WO2023280752A1

WO2023280752A1 - Computer-assisted method for stable processing of an audio signal using an adapted lms algorithm

Info

Publication number: WO2023280752A1
Application number: PCT/EP2022/068392
Authority: WO
Inventors: Ludwig KOLLENZ
Original assignee: Austrian Audio Gmbh
Priority date: 2021-07-05
Filing date: 2022-07-04
Publication date: 2023-01-12
Also published as: EP4367900A1; EP4117306A1

Abstract

The invention relates to a computer-assisted method for stable processing of an input audio signal of an acoustic system comprising: at least one audio signal input; at least one audio signal output; a data connection between the audio signal input and the audio signal output, which connection is influenced by an adaptive filter; and an LMS algorithm which controls the adaptive filter; the controlling algorithm being a novel combination of a classic LMS algorithm and a PEAK filter that ensures stable processing of the input audio signal of the acoustic system.

Description

Computer-aided method for stable processing of an audio signal using an adapted LMS algorithm

The invention relates to a computer-aided method for processing an audio signal using an adapted LMS algorithm according to the preamble of claim 1.

LMS algorithms, the collective term for algorithms in which the squares of error are minimized (least mean squares) have long been used in electroacoustics, in particular for adaptive filters in numerous variants and for the most diverse areas of application. The currently most used variants are known by the following abbreviations: LMS, NLMS, sign-LMS, LMS with variable increment (VSS-LMS), LMS with correlation factor or NLMS and for special applications variants such as filtered-x LMS or filtered-e LMS ( see International Journal of Electrical and Computer Engineering Vol 7, No 5, October 2017, LMS Adaptive Filters for Noise Cancellation: A Review, pp 2520ff).

The application of LMS algorithms is known from the prior art, for example in the case of feedback suppression. In particular, sign-LMS can be used in this context.

US 2008/0063230 shows the use of a classic LMS algorithm in conjunction with an Adaptive Digital Filter (ADF). The ADF is controlled here by the LMS algorithm.

Another example of use is US Pat. No. 9,807,503, in which an equalizer is used to optimize the reproduction of a digital acoustic source (e.g. Bluetooth audio). The equalizer is controlled by an adaptive filter whose filter coefficients are obtained, for example, by using LMS and then smoothed using linear interpolation and/or the exponential method. WOOO 19605 shows a standard implementation of an adaptive feedback canceller using an NLMS algorithm.

US 2017/0201276 shows an LMS application (standard system estimation) in which the input signals of the LMS algorithm are delayed differently.

Furthermore, the NPL document "Design the adaptive noise canceller based on an improved LMS algorithm and realize it by DSP" by Xu Yanhong and Zhang Ze shows the possibility of using an improved sign-LMS as an adaptive system estimator.

A disadvantage of all solutions is the potential instability of the system. Exponential smoothing uses the same coefficient for the attack and release cases. Due to this property, the filter coefficients of adaptive filters can converge towards infinitely high values under certain unfavorable conditions. Such overshooting of the filter coefficients is prevented by the word length of the processor used. This property is called clipping. A typical case of such clipping is acoustic feedback suppression: if feedback occurs (the microphone records the loudspeaker), this can be heard as a very loud sine wave (“howling”). In this case, an adaptive filter is configured to filter the speaker output and subtract at the microphone. In the best case, it filters the feedback sine exactly and subtracts it from the microphone, ergo the sine is eliminated. However, if the system is confronted with a sinusoidal tone from an external source (e.g. music), the adaptive filter tries to eliminate it. Since it does not get weaker despite subtraction (the source is no longer the loudspeaker but external), the filter coefficients swell towards infinity. The result is a distorted sound. In order to avoid this erroneous convergence, it is known in the prior art to monitor the coefficients using an external algorithm and to limit them at a specific threshold value, which is useful in terms of the result, but entails greater effort in terms of complexity and computing power. A problem occurs, for example, with hearing systems located in the ear, such as in-ear headphones or hearing aids, when the device is removed from the ear. By eliminating passive damping, a method that uses a classic LMS algorithm tries to increase the gain of the ANC filter to infinity, since there is no longer any ANC power. Thus, additional sensors must be used to detect whether the device is in the ear or not.

