closed room, with multiple reflections of the propagating sound, which was generated by the sound event, at a fixed location resulting in a continuous or an almost continuous decay of the sound event occurs.

In the present case, a level measurement means in particular that a mathematical function is formed by the level measurement or that the level measurement can be represented as such a function by which an amplitude of the audio signal and/or an envelope of the amplitude and/or a square of the absolute value of the amplitude is mapped onto a corresponding level value in a preferably strictly monotonic manner, and particularly preferably without inflection points. Here, in particular, such functions should also be considered in which the connection between their input variable and the displayed level value is not only logarithmic, but rather the term level measurement should also include more general functions with a suitable monotonic behavior.

A decay time of a level measurement is to be understood here in particular as the time which elapses after a signal contribution in the audio signal and a corresponding level deflection of the level measurement until the level measurement drops to zero or to zero in the absence of further signal contributions in the audio signal has dropped a predetermined fraction of the level deflectionB. A settling time of a level measurement is to be understood in particular as the time which elapses after a spontaneously occurring, stationary signal contribution in the audio signal until the level measurement has reached a predetermined proportion of the asymptotic limit value for the signal level, which corresponds to the stationary signal contribution. In this case, a shorter settling time means in particular a faster response of the level measurement to a spontaneously occurring signal contribution in the audio signal.

The settling time and the decay time are set for each of the two level measurements using the respective settling and decay parameters. If, for example, the level measurements are implemented by an asymmetric, smoothing function of the amplitude (e.g. a recursive averaging function or similar), the adjustment parameter can be given by the weighting factor of the next amplitude contribution for the rising edge, and the decay parameter can accordingly be given by the weighting factor of the respective next amplitude contribution for the falling edge.

A sound event includes, in particular, any sound-generating event in the real acoustic situation, which is represented by the audio signal and/or is implemented by means of appropriate converters to provide the audio signal, with a clear end being able to be assigned to the sound-generating event in terms of time. In this sense, the suppression of the acoustic reverberation of the sound event in the audio signal means in particular a suppression of those signal contributions which correspond to the acoustic reverberation of the sound event in the real acoustic situation.

The decay behavior of the two level measurements, characterized by the respective decay time, depends on the specific behavior of a room in which the provided audio signal is recorded. Immediately after the sound event, the first wavefront enters the audio signal directly and without further reflection, followed by the first reflections at various boundaries and/or objects in the room, and their delays in transit time especially dependent on the size of the room. These first reflections, which still represent a kind of weakened and delayed version of the original sound event, now go into the audio signal on the one hand, but on the other hand generate further, cascading reflections. With increasing order of reflection and superimposition of the individual wavefronts, the ability to distinguish individual reflections is lost; After the early reflections, which can usually still be isolated, an (almost) continuous, diffuse reverberation trail forms, which falls off exponentially.

The contribution of the diffuse reverberation in the audio signal can be determined at least approximately and implicitly on the basis of the difference between the two level measurements, particularly if the two decay times are selected appropriately. This applies in particular in the event that the second decay time is preferably set using the second decay parameter (which is preferably estimated for the present environment or the present room) in such a way that the decay behavior of the second level measurement is essentially determined by the contributions of the original Sound event is determined, which are gradually less and less included in the second level measurement. The first decay time is preferably set using the first decay parameter in such a way that the decay behavior of the first level measurement, at least after a short time of early reflections, is essentially determined by the contributions of the diffuse reverberation, which continue to "feed" the decaying first level measurement. In particular, the short time of the early reflections can be recognized based on the difference between the two level measurements, as can the diffuse reverberation that follows or transitions from it.

If the contribution of the diffuse reverberation or the decaying contribution of the sound event is known in the second level measurement, for example by using a so-called minimum tracker on the difference between the second and the first level measurement (this difference is generally negative), and it is additionally determined when the minimum reached is left again (or is exceeded by a predetermined minimum value), the reverberation noise level can be estimated from this knowledge using the second level measurement, which in particular the proportion can represent in the second level measurement, which (essentially) is based on or corresponds to the contributions of the diffuse reverberation (instead of the actual, decaying sound event).

