EP0930801B1 - Circuit and method for adaptive suppression of acoustic feedback - Google Patents

Description

The present invention relates to a circuit and a method for the adaptive suppression of an acoustic feedback according to the preambles of the independent claims. It is used, for example, in digital hearing aids.

In acoustic systems with a microphone, a loudspeaker or listener and an intermediate electronic signal processing part, acoustic feedback can occur between the loudspeaker or handset on the one hand and the microphone on the other hand. The acoustic feedback causes unwanted distortions and leads in extreme cases to unstable behavior of the system, for example. An unpleasant whistling. Since the unstable operation is not acceptable, the signal gain of the signal processing part often has to be set smaller than effectively desired.

The suppression of the acoustic feedback in digital hearing aids can basically be addressed with different approaches. The best results are currently achieved with the adaptive filtering method.

Various systems with adaptive filtering are known. Basically, in such systems, an acoustic input signal is picked up and converted into a digital electrical signal. From this, an echo estimate is subtracted. The echo-compensated signal is transformed with a necessary hearing correction into a digital output signal, converted into an analog electrical signal and emitted as an acoustic output signal. The acoustic signal is deformed on its way back to the microphone according to a feedback characteristic and superimposed on an externally incident acoustic signal to a new acoustic input signal. To calculate the echo estimate, the fixed delays contained in the system are modeled and the unknown feedback characteristic is modeled.

Unfortunately, such well-known systems with adaptive filtering are not enough to achieve a distortion-free transmission in a realistic environment while at the same time satisfying convergence behavior. The difficulties stem from the fact that real signals such as speech or music have a non-negligible autocorrelation function. The adaptive filter effectively interprets the autocorrelation of the signal as a feedback effect and partial cancellation of the desired signal is the consequence. This effect is most extreme for purely periodic signals (eg alarm sounds). The system can be improved if the feedback characteristic is modeled using decorrelated signals. There are different approaches to this, which are explained below.

A first approach involves the use of an artificial noise signal. Such a system is, for example, from the European patent applications EP-415 677 . EP-634 084 and EP-671 114 known by the company GN Danavox AS. The common feature of such systems is the use of an artificial noise signal to decorrelate the signals. The noise signal is switched on only when needed instead of the output signal or continuously to Added output signal. The disadvantage of these systems is the effort required for the control of the noise signal power such that the noise remains as inaudible and still a sufficiently good convergence speed can be achieved.

A second approach involves the use of fixed orthogonal transformations. Such a system from Phonak AG was, for example, as a European patent application EP-585 976 released. The common feature of such systems is the use of fixed orthogonal transforms to decorrelate the signals. The filtering and updating of the coefficients in these systems does not take place directly in the time domain. The disadvantage of these systems, in addition to the generally greater computational effort, is the additional delay in the signal processing path caused by the block-by-block processing.

A third approach involves the use of adaptive decorrelation filters. Such a system was, for example, in Mamadou Mboup et al., "Coupled Adaptive Prediction and System Identification: A Statistical Model and Transient Analysis", Proc. 1992 IEEE ICASSP, 4; 1-4, 1992 , described. The feasible with this approach systems differ by the different arrangement and implementation of the decorrelation filter. The disadvantage of the published system is the use of relatively slow transversal filter decorrelators, which due to their structure can not adapt very fast to the changing statistical properties of their input signals. The coefficients of the two decorrelation filters are generally determined by decorrelation of the output signal reaching the loudspeaker or receiver. This is intended to make the convergence speed frequency-independent. A special weighting of the frequencies which are particularly critical for the feedback behavior with high gains in the signal processing path is thus not present.

It is an object of the invention to provide a circuit and a method for the adaptive suppression of an acoustic feedback, which do not have the disadvantages of the known systems. In particular, an optimal convergence behavior with minimal, inaudible distortions and without additional signal delay should be achieved with the least possible effort.

The object is achieved by the circuit and the method as defined in the independent claims.