There is thus a need for an LMS algorithm that requires less acoustic system complexity and computational power requirements and is not only usable in ANC systems, but can be used anywhere adaptive filters are used.

According to the invention, these problems are solved by using a new PEAK-LMS algorithm, which has the features specified in the characterizing part of claim 1, in other words, in which a signum function is calculated from an input signal sequence and a filter output sequence, which together with a sequence of initial filter values and a sequence of time constants for the attack and the release case, which do not have to be the same, are used to calculate a sequence of adapted filter values. In one embodiment, upper and lower limits are set for the coefficients in the method/algorithm itself, so that the risk of upward convergence is intrinsically excluded, which also eliminates the need for complex, external monitoring.

The invention is illustrated below by examples. Here, Fig. 1 shows the schematic structure of a feedback suppression of an acoustic system, Fig. 2 shows stable values for the coefficients of the filter, which were obtained according to the method described in Fig. 1 using the PEAK-LMS algorithm, Fig. 3 filter coefficients of an LMS algorithm for a system with sinusoidal outer tones, 4 filter coefficients of a PEAK-LMS algorithm for a system with sinusoidal external tones, FIG. 5 the application of the invention to the control of an amplifier, FIG known form is used, Fig. 7 the unit step response of the filter for the attack case, Fig. 8 the unit step response of the filter for the attack and the release case, Figs. 9a and 9b in two further variants the adaptation speed and the

Residual error of the PEAK filter as used by the method according to the invention, FIG. 10 the convergence of a filter coefficient of an advantageous embodiment of the invention, in which this method is also applied to the adaptation speed, FIG 12 shows an example of an echo suppression application, FIG. 13 shows an example of an adaptive ANC application, and FIG. 14 shows a variant of an adaptive ANC application

beamforming, FIG. 15 by way of example a variant of the application of an adaptive equalizer and FIG. 16 by way of example a second variant of the application of an adaptive equalizer.

1 shows the schematic structure of feedback suppression of an acoustic system, as is known from the prior art, but can also be used for the method according to the invention. The microphone 1 records the output of the driver (loudspeaker 2), resulting in a so-called acoustic feedback 5 in the form of the feedback path h(n). In the unfiltered case, this can be audible as a very loud sine wave. The function block for adapting the filter 4 (adaptation block for short) calculates filter coefficients for an adaptive filter 3 from the reference signal d(n) (reproduction) and the error signal e(n) (result of the subtraction), which are then used as the filter output y(n ) are issued. The adaptation block 4 thus attempts the transfer function of the feedback path 5 estimate and apply by means of adaptive filter 3, estimate. In the example shown, the filtered output of the driver is subtracted at the circuit's microphone input (downstream of the physical microphone). If the transfer function is correctly estimated, the filter “clips” the sine and subtracts it from itself at the input, thereby suppressing feedback. LMS, SignLMS and PEAK-LMS are entered in adaptation block 4 in FIG. These represent variants for the adaptation algorithm. These three variants are compared below under identical conditions:

In a simulation, white noise (duration 1 second, sampling rate 48kHz) is fed to the microphone. The same learning rate is always used for the three algorithms. For PEAK-LMS, the individual filter coefficients are limited by the upper limit Δ _pos = 1 and the lower limit Δ _neg = -1. For this use case, the attack time is equal to the release time. The time constants for the attack and release times must be selected as a function of the sampling rate in a manner known to those skilled in the art from the prior art. For the sampling rate of 48kHz given here, for example, it would be 300ms.

Learning rate is one factor by which adaptation occurs. LMS is a gradient method, where the gradient is based on the least squares error. The learning rate thus defines an increment by which the coefficients of the LMS algorithm are adapted. For a learning rate of 0.00005, the LMS algorithm does not produce a valid result for the filter coefficients because the adaptation is too slow for a learning rate of 0.00005. A Sign-LMS algorithm shows a similar behavior and becomes unstable for a too high learning rate, while it adapts too slowly for a too low learning rate. For this example, a learning rate of 0.00001 is close to the optimum for an LMS or Sign-LMS algorithm. The PEAK-LMS algorithm shows a significantly higher tolerance: for 0.00005 the suppression performance is reduced, but the algorithm remains with stable values for the coefficients.