The reverberation noise level can then be used to determine an amplification parameter, so that an acoustic and in particular diffuse reverberation of the sound event in the audio signal is suppressed by weakening the audio signal as a function of the amplification parameter, particularly in response to a contribution from the sound event in the audio signal. The amplification parameter can be determined as an amplification factor, e.g. depending on the reverberation noise level as an interference signal and the first level measurement as a useful signal, which is applied to the audio signal.

In this case, the first level measurement and/or the second level measurement is expediently implemented by a weighted mean value function. This allows the different decay behavior of the level measurements to be implemented particularly easily with identical adjustment behavior by applying different weighting factors of the recursion to new audio signal contributions to the first or second level measurement for the falling edge. A weighting factor for a subsequent value of the weighted mean value function is preferably selected "asymmetrically" depending on a rising or falling level, i.e. a value of the audio signal at a point in time is compared with the level value at that point in time according to the weighted mean value function , and a weighting factor for the input of the new value of the audio signal into the level measurement is selected depending on whether the value of the audio signal is larger or smaller than the current value of the level measurement.

1. (i) $${\mathrm{pj}}^2\left(\mathrm{n}\right)=\left(1-\mathrm{c}\left(\mathrm{n}\right)\right)\cdot {\mathrm{a}}^2\left(\mathrm{n}\right)+\mathrm{c}\left(\mathrm{n}\right)\cdot {\mathrm{pj}}^2\left(\mathrm{n}-1\right)$$
   
   mit
   c (n) = c_steig für a² (n) > pj² (n-1), c_fall sonst,

wobei pj² (n) den Pegelwert der Pegelmessung pj zum diskreten Zeitindex n bezeichnet, und c (n) einen Steilheitsparameter, welcher in Abhängigkeit der o.g. Bedingung für a² (n) auf eine der beiden Konstanten c_steig, c_fall gesetzt wird.The first level measurement or the second level measurement is advantageously implemented by an asymmetric recursive low-pass filter of preferably the first order. For an audio signal a(n) in the discrete time domain, the respective level measurement pj (j=p1, p2) can then be represented as

1. (i) $${\mathrm{pj}}^2\left(\mathrm{n}\right)=\left(1-\mathrm{c}\left(\mathrm{n}\right)\right)\cdot {\mathrm{a}}^2\left(\mathrm{n}\right)+\mathrm{c}\left(\mathrm{n}\right)\cdot {\mathrm{pj}}^2\left(\mathrm{n}-1\right)$$
   
   with
   c(n) = c _rise for a ² (n) > pj ² (n-1), c _fall otherwise,

where pj ² (n) denotes the level value of the level measurement pj for the discrete time index n, and c (n) a slope parameter which is set to one of the two constants c _rise , c _fall depending on the above condition for a ² (n).

A physical decay time constant of a given environment is expediently determined, in which a sound level of a sound signal on which the audio signal is based (i.e. in particular a sound signal from which the audio signal is generated) has fallen to a predetermined proportion of an initial value, with the second Decay parameter is chosen such that the second decay time of the second level measurement is given by said physical decay time constant for the present environment. A drop by 60 dB, which corresponds to the time constant T60, is preferably used here as the drop to the predetermined proportion of an output value. A drop of 60 dB generally corresponds to complete decay down to the background noise.

Conveniently, the difference between the first level measurement and the second level measurement is compared with a specified first limit value, with the event that the magnitude of the difference between the two level measurements exceeds the magnitude of the first limit value, the presence of a contribution from the diffuse reverberation and/or a decaying contribution of the sound event is determined in the second level measurement. This includes, in particular, that the difference is formed and, if this is negative, the presence of a diffuse reverberation is determined when this difference is below the first limit value, which is now also negative. As soon as the difference exceeds the first limit value again (or the absolute value of the difference falls below the absolute value of the first limit value), there is no longer any diffuse reverberation in this implementation, so that there is preferably no weakening either. The weakening of the audio signal is preferably controlled as a function of a comparison of said difference with the first limit value.