The present invention belongs to the group of systems with adaptive decorrelation filters. It makes use of the knowledge that cross-member filter structures are particularly suitable for rapid decorrelation. Such cross-member filter structures are known from speech signal processing and are used there for linear prediction. Algorithms for the decorrelation of a signal by cross-link filter are known and can be taken from the literature, for example S. Thomas Alexander, "Adaptive Signal Processing", Springer-Verlag New York, 1986 ,

The present invention models the feedback path and adaptively follows its temporal changes by means of optimized tracking. The fed back signal components are constantly removed from the input signal. This significantly increases the signal gain allowed for stable operation. This allows the use of higher reinforcements (eg in the case of severe hearing damage) or a pleasant, more open supply (eg in case of slight hearing damage).

The inventive circuit is used in an acoustic system with at least one microphone for generating an electrical input signal, at least one speaker or handset and an intermediate electronic signal processing part. It contains a filter for Modeling a feedback characteristic, an updating unit for calculating current coefficients for the filter, a subtractor for calculating an echo-compensated input signal by subtracting an echo estimate from a digital input signal supplied by the filter, a delay element for calculating a delayed output signal, and two adaptive cross-link decorrelation filters. A first cross-sectional decorrelation filter is arranged to decorrelate the echo-canceled input signal, and a second cross-sectional decorrelation filter is arranged to decorrelate the delayed output signal by means of coefficients derived from the first cross-sectional decorrelation filter. The two cross-link decorrelation filters are configured to calculate their cross-member coefficients by means of adaptive decorrelation of the echo-canceled input signal.

The first decorrelation filter, a cross-member decorrelator, extracts from the echo-canceled signal the noise-like components contained therein. In parallel, in the second decorrelation filter, a cross-gate filter, the delayed output signal is converted into a transformed signal with the coefficients derived from the cross-divisor decorrelator. The special feature of this arrangement is the interchanging of the cross-link decorrelator and the cross-link filter over the conventional arrangement, in which not the echo-canceled signal, but the delayed output signal is decorrelated. The circuit according to the invention has the great advantage that the spectral maxima present in the hearing correction are preserved in the transformed signal. These maxima usually correspond to the most critical frequencies for the feedback, and these should certainly be taken into account in the updating of the filter coefficients with the correspondingly large weighting.

In the method according to the invention for the adaptive suppression of the acoustic feedback, an electrical input signal is generated with at least one microphone, a feedback characteristic is modeled with a filter, current updating coefficients for the Calculates a filter, a subtracter calculates an echo-canceled input signal by subtracting an echo estimate provided by the filter from a digital input signal, and a delay element calculates a delayed output signal. The echo-canceled input signal is decorrelated with a first cross-divisional decorrelation filter, and the delayed output signal is decorrelated with a second cross-divisional decorrelation filter by means of coefficients derived from the first cross-divisional decorrelation filter. The cross-member coefficients of the two cross-member decorrelation filters are calculated by adaptive decorrelation of the echo-canceled input signal.

The present invention differs significantly from all previously published systems for suppressing the acoustic feedback. What is new are the special arrangement and realization of the blocks for decorrelation and normalization, the control of the forgetting factor and the step size factor, as well as the possibility of staggered updating in the combination according to the invention. The present invention allows maximum convergence speeds with minimal distortion, since the updating of the filter coefficients takes place in terms of time and frequency mainly where the large amplifications in the auditory correction occur.

Fig. 1

ein allgemeines System zur adaptiven Unterdrückung der akustischen Rückkopplung gemäss Stand der Technik,

Fig. 2

ein System mit Verwendung eines Rauschsignals gemäss Stand der Technik,

Fig. 3

ein System mit Verwendung von orthogonalen Transformationen gemäss Stand der Technik,

Fig. 4

ein System mit Verwendung von adaptiven Dekorrelationsfiltern gemäss Stand der Technik,

Fig. 5

das erfindungsgemässe System,

Fig. 6

eine Detailzeichnung eines Verzögerungselements des erfindungsgemässen Systems,

Fig. 7

eine Detailzeichnung eines Filters des erfindungsgemässen Systems,

Fig. 8

eine Detailzeichnung einer Aufdatierungseinheit des erfindungsgemässen Systems,

Fig. 9

eine Detailzeichnung einer Normierungseinheit des erfindungsgemässen Systems,

Fig. 10

eine Detailzeichnung einer Geschwindigkeitssteuerungseinheit des erfindungsgemässen Systems,

Fig.11

eine Detailzeichnung eines Kreuzglied-Dekorrelators des erfindungsgemässen Systems,

Fig. 12

eine Detailzeichnung eines Kreuzglied-Filters des erfindungsgemässen Systems und

Fig. 13

eine Detailzeichnung einer Kontrolleinheit des erfindungsgemässen Systems.