FIG. 2 shows values for coefficients for a filter function c(n), which are calculated according to the method described in FIG. 1 using the PEAK-LMS algorithm at a learning rate of 0.00005 were obtained for a FIR filter. It can be clearly seen that the values of the coefficients remain stable. The abscissa shows the number of the filter coefficient and the ordinate shows the value of the coefficient.

3 shows an example of the filter coefficients of a prior art LMS algorithm for a system with sinusoidal external tones. The abscissa shows the number of the filter coefficient and the ordinate shows the value of the coefficient. A typical problem for feedback suppression is the case of an "external" sinusoidal tone, i.e. one not brought into the system by feedback. In this case, the adaptive filter recognizes the sine tones and wants to suppress them, since they are mistakenly regarded as feedback. For a sound whose source is not feedback, but e.g. music, the problem arises with the adaptive filter that the supposed feedback does not become weaker due to the subtraction at the microphone and the filter gets into an incorrect setting (this effect is called "entrainment"). For an LMS algorithm, this means that the filter coefficients tend towards infinity, since the gradient - due to the constant error - always remains the same. The strong increase in the filter coefficients can be seen particularly clearly in the case of the first two coefficients. The number range -4000 to +2000, in which the coefficients are located, can no longer be mapped in processors and the result of the filtering cannot be played under any circumstances, since digital-to-analog converters (DA converters) in acoustics are limited to a value range ± 1 are restricted. In the example above, the filter coefficients reach ±oo after a short time and the simulation stops because the coefficients can no longer be calculated.

Fig. 4 shows an example of the filter coefficients for a filter function c(n), determined using the PEAK-LMS algorithm according to the invention, for a system with sinusoidal outer tones, identical to that in Fig. 3. The abscissa shows the number of the filter coefficient and the ordinate the value of the coefficient. By limiting the filter coefficients to λ _pos =1 and X _neg =−1, the instabilities occurring in a classic LMS algorithm can be avoided (see FIG. 3). In addition to the advantage of the general stability of the filter, the range of values ± 1 is, for example, in conventional fixed-point processors for audio (eg 24-bit) without difficulties can be represented. While the invention is not limited to fixed-point processors, these are widely used in audio and video applications. In the example shown, PEAK-LMS prevents the coefficients from rising too far. In the extreme case, the PEAK-LMS algorithm would limit the filter coefficients to ±1, but in the case shown it generally prevents an incorrect setting. It is also immediately apparent that a method according to the invention that is based on a PEAK LMS algorithm can work more efficiently than one that works, for example, with a classic NLMS algorithm, resulting in savings in the required computing power that either for others processes can be used, or can contribute to an increase in battery life.

As an application example, FIG. 5 shows the regulation of an amplifier based on the estimation of the gain of a transfer function. In the case of an in-ear ANC (Active Noise Canceling) hearing system, the gain of a transfer function must be controlled in real time. In this context, real-time means with the lowest possible latency and time-invariant, i.e. the processing cannot take place independently of the temporal signal. The processing of an audio file that has already been recorded would therefore not be real-time, since a calculation is theoretically not subject to any time pressure. For ANC, on the other hand, incoming microphone data must be processed as soon as possible. Depending on the wearing situation (insertion depth; individual shape of the ear canal), the amplification of the primary path (outer microphone) and/or the amplification of the secondary path (inner microphone) must be adapted in order to optimize the performance of the ANC system.

In this example, too, conventional LMS variants will always try to converge towards a theoretical optimum. A problem arises, for example, when the device is removed from the ear: there is no longer any passive damping, which means that the system becomes acoustically open. In this case, the LMS algorithm would try to increase the gain of the ANC filter to infinity, since there is no longer any ANC performance. Thus, additional sensors must be used to detect whether the device is in the ear or not. Due to the automatic limitation of the PEAK-LMS algorithm at Ä _pos = 1 and l he _eg = — 1 clipping and the resulting artefacts on the driver (loudspeaker) are avoided.