In the event that the amount of the difference from the two level measurements exceeds the amount of the first limit value, the difference between the amount of the difference and the amount of the limit value is preferably determined as the decaying contribution of the sound event in the second level measurement. This includes, in particular, that in the case of a negative difference in the level measurements, the value by which the difference falls below the first limit value is determined as the contribution of the diffuse reverberation, which in particular is quantitatively included in the reverberation noise level. In other words, the first limit value provides on the one hand a binary criterion for whether diffuse reverberation or a decaying contribution of the sound event is present at all, and on the other hand a quantitative measure for said decaying contribution if it is present. In particular, a minimum tracker can be used for said difference, and the determined minimum can be compared with the first limit value, for example. The decaying contribution of the sound event includes in particular those parts in the second level measurement which are based essentially or solely on the actual sound event and only gradually decrease or disappear due to the time delay as a result of the smoothing. It is particularly advantageous here if the second decay time of the second level measurement is selected as the decay time constant T60; During a phase of the late, especially diffuse reverberation, the first level measurement, which initially shows a faster first decay time (and thus a faster decay behavior) for early reflections, also decays with this decay rate in the room as a result of the sound power of the diffuse reverberation.

A time-dependent correction function is preferably generated based on the decaying contribution of the sound event, and in particular also based on the magnitude of the first limit value, with the reverberation noise level being generated based on a subtraction of the correction function from the second level measurement. The time-dependent correction function can in particular be given by a base value which is dependent on the first limit value and on the contributions of the decaying sound event determined as described above.

• (ii) $$\mathrm{G}\left(\mathrm{n}\right)=1-\mathrm{prn}\left(\mathrm{n}\right)/\mathrm{p}1\left(\mathrm{n}\right)$$
  
  mit prn (n) = p2 (n) - d (n) als dem Nachhall-Störpegel, p1 und p2 als der ersten und der zweiten Pegelmessung, sowie d (n) als der Korrekturfunktion gemäß
• (iii) $$\mathrm{d}\left(\mathrm{n}\right)=\min \left[\mathrm{p}1\left(\mathrm{n}\right)-\mathrm{p}2\left(\mathrm{n}\right),\mathrm{th}1\right]$$
  
  mit dem (negativen) ersten Grenzwert th1.

The amplification parameter is advantageously determined using a spectral subtraction, for which a quotient of the reverberation noise level and the first level measurement is used in particular. The gain parameter can then be given, for example, as

• (ii) $$\mathrm{G}\left(\mathrm{n}\right)=1-\mathrm{prn}\left(\mathrm{n}\right)/\mathrm{p}1\left(\mathrm{n}\right)$$
  
  with prn (n) = p2 (n) - d (n) as the reverberation noise level, p1 and p2 as the first and second level measurements, and d (n) as the correction function according to
• (iii) $$\mathrm{i.e}\left(\mathrm{n}\right)=\mathrm{at\; least}\left[\mathrm{p}1\left(\mathrm{n}\right)-\mathrm{p}2\left(\mathrm{n}\right),\mathrm{th}1\right]$$
  
  with the (negative) first limit value th1.

In an advantageous embodiment, the audio signal is broken down into a plurality of frequency bands, with the first level measurement and the second level measurement being carried out for each frequency band, with the respective amplification parameter being determined for a plurality of frequency bands, in particular by forming the difference between the two level measurements for each frequency band, using which determines the respective decaying contribution of the sound event in the frequency band and from this the respective reverberation noise level is determined, and wherein the respective amplification parameter in the frequency band is applied to the signal portion of the audio signal to suppress the acoustic reverberation. The gain parameter G(n) in equation (ii) is in this case to be replaced by a corresponding plurality of gain parameters G(n,k) in the time-frequency domain, where k is the band index.

The invention further calls a method for suppressing an acoustic reverberation in an audio signal of a hearing instrument, in particular a hearing aid, wherein using an input transducer of the hearing instrument from a sound signal of the environment, the audio signal is provided, and acoustic reverberation in the audio signal is suppressed by the method described above, and a hearing instrument, which can be given in particular as a hearing aid, with an input converter for generating an audio signal and a signal processing unit, which is used to carry out the procedure described above is set up.