In the following, the invention and for comparison, the prior art will be described in detail with reference to figures. Here are shown in block diagrams:

Fig. 1

a general system for adaptive suppression of the acoustic feedback according to the prior art,

Fig. 2

a system using a noise signal according to the prior art,

Fig. 3

a system using orthogonal transformations according to the prior art,

Fig. 4

a system using adaptive decorrelation filters according to the prior art,

Fig. 5

the system according to the invention,

Fig. 6

a detailed drawing of a delay element of the inventive system,

Fig. 7

a detailed drawing of a filter of the inventive system,

Fig. 8

a detailed drawing of an updating unit of the system according to the invention,

Fig. 9

a detailed drawing of a normalization unit of the inventive system,

Fig. 10

a detailed drawing of a speed control unit of the inventive system,

Figure 11

a detailed drawing of a cross-member decorrelator of the inventive system,

Fig. 12

a detailed drawing of a cross-member filter of the inventive system and

Fig. 13

a detailed drawing of a control unit of the inventive system.

A well-known system for the adaptive suppression of the acoustic feedback is in FIG. 1 shown. An acoustic input signal a _in (t) is picked up by a microphone 1 and initially converted into an electrical signal d (t). A subsequent AD converter 2 determines a digital Input signal d _n . Of these, an echo estimate y _{n is} subtracted in a subtractor 3. The echo-compensated signal e _n is transformed into a digital output signal u _n with a correction 4 adaptable to the respective application, for example an individual hearing correction for a hearing impaired person. The DA converter 5 performs a conversion into an electrical signal u (t), which is emitted via a loudspeaker or handset 6 as an acoustic output signal a _out (t). The acoustic output signal a _out (t) is deformed on its way back to the microphone 1 in accordance with a feedback characteristic 7 characterized by an impulse response h (τ) to a signal y (t) and superimposed on an externally incident acoustic signal s (t) (8 ). The remaining components in the system are a delay element 9, a filter 10 and an updating unit 11. The delay element 9 simulates the fixed delays contained in the system, resulting in a delayed signal x _n . The filter 10 models the unknown feedback characteristic. In the updating unit 11, the current coefficients w _n for the filter are continuously calculated. In this case, a variant of the LMS algorithm (Least Mean Square) is usually applied.

Because of the non-negligible autocorrelation function of real acoustic signals s (t), the well-known system does not suffice to achieve a distortion-free transmission in a realistic environment while at the same time satisfying convergence behavior. The system can be improved if the updating unit works with decorrelated signals.

FIG. 2 shows a system which uses an artificial noise signal to decorrelate the signals. Such a system is, for example, from the European patent applications EP-415 677 . EP-634 084 and EP-671 114 known by the company GN Danavox AS. The artificial noise signal is generated in a noise generator 17 and added via a power control unit 18 to the digital output signal u _n (19). The artificial noise signal is also fed via a delay element 20 to the updating unit 11. The noise signal is either switched on only if necessary instead of the output signal u _n or continuously added to the output signal u _n .

FIG. 3 shows a system which uses fixed orthogonal transformations to decorrelate the signals. Such a system from Phonak AG was, for example, as a European patent application EP-585 976 released. The echo-compensated signal e _n and the output signal u _n are transformed into the frequency domain via transformation units 21 and 22 and the echo estimate y _n is recovered via an inverse transformation 23. The filtering and updating of the coefficients in these systems does not take place directly in the time domain.

FIG. 4 FIG. 12 shows a system which uses adaptive decorrelation filters 12, 13 to decorrelate the signals. Such a system was, for example, in Mamadou Mboup et al., "Coupled Adaptive Prediction and System Identification: A Statistical Model and Transient Analysis", Proc. 1992 IEEE ICASSP, 4; 1-4, 1992 , described. The echo-canceled signal e _n and the delayed output signal x _n are decorrelated by the adaptive decorrelation filters 12, 13. The coefficients a _{n of} the two decorrelation filters 12, 13 are calculated in block 13 by means of decorrelation of the delayed output signal x _n .