The method according to the invention, which is carried out using the PEAK-LMS algorithm, can be used not only in ANC systems, but wherever adaptive filters are used, for example in echo suppression, feedback suppression, system estimation ( Recognition of an unknown transfer function), channel equalization (also applies to HF technology), adaptive inverse control, hum suppression (50Hz hum of the power supply filter for e.g. ECG sensors), vector voltmeter, separation of signals with different correlations, interference suppression in audiological measurement systems ( e.g. measurement of otoacoustic emissions) and many others.

The precise design of the PEAK-LMS algorithm for use in the method according to the invention is explained below.

Figure 6 shows a block diagram of an example using an LMS algorithm in its known form. x(n) describes the input signal, h(n) describes the transfer function to be estimated (transfer function of the acoustic system), y(n) describes the output of the filter, c(n) describes the coefficients of the adaptive filter, d(ri) describes the reference signal for the LMS algorithm and e(n) the error signal which follows from the difference between the value y obtained and the predetermined value x, in each case in the nth step. Furthermore, FIG. 6 shows the LMS block that determines the necessary adaptations of the filter c(n).

The LMS algorithm in its known form reads (formula I): e(n) = x(n) — y(n) c(n + 1) = c(n) + m * e(n ) * x(n )

The vectors are printed in bold, as is usual in this context, and the scalars are normal. For example, x(n) is the element of vector x(n) at time n. Specifically, c(n+1) stands for the result of the nth adaptation of c, c(n) for the value of c obtained in the nlth step; m (also mu) for the adaptation rate; e(n) for the error signal in step n, x(n) for the input signal sequence representing the input data of the audio signal up to time n and y(n) for the output of the filter in the nth step.

As already explained, the basic form from the prior art shown in FIG. 6 is sensitive to level fluctuations of the input signal x(n) in that it has a negative effect on the convergence of the algorithm. As already explained, various adaptations of the LMS method have therefore been proposed which, under the designations NLMS and sign-LMS, are part of the prior art and reduce this sensitivity. However, NLMS is unfavorable on fixed-point processors because it requires division for normalization. Divisions mean increased computational effort and are subject to quantization effects on fixed-point systems. Sign-LMS, on the other hand, does not require any division, but does so at the expense of convergence speed and the occurrence of a larger residual error of the adaptation, which can lead to poorer filter properties.

For a sign-LMS algorithm, the formulas II derived from formulas I are: e(n) = x(n) — y(n) c(n + 1) = c(n) + m * sgn(e(n ) * x(n ))

The meaning of the vectors and scalars is analogous to the case of the classical LMS algorithm (see above). As can be seen directly, a small step ( μ * sgn(e(n) * x(n)) ) is added or subtracted to a coefficient. If the setting is incorrect (as explained above with reference to feedback suppression, filter coefficients can converge to infinity under unfavorable conditions), this change can theoretically occur to infinity, which causes problems in a finite system. This problem can be avoided by the so-called PEAK algorithm, which is shown in the following formulas III:

The time constants a and ß stand for the attack and release case and thus determine the rise and fall time of the algorithm, a and ß are purely numerical, their temporal significance depends on the selected sampling rate. Both move in the interval [0; 1] from R

A filter using a sign-LMS algorithm with a PEAK filter is a one-pole, recursive filter per coefficient, which has an attack and a release time constant. If the input signal is greater than the filter output, the attack time constant is applied, otherwise the release time constant (see formulas III).

Figure 7 shows the unit step response for a single coefficient approximated according to the method of the invention (value towards which the filter tends for an input signal which jumps from 0 to 1 and then remains constant) of the filter. The influence of the attack time constant can also be seen here. The abscissa shows the approximation steps based on their discrete time axis and the ordinate shows the value of the filter coefficient.

By combining the sign-LMS with the PEAK filter, each individual coefficient is no longer calculated by adding a step size, but via the step response of the PEAK filter. This results in the formulas IV:

where w(n) is a signum function that represents the summary of the signum term of formulas II. The inputs to the algorithm e(n) and x(n) could be filtered before being fed to the algorithm. This has the advantage that the adaptation can be focused on a frequency range, usually with the help of an appropriate pre-filter (typically a high, low, bandpass filter, or a combination of these). The desired frequency range depends on the application and can be defined individually by a specialist according to the requirements. Due to the exponential function of the PEAK filter, the LMS method can no longer converge to infinity, even if, for example, the increment (adaptation rate) m is constantly added for a coefficient due to an incorrect setting (see feedback cancellation example above).

