This invention generally relates to a digital Acoustic Echo Control (AEC) in telephones and more specifically, to introducing an additional residual echo filter and modifying a control block means for achieving more consistent echo cancellation results and enhancing output signal quality.
The invention is related to a digital Acoustic Echo Control (AEC) unit of telephones. The purpose of the AEC is to prevent the far-end speaker's speech circulating back as an echo after coming out from the near-end phone user's loudspeaker and partly picked up by the phone's microphone. A general concept is illustrated in
The need for an AEC unit in the hands-free telephones basically arises from an acoustic echo path with an impulse response g(i) from a local loudspeaker 16 to a local microphone 18. The objective of the echo canceller 10 with an impulse response c(i) is to find a replica of the echo path in order to compensate for an echo signal d(i) 22 of a voice signal x(i) 20 received by a loudspeaker 16 that provides an acoustic output signal in response to the voice signal x(i) 20, thus generating in the microphone 18 the echo signal d(i) 22 which is one of the components of a microphone signal y(i)=d(i)+s(i)+n(i) 28, where y(i) is a microphone speech signal and n(i) is a background noise signal. As the system identification process is always performed in the presence of observation noise (local speech plus background noise), s(i)+n(i), the objective of c(i)=g(i) cannot be reached exactly. The echo canceller 10 generates an estimate echo signal d′(i) 32 which is negatively added to the microphone signal 18 by the adder 30 which generates an echo reduced microphone signal e(i) 34 containing the partially compensated echo signal. The echo reduced microphone signal e(i) 34 is further provided to the gradient adaptation means 12 and to the residual echo suppressor 14. The gradient adaptation means 12 further provides a control signal 15 to the echo canceller 10 by determining a gradient of the controlled signal based on predetermined criteria using the voice signal x(i) 20 and the echo reduced microphone signal e(i) 34 as input signals. The purpose of the residual echo suppressor 14 is further reducing of residual echo components of the echo reduced microphone signal e(i) 34. The resulting output system signal s′(i)+n′(i) 36 after residual echo suppression by the residual echo suppressor 14 is then transmitted to the far speaker.
The basic principles of how to generate and control the echo canceller 10 and the residual echo suppressor 14 are well known. However, there are some problems involved in controlling them efficiently in a most optimal way. The key variable in the whole control issue is the residual echo, b(i)=d(i)−d′(i) which, unfortunately, cannot be directly determined since it is inherently embedded in the echo reduced microphone signal e(i)=b(i)+s(i)+n(i) 34.
The echo canceller module 21 of
The object of the present invention is to provide a novel method for achieving more consistent echo cancellation results and enhancing output signal quality by introducing an additional residual echo filter and modifying a control block.
According to a first aspect of the present invention, an echo cancellation system comprises: a microphone, responsive to a resulting echo signal from a loudspeaker that provides an acoustic output in response to a speech signal, for providing an echo signal which is a component of a microphone signal; and a control block means, responsive to the speech signal, to an echo reduced microphone signal and to a further echo reduced microphone signal, for providing a first control signal to an echo canceller, a second control signal to a residual echo suppressor and a third control signal to a residual echo filter; wherein said control signals are provided for optimizing cancellation of the echo signal
According further to the first aspect of the invention, the first control signal may be a transfer function signal provided to the echo canceller, wherein said transfer function signal weights the voice signal.
Further according to the first aspect of the invention, the second control signal may be a further transfer function signal provided to the residual echo suppressor, said further transfer function signal weights an echo reduced microphone signal.
Still further according to the first aspect of the invention, the third control signal may be a residual transfer function signal provided to the residual echo filter, said residual transfer function signal weights the voice signal.
According further to the first aspect of the invention, the echo cancellation system may further comprise the residual echo filter, responsive to the speech signal and to the third control signal, for providing a further estimate echo signal to a further adder.
According still further to the first aspect of the invention, the echo cancellation system may further comprise the residual echo suppressor, responsive to an echo reduced microphone signal and to the second control signal, for providing an output system signal.