An embodiment of a system according to the invention is shown in FIG FIG. 5 shown. In addition to the blocks 1 to 11 described above, the system according to the invention uses adaptive cross-member decorrelation filters, namely a cross-member decorrelator 12 and a cross-member filter 13 running parallel to it. For rapid decorrelation, the cross-member filter structures known from speech signal processing prove to be particularly suitable , They are used there for linear prediction. Algorithms for decorrelating a signal using cross-link filters are known.

The cross-correlator member 12 extracted from the echo-compensated signal e _n is given by noise-like components e ^M _n. In parallel, the cross member filter 13 with originating from the cross-over element de-correlator 12 coefficients k _n the delayed output signal x _n x ^M _n converted into a transformed signal. The special feature of this arrangement is the permutation of the two adaptive decorrelation filters 12 and 13 compared to the usual procedure, namely, not the echo-canceled signal e _n , but the delayed signal x _{n is} decorrelated. However, the arrangement according to the invention has the great advantage that the spectral maxima present in the hearing correction 4 are maintained in the transformed signal x ^M _n . These maxima usually correspond to the most critical frequencies for the feedback, and these should definitely be taken into account in the updating of the filter coefficients w _n with the correspondingly large weighting.

The order of the two cross-link decorrelation filters 12, 13 is determined by a compromise between the desired degree of decorrelation and the associated computation effort. For the special case of second-order filters (M = 2), a considerable improvement in the system behavior is again achieved by means of an upper boundary of the second cross-member coefficient k _2n . This upper limit of the second cross-member coefficient has the consequence that pure sine tones are not completely decorrelated. This in turn has the great advantage that the whistling sounds occurring during unstable operation are compensated much faster.

Furthermore, the system according to the invention contains a control unit 14. The control unit 14 continuously compares the power of the input signal d _n with the power of the echo-compensated signal e _n . The ratio of the two powers determines which forgetting factor λ _n in the updating unit 11 Application comes. If the power of the echo-compensated signal is greater than the power of the input signal, this is almost always an indication that the echo estimate y _n and thus the coefficients w _{n of} the filter 10 are too large in terms of magnitude. By setting λ _n <1, the coefficients rapidly converge to a more appropriate value. In normal operation, however, λ _n = 1 is set. The described control of the forgetting factor λ _n provides improved convergence behavior with rapid changes in the feedback path. An internal feedback generated temporarily by the system is recognized immediately and quickly adapted to the external feedback path.

As a further difference from other systems, the updating unit 11 includes a normalization unit 15 and a speed control unit 16. The arrangement of the blocks described below is shown in FIG FIG. 8 which illustrates a specification of the updating unit 11. The normalization unit 15 allows the application of the NLMS algorithm (Normalized Least Mean Square). It calculates the power of the signal e ^M _n . The special feature of this arrangement is that the normalization with respect to e ^M _n and not as usual with respect to x ^M _n . Thus, the convergence speed becomes dependent on the ratio of the powers of x ^M _n and e ^M _n . This ratio is essentially given by the gain contained in the hearing correction 4. The gain in the hearing correction is not constant in the general, non-linear case (eg compression method). In the method according to the invention, therefore, the convergence behavior of the adaptive filter 10 modeling the feedback characteristic 7 depends on the temporal behavior of the hearing correction 4, ie on the temporal course of its amplification and frequency response. In times of high gain with particularly critical feedback behavior, a fast adaptation of the coefficients w _{n occurs} and in times of small amplification with uncritical feedback behavior a correspondingly slower adaptation takes place. The updating takes place mainly in the times when it is actually necessary. This approach combines rapid convergence in the critical case with an approximately distortion-free processing in the uncritical case.

The speed control unit 16 provides a step factor β _n for the NLMS algorithm. The speed control unit 16 delivers values for β _n beginning at the start value β _max and within the first seconds after start-up gradually decreasing to the final value β _min . This procedure allows a very fast convergence of the filter coefficients w _n from zero to their nominal values after startup. The resulting initial signal distortions are less severe than the otherwise much longer-lasting feedback whistles.