Figure 8 shows the unit step response of a filter for a single coefficient, approximated according to the method of the invention using the PEAK-LMS algorithm, for the attack and release cases. The abscissa shows the approximation steps based on their discrete time axis and the ordinate shows the value of the filter coefficient. The attack and release time constants in Formulas IV were equated. In this case, C(a,ß ) = a = ß applies. This means that the algorithm (example rectangular pulse) converges equally fast in both directions.

However, it is also possible to use different values for the attack and release time constants (a Y ß) for the PEAK-LMS algorithm, with which different convergence times can be obtained for positive and negative coefficients. An adaptation of formulas IV leads to formulas V:

These and other modifications open up possibilities and areas of application that are not accessible with the various LMS variants of the prior art. An example of this is the introduction of the filter coefficient l. The signum operator returns 1, or 0, or -1; by introducing the factor l, the limit of the coefficient can be controlled, as shown by formulas VI, which are an extension of formulas V:

The factor l is fundamentally freely definable and application-related. However, the interval [0,1] makes sense for audio phrases, but [-1 ; 1] make sense.

With this it is also possible to set different limits for the coefficients in the negative and positive area and also own limits for each individual coefficient. An example of this is Formulas VII, which are an extension of Formulas V:

L+ have values that are firmly defined in advance, which can be adapted by the specialist to the desired application.

By using different attack and release time constants, the adaptation of the filter can be handled extremely flexibly, while the advantages of the sign-LMS over other variants, as well as the NLMS, continue to be retained, since no divisions are required during the calculation process, which is advantageous when converting to a fixed-point DSP.

An example implementation of the above is adaptive gain control. An amplification should typically regulate adaptively in the range of ± 6 dBFS. By setting an upper limit for X _up = 2 and a lower limit for X _down = 0.5, the PEAK-LMS algorithm automatically regulates ± 6 dBFS without the need for additional calculations such as converting linear values to decibels and vice versa. The elegance of the solution is thus evident.

The interval ±6 dBFS here is application-related for an adaptive gain control in an ANC earphone. Other areas of application include feedback suppression, ANC, echo cancellation, adaptive beamforming, adaptive gain control and similar applications. The space of the gain control depends on the respective product and the calculated ANC filters. In any case, the sensible interval range is ± 10 dBFS. The volume of the pressure chamber created by the auditory canal and the in-ear headphones varies depending on the wearing situation, e.g. This in turn must be taken into account and compensated for by an adaptive ANC system.

9 shows the adaptation speed and the residual error of the PEAK filter as used by the method according to the invention. The abscissa shows the discrete time axis and the ordinate shows the value of the error size. As already briefly mentioned, the sign-LMS has a residual error of convergence, the size of which depends on the step size m: A larger m leads to fast convergence, but at the same time to a larger residual error (the algorithm often "oscillates" around an optimum without it to reach); on the other hand, a smaller m leads to slower convergence with smaller residual error. The variant used according to the invention shows a similar behavior. The variant in FIG. 9a shows a PEAK filter with a time constant C(α,β) of 0.001 and the variant in FIG. 9b shows a PEAK filter with a time constant C(α,β) of 0.0025. In LMS algorithms, it is generally known that the result can be improved by choosing the step size adaptively, depending on the size of the error. This is commonly known as variable step-size LMS. Formulas VIII can be specified by adapting formulas I:

An upper limit and a lower limit can again be introduced for m. The step size is chosen large when the error is large and small when the error is small, thus achieving a compromise between convergence speed and residual error, a and b are coefficients that ensure an adaptive step size and a weighting between m(h) and e ² (n) allow. For example, the ratio a = 1 - b is chosen. Typical values here would be 1 > a > 0.8. 10 shows the adaptation speed and the residual error of the PEAK filter as used in an advantageous embodiment of the method according to the invention. The abscissa shows the discrete time axis and the ordinate shows the value of the error size. In the case of the same attack and release time constants, formulas IX deliver:

Formula IX is a simplified case of formula VII for the case C _a ß (n) = a = ß. In other words, it is a case where α and β are the same but can be time-variant controlled. For the sake of simplicity, the formulas have therefore been set in relation to time. A significant advantage of this variant is the lower demand for computing capacity. The improved convergence of Figure 10 over Figures 9a and b is evident from the exponential given convergence behavior. If, in the embodiment of the invention presented, different time constants are also used for this method, the formula X is obtained as an extension of formula IX:

In this case there are no predetermined attack/release times, since the formula in the bottom row recalculates the time constants in each iteration. The invention is not limited to the example and its variants, but can be used wherever LMS algorithms are used and there is a risk of instability due to the external conditions or a rapid reaction is desired. The two terms “step response” and “step response” are used synonymously within the scope of this application.

In the formulas given, the algorithm was always defined in the time domain. A person skilled in the art can easily define this using appropriate methods, for example the Fourier transformation, with the same result over the frequency range.

In summary, it is a computer-aided method for the stable processing of an input audio signal of an acoustic system, comprising at least one audio signal input (e.g. a microphone, an external device for playing music, input of a digital interface (USB, Bluetooth,...), reading a stored audio file or any other source of an audio signal), at least one audio signal output (e.g. a loudspeaker, a recording device, output of a digital interface (USB, Bluetooth,..), writing a saved audio file or any other sink for an audio signal), one through a adaptive filter (3) affected data connection between the audio signal input and the audio signal output and an LMS algorithm (4) controlling the adaptive filter (3), wherein a) the LMS algorithm (4) on the basis of a reference signal d(n) and an error signal e(n) a transfer function h(ri) (this is the target value of the LMS algorithm, which can be approximated as best as possible but can never be fully reached), b) the estimate of the transfer function is applied as an adaptive filter (3) and there as one initial filter function c(n) is available (the estimation is thus transferred to the filter block and is available there as a mathematical filter function c(n) in the form of filter coefficients), c) the output of a filter output sequence as y(ri) (this is about a sequence of samples that should be as close as possible to the output sequence of h(ri)), d) a comparison of the output of the transfer function h(ri) and the filter output sequence y(ri) of the initial filter function n c(n), whereby the error signal e(n) results from the difference, e) characterized in that the LMS algorithm used to adapt the adaptive filter (3) is a PEAK-LMS algorithm, f) for the algorithm uses a signum function w(n) that over i. an input signal sequence x(n) representing the input data of the input audio signal (this is the actually desired audio input signal, h(ri) in contrast to this is an undesirable environmental influence such as feedback, echo, noise, or similar, which in the application x(n) undesirably changed), ii. the filter output sequence y(n) representing the output of the filter, iii. that defines the difference between an input signal sequence x(n) changed by undesired influences (e.g. environmental influences feedback, echo, noise, etc.) and the filter output sequence y(n) by e(n) = x(n) - y(n). , error signal e(n) (the task of the PEAK-LMS algorithm is to minimize this error signal), iv. is calculated in the form w(n ) = sign(e(n ) * x(n )), g) which together with the initial filter function c(n) and time constants C(a,ß ), where a is the time constant of the attack - and ß describes the time constants of the release case, can be used to calculate the sequence of adapted filter values c(n + 1) using the equation c(n + 1) = c(n) + C(a,ß ) * (w( n) - c(n)) to be calculated.

The invention is therefore a computer-assisted method for the stable processing of an input audio signal of an acoustic system, comprising at least one audio signal input, at least one audio signal output, a data connection between the audio signal input and the audio signal output which is influenced by an adaptive filter, and an LMS system controlling the adaptive filter. Algorithm, where the controlling algorithm is a novel combination of a classic LMS algorithm and a PEAK filter, which ensures the stable processing of the input audio signal of the acoustic system.

For better processing, the input signal sequence x(n) and the error signal e(n) can be subjected to a pre-filtering before being fed to the algorithm. High, low or bandpass filters and combinations thereof are common here.

