Recursive Bayesian controllers for non-linear acoustic echo cancellation and suppression systems

• US 8,924,337 B2
• Filed: 05/09/2011
• Issued: 12/30/2014
• Est. Priority Date: 05/09/2011
• Status: Expired due to Fees

• Current Assignee: RPX Corporation
• Original Assignee: Nokia Corporation
• Inventors: Malik, Sarmad, Enzner, Gerald, Vartiainen, Jukka Petteri, Sjoberg, Jari Petteri, Myllyla, Ville Mikael
• Primary Examiner(s): Chaki, Kakali
• Assistant Examiner(s): HANCHAK, WALTER P
• Application Number:
  US13/068,341
• Publication Number:
  US 20120290525A1
• Time in Patent Office: 1,331 Days
• Field of Search: None
• US Class Current: 706/52
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks

Abstract

Both a cascade and a multichannel joint Bayesian estimator are provided for suppressing acoustic echo. An expansion basis (Power/Fourier series) is selected to convert a sample-based input signal x_t into a DFT-domain multichannel signal [X_τ,1, . . . X_τ,p]. The posterior of unknown states (e.g., mean Ŵ_τ and covariance P_τ of the echo path W_τ and the mean â_τ and covariance Q_τ of the nonlinear coefficients a_τ; or channel-wise mean Ŵ_τ,i and multichannel covariance P_τ of a compound quantity formed by merging together the echo path W_τ and the i^th nonlinear coefficient a_τ,i) and model parameters θ_τ are estimated; and Kalman gain factor(s) K_τ are computed for optimal adaptation of the posterior of unknown states. An echo signal Ŷ_τ is estimated using the multichannel input signal [X_τ,1, . . . X_τ,p] and the adapted posterior; and an error signal E_τ is generated. Residual echo is suppressed by post-filtering the error signal E_τ with a weighting function ψ_τ which depends on the adapted posterior, and the filtered error signal ŝ′_t is then transmitted to a far-end.

Claims

1. A method for a handheld portable communication device, comprising:

performing an echo suppression process on the handheld portable communication device by performing at least the following;

estimating a model parameter set comprising at least one noise covariance as a function of at least an estimated or a predicted posterior of at least one unknown state quantity, a far-end signal and an error signal;

generating a gain factor in dependence on the at least one noise covariance;

using the gain factor to adapt the posterior of the at least one unknown state quantity;

estimating an acoustic echo signal using the adapted posterior of the at least one unknown state quantity, wherein the acoustic echo signal comprises an estimate of an acoustic echo caused at least by the far-end signal being played by a loudspeaker and resultant output picked up by a microphone;

subtracting the estimated acoustic echo signal from a microphone signal from the microphone to generate an error signal;

filtering the error signal to suppress residual linear and nonlinear error caused at least by the acoustic echo according to a weighting function which depends on the error signal, the far-end signal and the adapted posterior of the at least one unknown state quantity;

outputting the filtered error signal for transmission to a far-end,

in which the handheld portable communication device comprises;

a radio receiver which provides the far-end signal from which the microphone signal derives; and

a radio transmitter for sending the filtered error signal to a source from which the radio receiver received the far-end signal.

2. The method according to claim 1, executed in the frequency domain in which the far-end signal is converted to a multichannel input by operating on the far-end signal by an expansion series selected in dependence on the error signal.

3. The method according to claim 1, in which the at least one unknown state quantity comprises a mean echo path and its estimation error covariance, and in which estimating the acoustic echo signal comprises explicitly modeling coefficients of loudspeaker nonlinearity and unknown echo path states as random variables with a first-order Markov property, in which the posterior on the coefficients and the unknown echo path states are adapted by respective gain factors.

4. The method according to claim 3, in which:

the model parameter set comprises process noise covariance for the echo path state, process noise covariance for nonlinear expansion coefficients, and observation noise covariance;

the adapted posterior of the unknown echo path states is conditional on the process noise covariance for the echo path state and the observation noise covariance; and

the adapted posterior of the non-linear coefficients is conditional on the process noise covariance for nonlinear expansion coefficients and the observation noise covariance.

5. The method according to claim 4, in which the estimated acoustic echo signal is generated by convolving an estimated non-linearly mapped input signal with a mean of the unknown echo path states.

6. The method according to claim 1, in which the handheld portable communication device comprises a multichannel system, in which the gain factor comprises channelwise gain parameters and the estimated acoustic echo signal is generated from channelwise estimated echo signals.

7. The method according to claim 6, in which:

the model parameter set comprises multichannel process noise covariance for the echo path state and observation noise covariance; and

the channelwise gain parameter is generated in dependence on an estimate of the near-end noise covariance via a common inverse-term.