The method in the hearing instrument and the hearing instrument itself share the advantages of the method for suppressing reverberation described above. The advantages specified for the method for suppressing an acoustic reverberation in an audio signal and for its developments can be transferred analogously to the method in the hearing instrument and to the hearing instrument itself.

Fig. 1

in einem Zeitdiagramm eine erste Pegelmessung und eine zweite Pegelmessung desselben Audiosignals mit identischen Einregelzeiten und jeweils unterschiedlichen Abklingzeiten,

Fig. 2

in einem Zeitdiagramm eine Differenz der Pegelmessungen nach Fig. 1, und eine hieraus ermittelte Korrekturfunktion,

Fig. 3

in einem Zeitdiagramm die zweite Pegelmessung und ein anhand der Korrekturfunktion nach Fig. 2 ermittelter Nachhall-Störpegel,

Fig. 4

in einem Zeitdiagramm der Nachhall-Störpegel nach Fig. 3 und die erste Pegelmessung, und

Fig. 5

in einem Blockdiagramm ein Hörinstrument.

An exemplary embodiment of the invention is explained in more detail below with reference to drawings. Here each show schematically:

1

in a timing diagram, a first level measurement and a second level measurement of the same audio signal with identical settling times and different decay times,

2

shows a difference in the level measurements in a time diagram 1 , and a correction function determined from this,

3

in a time diagram the second level measurement and a correction function based on 2 determined reverberation noise level,

4

in a time chart of the reverberation noise level 3 and the first level measurement, and

figure 5

in a block diagram a hearing instrument.

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

In figure 1 the level values P of a first level measurement p1 (continuous line) and a second level measurement p2 (dashed line), which are each carried out on an audio signal not shown in detail, are shown schematically in a time diagram against a time t. At a point in time T0 there is an isolated sound event 4 (dotted line) in the audio signal, which on the one hand has a clearly defined end and on the other hand, due to the physical environment in which the audio signal was recorded for generation, causes contributions of acoustic reverberation in the audio signal (not drawn). In this case, the sound event 4 should only have a very short duration z. The entire sound energy of the sound event 4 is therefore concentrated in this time period z. This can be the case, for example, with a bang, a hit, a clapping, but also with consonants of speech, especially plosives, and similar noises of very short duration.

The first level measurement p1 has a first settling time 6, which elapses after the point in time T0, at which the sound event 4 begins, until the first level measurement has assumed a predetermined proportion 8 of the asymptotic level 10, with the asymptotic level 10 corresponding to that level which the first level measurement for a stationary, continuous sound event with a signal level identical to sound event 4 would assume. The second settling time 12 of the second level measurement p2 is presently identical to the first settling time 6 of the first level measurement.

For this reason, the first level measurement p1 and the second level measurement p2 have the same adjustment behavior and thus at time T1, which marks the end of the time z and thus the end of the sound event 4, have the same maximum value 14 for the level, which is just below lies at the asymptotic level 10.

At time T1, the decay behavior characterized by reverberation sets in in the real acoustic situation, which is represented by the audio signal, so that the first level measurement p1 and the second level measurement p2 now also change over to their decay behavior according to the respective decay time.

The adjustment behavior and the decay behavior of the first and the second level measurement p1, p2 are shown here in a schematic linear manner, which also results in a peak in this representation during the transition from the adjustment behavior to the decay behavior. However, the adjustment behavior can increase more slowly before said transition, and the transition can in particular also be smoothed out.

The first level measurement p1 and the second level measurement p2 have a first decay time 16 and a second decay time 18, respectively, with the second decay time 18 being greater than the first decay time 16. The second decay time can be related to the decay time constant T60 as follows , after which a sound level has decreased by 60 dB from a maximum value, and is often used as a measure of decay in a room: the "decay rate" of the second level measurement is given by the difference between the maximum value 14 and the initial value 24, divided by the second decay time 18; In the present case, this decay rate corresponds to a rate of 60 dB/T60. The first decay time 16 can be given here, for example, by half the second decay time 18 (or a similar value), and determines in particular the beginning of the decay behavior 15 of the first level measurement p1 (see dotted line for extrapolation of the first decay time 16).