In the following, application examples for acoustic systems are shown for a better understanding, in which the method according to the invention with PEAK-LMS- algorithm can be applied. For the sake of simplicity, ADC (analog-to-digital converter) and DAC (digital-to-analog converter) blocks are omitted in graphics, and acoustic paths are considered discrete H(z) instead of analog H(s), which in the previous figures correspond to h( ri) . Reference signal d(n) , error signal e(n), filter output sequence y(n) and input signal sequence x(n) are shown without the sample index (n) for the sake of simplicity. Furthermore, H(z) is used instead of the filter function c(n) to make it clear that the goal is either to form a complementary function or to estimate the objective function.

Fig. 11 shows the general case of feedback suppression ("Feedback Cancellation"; applicable to all types of headphones, hearing aids, ANC systems, etc.) analogous to Fig. E A sound source is additively mixed with the sound of loudspeaker 2 and Microphone 1 of the system recorded. The task of the PEAK-LMS algorithm is to estimate the acoustic feedback path H(z), to filter the loudspeaker signal accordingly and to subtract it from the microphone signal. The error signal e (which is to be minimized) is the result of the subtraction and the reference d is the result for the output (usually it is taken directly from the DAC buffer). M(z) denotes the transfer function of the system (e.g. equalizer, multi-band compression, etc.). In the ideal case, H(z) is optimally approximated in H(z) (H(z) = H(z)), i.e. the sound of loudspeaker 2, which is mixed in front of microphone 1, is subtracted behind the ADC. This creates an infinite cancellation for the feedback sine. An example of a suitable processor is: "Onsemi Ezairo 7100" for hearing aids. This has a block floating point in an auxiliary processor.

12 shows the application of echo cancellation. Similar to feedback cancellation, a PEAK-LMS algorithm has to estimate the acoustic path of the echo (H(z)). This acoustic path mixes with the speaker in front of microphone 1 and must be subtracted behind it. In contrast to feedback suppression, the microphone here does not play to loudspeaker 2, but to a receiving sink (examples: radio microphones and monitoring boxes, headsets in the field of mobile communications and intercom). The labels are analogous to Fig. 11. 13 shows the application of an adaptive ANC. In adaptive ANC, two paths can be estimated: feedforward and feedback. Depending on the system, there is only one of the two paths or both. H(z) denotes the passive damping of the system and a PEAK-LMS algorithm has to estimate this path from feedforward microphone 6 and feedback microphone 7 in H(z). If this path is output inverted at loudspeaker 2, H(z) and H(z) cancel out. Another path is /(z) - this is the acoustics inside the ear cup of the ANC listener. Again, a PEAK-LMS algorithm must estimate /(z) in /(z) to achieve cancellation. /(z) in /(z) are analogous to H (z) and H(z) and are named differently because they are different paths. An example of a suitable processor is: "AD AU 1860". This is a fixed-point DSP with no floating-point capability.

14 shows a variant of the application of an adaptive beamforming. Two microphones 1 are operated here as an endfire array, ie one signal is delayed in relation to the other and subtracted (the delay is represented by the block z ^~n ). This is done for two paths, with the effect of producing two opposite cardioid polar patterns as the polar pattern. A PEAK-LMS algorithm now estimates the transfer function between one cardioid and the other, filters the former and subtracts its signal from the latter. The result is an adaptive beamforming where the nulls of the final directivity are controlled by the adaptive filter. The speaker is not shown in this simplified schematic. An example of a suitable processor is: "Onsemi BelaSigna 300". It is a fixed-point processor with a block-floating-point unit.

15 shows a variant of the application of an adaptive equalizer. In this application, the block of a PEAK-LMS algorithm is arranged to remove noise (noise disturbance, corresponding to H(z )) from a signal. If a useful signal (source) is disturbed by additive noise, the PEAK-LMS algorithm controls a filter in the channel and equalizes the channel according to the difference in the source and the disturbed transmission. The speaker is not shown in this simplified schematic. The block z ^~n represents a delay element.