According still yet further to the first aspect of the invention, the echo cancellation system may further comprise the echo canceller, responsive to the voice signal and to the first control signal, for providing an estimate echo signal to an adder. Further, the echo cancellation system may further comprise the residual echo filter, responsive to the speech signal and to the third control signal, for providing a further estimate echo signal to a further adder. Still further, the echo cancellation may further comprise the residual echo suppressor, responsive to the echo reduced microphone signal and to the second control signal, for providing an output system signal.
According further still to the first aspect of the invention, the echo cancellation system may further comprise an adder, responsive to a microphone signal and to an estimate echo signal, for providing an echo reduced microphone signal.
According yet further still to the first aspect of the invention, the echo cancellation system may further comprise a further adder, responsive to the echo reduced microphone signal and to a further estimate echo signal, for providing the further echo reduced microphone signal.
Yet still further according to the first aspect of the invention, the residual echo filter, the echo canceller and the residual echo suppressor may operate in a time domain, and said first, second and third control signals are provided then in the time domain as well. Further, the residual echo filter, the echo canceller and the residual echo suppressor may operate in a frequency domain, and said first and second control signals are provided then in the frequency domain as well.
Still yet further according to the first aspect of the invention, the residual echo filter and the echo canceller may operate in a time domain, and said first and third control signals are then provided in the time domain as well. Further, the residual echo suppressor may operate in a frequency domain, and said second control signal is provided then in the frequency domain as well.
According to a second aspect of the present invention, a method for acoustic echo control, comprising the steps of: providing an echo signal which is a component of a microphone signal of a microphone which is responsive to an echo signal from a loudspeaker that provides an acoustic output signal in response to a voice signal; and providing a first control signal to an echo canceller, a second control signal to a residual echo suppressor and a third control signal to a residual echo filter by a control block means which is responsive to the speech signal, to an echo reduced microphone signal and to a further echo reduced microphone signal for optimizing cancellation of the echo signal.
According further to the second aspect of the invention, the first control signal may be a transfer function signal provided to the echo canceller, said transfer function signal weights the voice signal.
Further according to the second aspect of the invention, the second control signal may be a further transfer function signal provided to the residual echo suppressor, said further transfer function signal weights an echo reduced microphone signal.
Still further according to the second aspect of the invention, the third control signal may be a residual transfer function signal provided to the residual echo filter, said residual transfer function signal weights the voice signal.
According further to the second aspect of the invention, wherein prior to the step of providing the first, the second and the third control signals, the method may further comprise the step of: determining the first, the second and the third control signals by a statistical adaptive-filter controller.
According still further to the second aspect of the invention, the method may further comprise the steps of: providing an estimate echo signal by the echo canceller using the first control signal provided by the controlled block means; and providing the echo reduced microphone signal by an adder by adding the estimate echo signal to the microphone signal. Further, the method may further comprise the steps of: providing a further estimate echo signal by the residual echo filter using the third control signal provided by the controlled block means; and providing a further echo reduced microphone signal by an adder by adding a further estimate echo signal to a microphone signal. Still further, the method may further comprise the steps of: providing an output system signal by the residual echo suppressor using the second control signal provided by the control block means.
For a fuller understanding of the nature and objects of the present invention, reference is made to the following detailed description taken in conjunction with the following drawings, in which:
This invention generally relates to a digital Acoustic Echo Control (AEC) in telephones. It discloses an additional adaptive filter referred to as a residual echo filter, which is placed after the usual echo canceller module 21 of
β(i)=E{∥g(i)−c(i)∥2},
which can be utilized to establish an efficient control for the AEC system and construct the appropriate filters (bold font in the equation is used for vectors). This invention also describes a control block means performing a joint operation control of the echo canceller, residual echo suppressor and residual echo filter for achieving more consistent echo cancellation results and enhancing output signal quality. This is illustrated in
The blocks 10 and 14 are described above in regard to
The first control signal 46 can be a transfer function signal provided to an echo canceller 10, said transfer function signal weights the voice signal 20. If the transfer function signal 46 is in a time domain, it is an impulse function c(i) (defined in regard to
The second control signal 54 can be a further transfer function signal of the residual echo suppressor 14, said further transfer function signal weights an echo reduced microphone signal 34 for generating a high quality undistorted microphone signal 36.