The updating unit 11 can be designed so that only a certain small, cyclically changing part of the (N + 1) filter coefficients is updated at each discrete point in time. This considerably reduces the required computing effort. The system does not have to be made slower than it needs to be to prevent audible distortion anyway.

$$T=1/f_s{f}_s=16\mathit{kHz}$$

Für den Subtrahierer 3 und die Hörkorrektur 4 gelten die nachfolgenden Beziehungen. Die Funktion f() steht für eine beliebige nichtlineare Funktion ihrer Argumente. Sie ergibt sich aufgrund des ausgewählten Verfahrens zur Korrektur des individuellen Hörverlustes. $$e_n=d_n-y_n$$

$$u_n=f\left(e_0,,e_1,,e_2,\dots ,e_n\right)$$

In the following, a specific embodiment of the present invention is based on FIG. 5 described in more detail. The microphone 1, the AD converter 2, the DA converter 5 and the handset 6 are considered ideal in the consideration. The characteristics of the real acoustic and electrical transducers can be considered as part of the feedback characteristic 7. For the AD converter 2 and the DA converter 5, the following relationships apply. Here, T and f _s denote the sampling period and the subscript n the discrete time. $$d_n=d\left(n\cdot T\right)u\left(n\cdot T\right)=u_n$$

$$T=1/f_s{f}_s=16\mathit{kHz}$$

For the subtractor 3 and the hearing correction 4, the following relationships apply. The function f () stands for any nonlinear function of its arguments. It results from the selected procedure for correcting individual hearing loss. $$e_n=d_n-y_n$$

$$u_n=f\left(e_0,,{\mathrm{<figure-callout\; id="1"\; label="e"\; filenames="imgf0001.png,imgf0002.png"\; state="\{\{state\}\}">e</figure-callout>}}_1,,{\mathrm{<figure-callout\; id="2"\; label="e"\; filenames="imgf0001.png,imgf0002.png"\; state="\{\{state\}\}">e</figure-callout>}}_2,...,e_n\right)$$

$$a_{\mathit{in}}\left(t\right)=s\left(t\right)+y\left(t\right)$$

The acoustic transmission path is modeled by means of the feedback characteristic 7 and an adder 8. The operator * is to be understood as a convolution operator and h (τ) stands for the impulse response of the feedback. The externally incident signal is denoted by s (t). $$y\left(t\right)=a_{\mathit{out}}\left(t\right)*H\left(\mathit{\tau }\right)$$

$$a_{\mathit{in}}\left(t\right)=s\left(t\right)+y\left(t\right)$$

$$L=16\dots 24\left(L\cdot T=1\mathit{ms}\dots 15\mathit{ms}\right)$$

The delay element 9 is in FIG. 6 and the following relationships apply. The delay length L must be matched to the sum of the delays of the acoustic and electrical converters. $$x_n=u_{n-L}$$

$$L=16...24\left(L\cdot T=1\mathit{ms}...15\mathit{ms}\right)$$

$$\begin{array}{l}r=1/128,1/64,1/32,1/16,1/8,1/4,1/2,1/1\\ N=32\dots 64\left(N\cdot T=2\mathit{ms}\dots 4\mathit{ms}\right)\end{array}$$

The filter 10 is in FIG. 7 and the following relationships apply. Underlined sizes mean the similar elements combined into vectors. The factor r allows range selection, so that the filter coefficients are independent of the hearing correction 4 can always be kept in the range -1 ≤ w _kn <1. The filter order N must be matched to the length of the impulse response h (τ). $$y_n=r\cdot {\underset{}{w}}_n^T\cdot {\underset{}{x}}_n=r\cdot {\displaystyle \underset{k=0}{\overset{N}{\Sigma }}}{w}_{k.n}\cdot x_{n-k}$$

$$\begin{array}{l}r=1/128.1/64.1/32.1/16.1/\mathrm{8th}.1/4.1/2.1/1\\ N=32...64\left(N\cdot T=2\mathit{ms}...4\mathit{ms}\right)\end{array}$$

$$w_{k,n+1}={\mathit{\lambda }}_n\cdot w_{k,n}+{\beta }_n\cdot \frac{e_n^M}{n_n}\cdot x_{n-k}^M$$