The first and the second level measurement p1, p2 are each implemented as an asymmetric recursive first-order low-pass filter, so that a corresponding decay parameter of the filter controls the remaining of existing level values for the next time in the level measurement p1, p2 (cf. Equation (i), above), the respective decay time 16, 18 can be adjusted. For the second level measurement p2, the second decay time 18 set in this way is slow enough that the decay behavior due to the linear decline of the filter is given. The acoustic contributions of the diffuse reverberation of the reverberation trail may make contributions to the value of the second level measurement p2, but they do not determine its decay behavior.

For the first level measurement p1, on the other hand, the first decay time 16 is such that after a first peak 15, which corresponds to the maximum value 14 and is given by the sound event 4, and after a rapid fall 17 corresponding to the first decay time 16 (which may still the first and early reflections in the audio signal provide contributions), at a point in time T2 a transition to a flatter flank 19 takes place. In this, the gradually decreasing contributions of the diffuse reverberation "feed" the first level measurement p1, and thus determine its decay behavior in this range.

Since the decay behavior is actually determined by the diffuse reverberation in the flatter edge 19 of the first level measurement p1, which decays in the room with the decay time T60, this exponential decay behavior is parallel to the decay behavior from time T2 in the logarithmic representation of the first level measurement p1 the second level measurement p2 until the first level measurement at a point in time T3 has dropped to the initial value 24 before the sound event 4, the diffuse reverberation in the room in which the audio signal was generated has thus faded away. As a result of the longer second decay time 18, however, the second level measurement p2 only falls to the initial value 24 at a later point in time T4. The output value 24 can be given, for example, by a noise background of the audio signal.

In figure 2 is a schematic time diagram showing a difference Δ from the first level measurement p1 and the second level measurement p2 figure 1 shown (dashed line). As a result of the fact that p1 and p2 are both equal to the initial value 24 up to time T0, and as a result of the identical first and second adjustment times 6, 12 having the same adjustment behavior up to time T1, the difference is Δ=p1 - p2 up to time T1 exactly equal to 0. As a result of the different first and second cooldown 16, 18 (see 1 ) the difference Δ drops to a minimum value min<0 by time T2, and remains there due to the identical decay behavior of the two level measurements p1, p2 between the times T2 and T3 (see 1 , parallel course) at said minimum value min. Only after time T3, from which the first level measurement p1 has already assumed the initial value 24, to which the second level measurement p2 still falls up to time T4 (see 1 ), the absolute value of the difference Δ decreases again, ie the difference Δ increases again to the value zero by time T4.

The difference Δ is now compared with a first limit value th1, a contribution of a diffuse reverberation in the second level measurement p2 being to be determined on the basis of the comparison in a manner to be described below. The amount of the first limit value is less than the amount of the minimum value min, ie | th1 | < | min |, or 0 > th1 > min.

The first limit value th1 is preferably to be selected in such a way that the minimum value min of the difference Δ can be reliably identified by a corresponding comparison. It is therefore assumed that in the case Δ < th1 (i.e. when the difference Δ is below the first limit value th1), there is a contribution 30 of diffuse reverberation in the audio signal, since on the one hand for the period between the relevant times T2 and T3 the The decay behavior of the first level measurement p1 is given directly by said diffuse reverberation, while the decay behavior of the second level measurement p2 is still largely given by the original sound event 4 as a result of the slow second decay time 18, with the contribution 30 of the diffuse reverberation also being included in the second level measurement p2 entrance finds.

A correction function d (solid line) according to equation (iii) is now formed on the basis of the difference Δ=p1−p2 and the first limit value. This correction function d is formed by the first limit value th1 up to a point in time T1'>T1 (with T1'<T2), and changes to the difference Δ for t>T1'. At a point in time T3` > T3 (with T3'< T4), the correction function d is given by the first limit value th1.