The third control signal 48 can be a residual transfer function signal provided to the residual echo filter 42, said residual transfer function signal weights the voice signal 20. If the transfer function signal 48 is in the time domain, it is an impulse function h(i) of the residual echo filter 42. The residual echo filter 42 generates a further estimate echo signal b′(i) 50 which is negatively added to the echo reduced microphone signal e(i) 34 by an adder 44 which generates the further echo reduced microphone signal f(i) 52 with the further partially compensated echo signal. The further echo reduced microphone signal f(i) 52 is further provided to the CBM 40.
The echo cancellation system 11 can operate in a time domain or in a frequency domain. This implies that the echo canceller 10, the residual echo filter 42 and the residual echo suppressor 14 can operate in the time or frequency domain, and the first and second control signals can also be provided in the time or frequency domain, respectively. Other variations are possible. For example, the echo canceller 10 and the residual echo filter 42 with corresponding the first and third control signals 46, 54, respectively, can be implemented in the time domain and the residual echo suppressor 14 with the third control signal 48 being implemented in the frequency domain.
To further illustrate the performance of the system 11 of
The echo canceller 10 in
wherein c(i) is a transfer function signal in the time domain provided to the echo canceller 10 by the CBM 40 as the first control signal 46, the μ1(k) is a step-size determined as an estimation of the optimal step-size criteria,
wherein {overscore (x2(i))} is a voice signal power, {overscore (e2(i))} is an echo reduced microphone signal power, and 0<γ<1 is a smoothing coefficient. The coupling factor, β1(i)=E{∥g(i)−c(i)∥2}, is estimated with the help of a residual transfer function signal h(i), the third control signal, 48 provided to the residual echo filter 42. Before showing how, some characteristics of the system mismatch vector,
Δ(i)=g(i)−c(i) (7),
are described. One of the well-known characteristics of the NLMS filter is its tendency to disperse the estimation error energy evenly among its filter coefficients. Furthermore, the errors in each coefficient can be assumed only weakly correlated. As a result, the system mismatch or error vector has a random noise like characteristic, which corresponds to a flat magnitude squared transfer function in the frequency domain.
The residual echo filter attached to the AEC system as illustrated in
h(i)≈Δ(i)=g(i)−c(i) (8).
Thus, the coupling factor can be estimated from the residual echo filter coefficients. Furthermore, the fact that the error energy is evenly distributed among the mismatch vector makes it possible to utilize shorter filter lengths for the residual echo filter 42 than the main filter (echo canceller 10). The coupling factor estimation can then be interpolated to represent the main filter as follows:
wherein N is an echo canceller 10 filter length, and L is a residual echo filter 42 length with N>L. The expectation operation is inherently estimated as an averaging by the residual echo filter 42.
The residual echo filter shown in
wherein h(i) is a residual transfer function signal in the time domain, the third control signal 48 provided to the residual echo filter 42. The further step-size, μ2(i), for the residual echo filter is also estimated according to the optimal rule described by Equation 4, as follows:
wherein {overscore (f2(i))} is a further echo reduced microphone signal power.
The difference is that a further coupling factor, 0<β2<1, is a constant. As a result, the filter does not adapt optimally, that, in turn, degrades the residual echo estimation performance. However, the estimate does not have to be accurate in order to provide an already very good overall echo cancellation performance. Moreover, the estimate accuracy can be enhanced by utilizing knowledge about the system mismatch.