16 shows the channel equalization as a second variant of the application of an adaptive equalizer. In this form, a filter /(z) is linked to a transfer function H(z) via convolution (in the time domain). The reference R(z) precedes the block of the PEAK-LMS algorithm (which is also made clear by the channel designation d), so the adaptation will equalize the channel in the sense of H(z ) * /(z) = Ä(z ).

As can be seen from the exemplary, non-exhaustive list of use cases, the possible area of use is wide. On the basis of the descriptions, a person skilled in the art is easily able to adapt the method according to the invention to his own tasks.

Claims

1. Computer-assisted method for processing an input audio signal of an acoustic system, comprising at least one audio signal input, at least one audio signal output, a data connection between the audio signal input and the audio signal output influenced by an adaptive filter (3) and an LMS algorithm (3) controlling the adaptive filter (3). 4), where a) the LMS algorithm (4) estimates a transfer function h(ri) on the basis of a reference signal d(ri) and an error signal e(n), b) the estimation of the transfer function is applied as an adaptive filter (3). and is present there as an initial filter function c(n), c) a filter output sequence is output as y(ri), d) a comparison of the output of the transfer function h(ri) and the filter output sequence y(ri) of the initial filter function c(n ) takes place, whereby the error signal e(n) results from the difference, e) characterized in that the use for the adaptation of the adaptive filter (3). th LMS algorithm is a PEAK-LMS algorithm, where f) a signum function w(n) is used for the algorithm, which over i. one representing the input data of the input audio signal,

input signal sequence x(n), ii. the filter output sequence y(n) representing the output of the filter, iii. that, the difference between the input signal sequence x(n) and the

Filter output sequence y(n) defined by e(n) = x(n) — y(n),

error signal e(n), iv. is calculated in the form w(n) = sign(e(n ) * x(ri)), g) which together with the initial filter function c(n) and time constants C(a,ß ), where a is the time constant of the attack - and ß describes the time constants of the release case, can be used to calculate the sequence of adapted filter values c(n+1) using the equation to calculate c(n + 1) = c(n) + C(a,ß ) * (w(n) - c(n)).

2. Computer-aided method for stable processing of an input audio signal of an acoustic system according to claim 1, characterized in that the time constant a of the attack and the time constant ß of the release case are the same.

3. Computer-assisted method for stable processing of an input audio signal of an acoustic system according to claim 1 or 2, characterized in that the input signal sequence x(n) is subjected to a pre-filtering before it is fed to the algorithm.

4. Computer-assisted method for stable processing of an input audio signal of an acoustic system according to any one of claims 1 to 3, characterized in that the error signal e(n) is subjected to a pre-filtering before it is fed to the algorithm.

5. Computer-supported method for stable processing of an input audio signal of an acoustic system according to claim 3 or 4, characterized in that high, low or bandpass filters and combinations thereof are used for pre-filtering.

6. Computer-assisted method for stable processing of an input audio signal of an acoustic system according to one of claims 1 to 5, characterized in that the signum function w(n) is expanded by a filter coefficient l, so that the following applies: w(n) = l * sgn( e(n) * x(n)).

7. Computer-aided method for stable processing of an input audio signal of an acoustic system according to claim 6, characterized in that the filter coefficient 1 lies in an interval [0,1].

8. Computer-aided method for stable processing of an input audio signal of an acoustic system according to claim 6 or 7, characterized in that the filter coefficient l depends on the sign of the sign function w(n) according to

can be different.

9. Computer-assisted method for stable processing of an input audio signal of an acoustic system according to one of claims 1 to 8, characterized in that the step size m of the PEAK-LMS algorithm is adaptive depending on the error size according to the equation system

is chosen, where a and b are coefficients which ensure an adaptive step size and allow a weighting between m(h) and e ² (n) and are linked via a = 1 - b.

10. Computer-aided method for stable processing of an input audio signal of an acoustic system according to any one of claims 1 to 9, characterized in that an adjustment of the adaptation speed is applied depending on the error size, according to the formulas

11. Computer-aided method for stable processing of an input audio signal of an acoustic system according to claim 10, characterized in that the time constant a of the attack and the time constant ß of the release case are different.

12. Computer-aided method for stable processing of an input audio signal of an acoustic system according to any one of claims 1 to 9, characterized in that the algorithm is applied to the frequency range.