The benefit of this correction function d is based on figure 3 clearly: In figure 3 is a schematic time diagram of the second level measurement p2 (solid line) as in figure 1 registered. Furthermore, a reverberation noise level prn (dashed line) is entered, which is determined by the second level measurement p2, from which the correction signal d is derived figure 2 was subtracted is given. Since the first level measurement p1 is determined by the contributions of the diffuse reverberation for the period between the times T2 and T3, and in this period said contributions are also included in the second level measurement p2, but this primarily by the contribution decaying according to the second decay time 18 of sound event 4 (i.e. the "isolated" decaying contribution, without newly added sound contributions), the correction function d between times T2 and T3 (or temporarily already between T1' and T3`) is essentially due to the decaying contribution of the sound event 4 given in the second level measurement p2. In other words, between the times T2 and T3, the actual sound power is essentially the same as the first level measurement p1, so that the (negative) correction function d is added to the second level measurement p2 in order to also arrive at the actual sound power in the reverberation noise level prn in this interval . The second level measurement p2 is always above the actual sound power and thus also above the power of the diffuse reverberation.

The reverberation noise level prn = p2 + d is therefore essentially given by the diffuse reverberation in the said period. Before time T1' and after time T3`, the reverberation noise level prn is given by the constant offset due to the first limit value th1 in the correction function d.

In figure 4 is shown schematically in a time diagram of the reverberation noise level prn (dashed line) and the first level measurement p1 (solid line). In the period of time between the times T2 and T3, in which the first level measurement p1 is determined by the diffuse reverberation, the two lines mentioned are essentially on top of each other. Even if a schematic, idealized representation is shown here, the real case will be similar, namely that both lines in the area of the diffuse reverberation each deliver almost the same values.

If an amplification factor is now determined in a frequency band according to equation (ii) and applied to the signal component of the audio signal 50 in the frequency band, the diffuse reverberation (between times T2 and T3) can be suppressed, while the other contributions of the audio signal 50 in the frequency band are preserved, since p1 > prn then applies there.

That based on figure 1 The method shown for suppressing the acoustic reverberation in the audio signal can be carried out in particular on a frequency band basis. For this purpose, the audio signal is divided into individual frequency bands in which the first and the second level measurement p1, p2 are carried out as shown. The attenuation of the audio signal can then be controlled individually for each frequency band using the amplification parameter G determined in this frequency band as a function of the adjustment behavior and the decay behavior.

In figure 5 a hearing instrument 40 is shown schematically in a block diagram, which has an input converter 42, a signal processing unit 44, and an output converter 46. The input transducer 42, which is given here by a microphone, generated from a sound signal 48 of the environment, in which a specific sound event 4 is received, that audio signal 50. An acoustic reverberation of the sound signal 4 in the audio signal 50 can now in the basis of figure 1 or the one based on figure 2 described manner in the signal processing unit 44 are suppressed. The resulting signal is further processed, in particular subjected to dynamic compression and frequency-band-dependent amplification, and an output signal 52 is generated therefrom, which is converted into an output sound signal 54 by the output converter 46 .

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 derived from this by a person skilled in the art without departing from the scope of protection of the invention.

Reference List

4

sound event

6

first settling time

8th

predetermined proportion

10

asymptotic level

12

second settling time

14

maximum value

15

Peaks

16

first cooldown

17

fall off

18

second cooldown

19

flatter flank

24

noise background

40

hearing aid

42

input converter

44

signal processing unit

46

output converter

48

sound signal

50

audio signal

52

output signal

54

output sound signal

i.e

correction function

at least

minimum value

P

level values

p1

first level measurement

p2

second level measurement

prn

reverberation noise level

t

Time

th1

first limit

T0 - T4

time

T1', T3`

time

e.g

duration

Δ

difference

Claims (15)