Realization of the β1 estimator is described below. The case β2>β1(i) can be achieved by choosing, for example, β2=0.1 . . . 1.0. The system coupling factor, β1(i), is usually smaller than this value, motivated by the physical nature of an acoustic echo path which is not performing power amplification. If β2>β1(i), the residual echo filter delivers an estimate of the time varying residual echo path h(i) quickly, but momentarily inaccurate. As a consequence, the corresponding system coupling factor estimate, β1(i), also becomes unreliable and biased. Thus, a bias correction (derived from the statistics of LMS type adaptive filters) can be applied to β1(i).
In this way, an unbiased estimate of β1(i) can be determined quickly and reliably at the same time. The corresponding echo canceller c(i) thus tracks room impulse response changes quickly but also delivers accurate estimates in the “steady state”. The corresponding postfilter (the residual echo suppressor 14 of
The residual echo suppression filter 14 can be constructed, e.g., according to Wiener rule in the frequency domain, implemented as the Discrete Fourier Transform (DFT) domain, as follows:
wherein W(Ω,k) is a further transfer function signal, the second control signal 54, provided to the residual echo suppressor 14 in the form of the DFT and, ΦE(Ω,k) and ΦB(Ω,k) are power spectral density (PSD) signals of the echo reduced microphone signal e(i) 34 and a residual echo signal b(i)=d(i)−d′(i) (a difference between an echo signal d(i) and an estimate echo signal d′(i) 32) with frequency and frame indexes Ω and k, respectively. The residual echo PSD can be estimated with the help of the coupling factor described by Equation 9 as follows:
ΦB(Ω,k)=β1(k)ΦX(Ω,k) (14),
wherein β1(k)=β1(i) is the coupling factor estimated at the time corresponding to the end of a frame. Since the residual echo filter has a flat magnitude squared transfer function in the frequency domain, its effect on the excitation signal, x(i), in the frequency domain can be modeled only with the level change. Thus, the frequency independent coupling factor, β1(k) can be directly utilized.
The residual echo filter 42 outputs a further estimate echo signal b′(i). This output can be optionally used to estimate directly the estimate echo signal power, e.g., with a first order recursive smoothing equation:
{overscore (b′2(i))}≈(1−γ)b′2(i)+γ{overscore (b′2(i−1))} (15),
wherein {overscore (b′2(i))} is a further estimate echo signal power of a further estimate echo signal b′(i) 50 and 0<γ<1 is a smoothing coefficient. The result can be directly used in Equation 3. However, the residual echo power estimation through the coupling factor using Equation 4 is preferred, since it is easier to handle, according to the present invention.
Equations 9-12 and 14 are novel and original and constitute a new methodology for enhancing the acoustic echo cancellation system using residual echo filter, according to the present invention.
In a step 78, the residual echo filter 42 with adder 44 performs NLMS algorithm using Equation 10. In a next step 80, the further echo reduced microphone signal f(i) 52 is provided to the CBM 40.
In a step 82, the echo canceller 10 with adder 30 perform NLMS algorithm using Equation 1. In a next step 84, the echo reduced microphone signal e(i) 34 is provided to the CBM 40, to the adder 30 and to the residual echo suppressor 14.
In a next step 86, the residual echo suppressor 14 minimizes the residual echo component of the echo reduced microphone signal 34 using the further transfer function signal W(Ω, k), the third control signal 54, which weights the echo reduced microphone signal 34 in the frequency domain for generating a high quality undistorted microphone signal 36 and for providing the signal 36 to the far user.
After steps 80, 84 and 86, in a next step 88, a determination is made whether communication (e.g., phone conversation) is still on. If not, the process stops. If communication is still on, the process returns to step 64.
This application claims priority from U.S. Provisional Patent Application Ser. No. 60/483,265, filed Jun. 27, 2003. This application discloses subject matter which is also disclosed and which may be claimed in copending, co-owned U.S. patent application Ser. No. 10/608,785, filed Jun. 27, 2003.
Number | Date | Country | |
---|---|---|---|
60483265 | Jun 2003 | US |