The updating unit 11 is in FIG. 8 and the following relationships apply. The formula is given in vector notation and element notation. $${\underset{}{w}}_{n+1}={\mathit{\lambda }}_n\cdot {\underset{}{w}}_n+{\beta }_n\cdot \frac{e_n^M}{n_n}\cdot {\underset{}{x}}_n^M$$

$$w_{k.n+1}={\mathit{\lambda }}_n\cdot w_{k.n}+{\beta }_n\cdot \frac{e_n^M}{n_n}\cdot x_{n-k}^M$$

$$\begin{array}{l}{c}_n=\left(c_{n-1}+2\right)\mathrm{mod}\left(N+1\right)\\ N=47K=4\end{array}$$

In the preferred embodiment, not all (N + 1) filter coefficients are updated simultaneously, but only K at a time. The following relationships hold, assuming that K is an integer divisor of (N + 1). The variable c _n is used as a count variable. $$k=K\cdot \mathrm{int}\left(\frac{c_{n-1}}{K}\right).....K\cdot \mathrm{int}\left(\frac{c_{n-1}}{K}\right)+K-1$$

$$\begin{array}{l}{c}_n=\left(c_{n-1}+2\right)\mathrm{mod}\left(N+1\right)\\ N=47K=4\end{array}$$

$$g=63/64h=1-g=1/64$$

The updating unit 11 in turn includes the normalization unit 15 and the speed control unit 16. The normalization unit 15 is shown in FIG FIG. 9 and the following relationships apply. The coefficients g and h determine the length of the time interval over which an averaging of the power of e ^M _n takes place. $$n_n=G\cdot n_{n-1}+H\cdot {\left(e_n^M\right)}^2$$

$$G=63/64H=1-G=1/64$$

$${\beta }_n=\{\begin{array}{ll}{\beta }_{n-1}& \left(c_n\ne 0\right)\\ \max \left(0.5\cdot {\beta }_{n-1},{\beta }_{\min }\right)& (c_n=0)\end{array}$$

$$c_n=\left(c_{n-1}+1\right)\mathrm{mod}P$$

$$P=4096\left(P\cdot T=256\mathit{ms}\right)$$

The speed control unit 16 is in FIG. 10 and the following relationships apply. The step size factor β _n is gradually reduced by a factor of 0.5 starting from β _max to β _min . The optimal values for β _max and β _min depend on the individual hearing correction (4). The variable c _n is used as a count variable. $${\beta }_{-1}={\beta }_{\mathrm{Max}}$$

$${\beta }_n=\{\begin{array}{ll}{\beta }_{n-1}& \left(c_n\ne 0\right)\\ \mathrm{Max}\left(0.5\cdot {\beta }_{n-1}.{\beta }_{\min }\right)& (c_n=0)\end{array}$$

$$c_n=\left(c_{n-1}+1\right)\mathrm{mod}P$$

$$P=4096\left(P\cdot T=256\mathit{ms}\right)$$

$$b_n^0=e_n$$

$$\begin{array}{l}{e}_n^i=e_n^{i-1}+k_{i,n}\cdot b_{n-1}^{i-1}\\ b_n^i=k_{i,n}\cdot e_n^{i-1}+b_{n-1}^{i-1}\\ d_n^i=g\cdot d_{n-1}^i+h\cdot \left[{\left(e_n^{i-1}\right)}^2+{\left(b_{n-1}^{i-1}\right)}^2\right]\\ n_n^i=g\cdot n_{n-1}^i+h\cdot \left[\left(-2\right)\cdot e_n^{i-1}\cdot b_{n-1}^{i-1}\right]\\ k_{i,n+1}=\frac{n_n^i}{d_n^i}\end{array}\}\left(1\le i\le M\right)$$

$$g=63/64h=1-g=1/64$$

$$M=2\dots 8$$

The cross-member decorrelator 12 is in FIG. 11 and the following relationships apply. In addition to the recursion formulas for the calculation of e ⁱ _n and b ⁱ _n , the quantities d ⁱ _n and n ⁱ _n must also be determined at each stage for tracking the coefficients k _in . The filter order M results from a compromise between the desired degree of decorrelation and the required computational effort. $$e_n^0=e_n$$