Method for suppressing acoustic reverberation in an audio signal (50), - wherein an audio signal (50) is provided, - wherein a first level measurement (p1) of the audio signal (50) is carried out, - a second level measurement (p2) of the audio signal (50) being carried out during the first level measurement (p1), - the first level measurement (p1) being carried out using a first adjustment parameter and a first decay parameter in such a way that the first level measurement (p1) has a first adjustment time (6) and a first decay time (16), - the second level measurement (p1) being carried out by means of a second adjustment parameter and a second decay parameter in such a way that the second level measurement (p2) has a second adjustment time (12) identical to the first adjustment time (6) and a second decay time (18 ) which is greater than the first decay time (6), and - where a difference (Δ) is formed from the first level measurement (p1) and the second level measurement (p2), characterized in that - a reverberation noise level is estimated on the basis of the difference (Δ) and on the basis of the second level measurement (p2), and - An amplification parameter for the audio signal (50) is determined on the basis of the first level measurement (p1) and on the basis of the reverberation noise level (prn). Method according to claim 1,
wherein, in response to a contribution of the sound event (4) in the audio signal (50), acoustic reverberation of the sound event (4) in the audio signal (50) is suppressed by weakening the audio signal (50) as a function of the amplification parameter, in particular by applying the amplification parameter to the audio signal (50). A method according to claim 1 or claim 2,
wherein the first level measurement (p1) and/or the second level measurement (p2) is implemented by a weighted average function. Method according to claim 3,
a weighting factor for a subsequent value of the weighted average function being selected as a function of a rising or falling level. Method according to one of the preceding claims,
wherein the first level measurement (p1) or the second level measurement (p2) is implemented by an asymmetric recursive low-pass filter of preferably first order. Method according to one of the preceding claims, wherein a physical decay time constant of an existing environment is determined, in which a sound level of a sound signal (48) on which the audio signal (50) is based has fallen to a predetermined proportion of an initial value, wherein the second decay parameter is chosen such that the second decay time (18) of the second level measurement (p1) is given by said physical decay time constant for the present environment. Method according to one of the preceding claims,
based on the difference between the first level measurement (p1) and the second level measurement (p2), the presence of a contribution (30) of diffuse reverberation in the audio signal (50) and/or a decaying contribution of the sound event (4) in the second level measurement (p2 ) is detected. Method according to claim 7, wherein the difference between the first level measurement (p1) and the second level measurement (p2) is compared with a specified first limit value (th1), and in the event that the amount of the difference (Δ) from the two level measurements (p1, p2) exceeds the amount of the first limit value (th1), the presence of the contribution (30) of a diffuse reverberation in the audio signal (50) or the decaying contribution of the sound event (4) is determined in the second level measurement (p2). A method according to claim 7 or claim 8,
where in the event that the amount of the difference (Δ) from the two level measurements (p1, p2) exceeds the amount of the first limit value (th1), the difference between the amount of the difference (Δ) and the amount of the limit value (th1) as the decaying contribution of the sound event (4) is determined in the second level measurement (p2). Method according to one of claims 7 to 9,
where a minimum tracker is applied to the difference between the first level measurement (p1) and the second level measurement (p2), and from this the presence of the contribution (30) and/or the contribution (30) of a diffuse reverberation in the audio signal (50) or the presence of the decaying contribution and/or the decaying contribution of the sound event (4) is determined in the second level measurement (p2). A method according to claim 9 or claim 10, a time-dependent correction function (d) being generated on the basis of the decaying contribution of the sound event (4), and the reverberation noise level (prn) being generated by adding the correction function (d) from the second level measurement (p2). Method according to one of the preceding claims,
wherein the amplification parameter is determined using a spectral subtraction, for which in particular a quotient of the reverberation noise level (prn) and the first level measurement (p1) is used. Method according to one of the preceding claims, wherein the audio signal (50) is broken down into a plurality of frequency bands, with the first level measurement (p1) and the second level measurement (p2) being carried out in each case by frequency band, the respective amplification parameter being determined for a plurality of frequency bands and being applied to the signal component of the audio signal (50) in the respective frequency band in order to suppress the acoustic reverberation. Method for suppressing an acoustic reverberation in an audio signal (50) of a hearing instrument (40),
wherein the audio signal (50) is provided from an ambient sound signal (48) using an input transducer (42) of the hearing instrument (40), and wherein acoustic reverberation in the audio signal (50) is suppressed by a method according to one of the preceding claims. Hearing instrument (40) with an input converter (42) for generating an audio signal (50) and a signal processing unit (44) which is set up for carrying out the method according to one of the preceding claims.

Images