$$b_n^0=e_n$$

$$\begin{array}{l}{e}_n^i=e_n^{i-1}+k_{i.n}\cdot b_{n-1}^{i-1}\\ b_n^i=k_{i.n}\cdot e_n^{i-1}+b_{n-1}^{i-1}\\ d_n^i=G\cdot d_{n-1}^i+H\cdot \left[{\left(e_n^{i-1}\right)}^2+{\left(b_{n-1}^{i-1}\right)}^2\right]\\ n_n^i=G\cdot n_{n-1}^i+H\cdot \left[\left(-2\right)\cdot e_n^{i-1}\cdot b_{n-1}^{i-1}\right]\\ k_{i.n+1}=\frac{n_n^i}{d_n^i}\end{array}\}\left(1\le i\le M\right)$$

$$G=63/64H=1-G=1/64$$

$$M=2...\mathrm{8th}$$

$$k_{\max }=0.921875$$

In the preferred embodiment with filter order M = 2, a complete decorrelation is prevented by limiting the second coefficient k _2n . The following relationships apply. $${\mathrm{<figure-callout\; id="2"\; label="k"\; filenames="imgf0001.png,imgf0002.png"\; state="\{\{state\}\}">k</figure-callout>}}_{2.n}=\mathrm{<figure-callout\; id="2"\; label="min\; k"\; filenames="imgf0001.png,imgf0002.png"\; state="\{\{state\}\}">min\; </figure-callout>}\left(k_{}\right)\left({}_{2.n},,k_{\mathrm{Max}}\right)$$

$$k_{\mathrm{Max}}=0.921875$$

$$b_n^0=x_n$$

$$\begin{array}{c}{x}_n^i=x_n^{i-1}+k_{i,n}\cdot b_{n-1}^{i-1}\\ b_n^i=k_{i,n}\cdot x_n^{i-1}+b_{n-1}^{i-1}\end{array}\}\left(1\le i\le M\right)$$

The cross-member filter 13 is in FIG. 12 and the following relationships apply. $$x_n^0=x_n$$

$$b_n^0=x_n$$

$$\begin{array}{c}{x}_n^i=x_n^{i-1}+k_{i.n}\cdot b_{n-1}^{i-1}\\ b_n^i=k_{i.n}\cdot x_n^{i-1}+b_{n-1}^{i-1}\end{array}\}\left(1\le i\le M\right)$$

$$n_n^e=g\cdot n_{n-1}^e+h\cdot {\left(e_n\right)}^2$$

$${\mathit{\lambda }}_n=\{\begin{array}{ll}{\mathit{\lambda }}_{\mathit{off}}& \left(n_n^e\le n_n^d\right)\\ {\mathit{\lambda }}_{n-1}& \left(n_n^d<n_n^e\le 2\cdot n_n^d\right)\\ {\mathit{\lambda }}_{\mathit{on}}& \left(n_n^e>2\cdot n_n^d\right)\end{array}$$

$$g=63/64h=1-g=1/64$$

$${\mathit{\lambda }}_{\mathit{off}}=1.0{\mathit{\lambda }}_{\mathit{on}}=0.99\dots 0.9999$$

The control unit 14 is in FIG. 13 and the following relationships apply. The forgetting factor λ _n results from the ratio of the two powers n ^d _n and n ^e _n . There is a hysteresis in the middle area. $$n_n^d=G\cdot n_{n-1}^d+H\cdot {\left(d_n\right)}^2$$

$$n_n^e=G\cdot n_{n-1}^e+H\cdot {\left(e_n\right)}^2$$

$${\mathit{\lambda }}_n=\{\begin{array}{ll}{\mathit{\lambda }}_{\mathit{off}}& \left(n_n^e\le n_n^d\right)\\ {\mathit{\lambda }}_{n-1}& \left(n_n^d<n_n^e\le 2\cdot n_n^d\right)\\ {\mathit{\lambda }}_{\mathit{on}}& \left(n_n^e>2\cdot n_n^d\right)\end{array}$$

$$G=63/64H=1-G=1/64$$

$${\mathit{\lambda }}_{\mathit{off}}=1.0{\mathit{\lambda }}_{\mathit{on}}=0.99...0.9999$$

The preferred embodiment can be easily programmed on a commercial signal processor or implemented in an integrated circuit. All variables must be suitably quantized and the operations optimized to the existing architecture blocks. Special attention is given to the treatment of the quadratic quantities (powers) and the division operations. Depending on the target system, there are optimized procedures for this. However, these are not in and of themselves subject of the present invention.

Claims (10)

1. Circuit for the adaptive suppression of acoustic feedback in an acoustic system having at least one microphone (1) for producing an electric input signal (d(t)), at least one loudspeaker or receiver (6) respectively and an interposed electronic signal processing part, incorporating a filter (10) for modeling a feedback characteristic (7), an updating unit (11) for calculating actual coefficients (w_n) for the filter (10), a A/D-converter (2) for converting the electric input signal (d(t)) into a digital input signal (d_n), a subtracter (3) for calculating an echo-compensated input signal (e_n) by subtracting an echo estimate (y_n) delivered by the filter (10) from the digital input signal (d_n), a delay element (9) for calculating a delayed output signal (x_n), a first adaptive de-correlation filter (12) for de-correlation of the echo-compensated input signal (e_n) as well as a second adaptive de-correlation filter (13) for de-correlation of the delayed output signal (x_n), characterized in that the two de-correlation filters (12, 13) are designed as lattice de-correlation filters, wherein the second de-correlation filter (13) is arranged for de-correlation of the delayed output signal (x_n) by means of coefficients (k _n) from the first de-correlation filter (12), and wherein the two de-correlation filters (12, 13) are configured for calculating their lattice coefficients (k _n) by adaptive de-correlation of the echo-compensated input signal (e_n).
2. Circuit according to claim 1, characterized by a normalization unit (15) arranged in the updating unit (11) for normalization of a de-correlated, echo-compensated input signal (e^M _n) delivered by the first de-correlation filter (12).
3. Circuit according to Claim 1 or 2, characterized by a control unit (14) for monitoring the ratio of the powers of the digital input signal (d_n) and the echo-compensated input signal (e_n) and for controlling a forget factor (λ_n) in the updating unit (11).
4. Circuit according to one of the claims 1 to 3, characterized by a speed control unit (16) for calculating a step size factor (β_n) in the updating unit (11).
5. Method for the adaptive suppression of acoustic feedback, implementable by means of the circuit according to claim 1, wherein an electric input signal (d(t)) is produced by at least one microphone (1), a feedback characteristic (7) is modelled with a filter (10), actual coefficients (w_n) for the filter (10) are calculated by an updating unit (11), an echo-compensated input signal (e_n) is calculated by a subtracter (3) by subtraction of an echo estimate (y_n) delivered by the filter (10) from the digital input signal (d_n) and a delayed output signal (x_n) is calculated with a delay element (9), wherein this is converted into a digital input signal by means of a A/D-converter (2), the echo-compensated input signal (e_n) is de-correlated with a first lattice de-correlation filter (12) and the delayed output signal (x_n) is de-correlated by a second lattice de-correlation filter (13), characterized in that with the second lattice de-correlation filter (13) the delayed output signal (x_n) by means of coefficients (k _n) from the first lattice de-correlation filter (12), and that the lattice coefficients (k _n) of the two de-correlation filters (12, 13)are calculated by the adaptive de-correlation of the echo-compensated input signal (e_n)
6. Method according to claim 5, characterized in that there is a normalizeation of a de-correlated, echo-compensated input signal (e^M _n) delivered by the first de-correlation filter (12) in the updating unit (11).
7. Method according to claim 5 or 6, characterized in that a control unit (14) monitors the ratio of the powers of the digital input signal (d_n) and the echo-compensated input signal (e_n) and controls a forget factor (λ_n) in the updating unit (11).
8. Method according to one of the claims 5 to 7, characterized in that in the updating unit (11) a step size factor £ is reduced stepwise from a starting value following the starting up of the hearing aid until the optimum operating value is reached.
9. Method according to one of the claims 5 to 8, characterized in that second order lattice de-correlation filters (12, 13) are used and there is an upper limitation to the second lattice coefficient k_2n.
10. Method according to one of the claims 5 to 9, characterized in that all the filter coefficients (w_n) are not simultaneously updated in the updating unit (11) and instead only a small, cyclically changing part thereof is updated.

Images