8. A handheld portable communication device, comprising at least one processor and at least one memory storing computer program instructions;

in which the at least one processor is configured with the at least one memory and the computer program instructions to at least;

perform an echo suppression process on the handheld portable communication device by performing at least the following;

estimate a model parameter set comprising at least one noise covariance as a function of at least an estimated or a predicted posterior of at least one unknown state quantity, a far-end signal and an error signal;

generate a gain factor in dependence on the at least one noise covariance;

adapt the posterior of the at least one unknown state quantity using the gain factor;

estimate an acoustic echo signal using the adapted posterior of the at least one unknown state quantity, wherein the acoustic echo signal comprises an estimate of an acoustic echo caused at least by the far-end signal being played by a loudspeaker and resultant output picked up by a microphone;

subtract the estimated echo signal from a microphone signal from the microphone to generate an error signal;

filter the error signal to suppress residual linear and nonlinear error according to a weighting function which depends on the error signal, the far-end signal and the estimated posterior of the at least one unknown state quantity;

output the filtered error signal for transmission to a far-end,

in which the handheld portable communication device further comprises;

a radio receiver which provides the far-end signal from which the microphone signal derives; and

a radio transmitter for sending the filtered error signal to a source from which the radio receiver received the far-end signal.

9. The handheld portable communication device according to claim 8, in which the at least one processor is configured with the at least one memory and the computer program instructions to estimate, generate, adapt, subtract and filter as said in claim 8 in the frequency domain, and further to:

convert the far-end signal to a multichannel input by operating on the far-end signal by an expansion series selected in dependence on the error signal.

10. The handheld portable communication device according to claim 8, in which the at least one unknown state quantity comprises a mean echo path and its estimation error covariance,

and in which the at least one processor is configured with the at least one memory and the computer program instructions estimate the echo signal by explicitly modeling coefficients of loudspeaker nonlinearity and unknown echo path states as random variables with a first-order Markov property, in which the posterior on the coefficients and the unknown echo path states are adapted by respective gain factors.

11. The handheld portable communication device according to claim 10, in which:

the model parameter set comprises process noise covariance for the echo path state, process noise covariance for nonlinear expansion coefficients, and observation noise covariance;

the adapted posterior of the unknown echo path states is conditional on the process noise covariance for the echo path state and the observation noise covariance; and

the adapted posterior of the non-linear coefficients is conditional on the process noise covariance for nonlinear expansion coefficients and the observation noise covariance.

12. The handheld portable communication device according to claim 11, in which the at least one processor is configured with the at least one memory and the computer program instructions to generate the estimated acoustic echo signal by convolving an estimated non-linearly mapped input signal with a mean of the unknown echo path states.

13. The handheld portable communication device according to claim 8, in which:

the handheld portable communication device comprises a multichannel system, in which the gain factor comprises a channelwise gain parameter; and

the at least one processor is configured with the at least one memory and the computer program instructions to generate the estimated acoustic echo signal from channelwise estimated echo signals.

14. The handheld portable communication device according to claim 13, in which:

the model parameter set comprises multichannel process noise covariance for an echo path state and observation noise covariance; and

at least one processor is configured with the at least one memory and the computer program instructions to generate the channelwise gain parameter in dependence on an estimate of the near-end noise covariance via a common inverse-term.

15. A method for a handheld portable communication device, comprising:

performing an echo suppression process on the handheld portable communication device by performing at least the following;

selecting, based on an error signal, an expansion basis from a plurality of basis function expansions for operating the estimator;

using the selected expansion basis for converting a sample based input signal into a frequency domain multichannel signal having a number of portions, where the number of portions is a total number of channels or an expansion order of the selected expansion basis;

estimating posteriors of at least two unknown state quantities;

estimating model parameters comprising at least one noise covariance as a function of the estimated posteriors of the at least two unknown state quantities, the input signal and a microphone signal from a microphone;

computing at least one gain factor for adapting the posteriors of the at least two unknown state quantities in dependence on at least one of the estimated model parameters, the input signal, and the estimated posteriors of the at least two unknown state quantities;

estimating an acoustic echo signal from the multichannel input signal and the adapted posteriors of the at least two unknown state quantities, wherein the acoustic echo signal comprises an estimate of an acoustic echo caused at least by the far-end signal being played by a loudspeaker and resultant output picked up by the microphone;

generating the error signal by subtracting the estimated echo signal from a frequency-domain version of the microphone signal;

suppressing residual linear and nonlinear error caused at least by the acoustic echo by filtering the error signal using a weighting function that depends at least on the error signal, the input signal and directly or indirectly on the adapted posteriors of the unknown state quantities;

outputting the filtered error signal for transmission to a far-end;

in which the handheld portable communication device comprises;

a radio receiver which provides the far-end signal from which the microphone signal derives; and

a radio transmitter for sending the filtered error signal to a source from which the radio receiver received the far-end signal.

16. The method according to claim 15, in which the at least two unknown state quantities comprise mean and covariance of an echo path, and mean and covariance of nonlinear coefficients.

17. The method according to claim 15, in which the handheld portable communication device comprises a multichannel system and the at least two unknown state quantities comprise channel wise mean and multichannel covariance of a compound quantity formed by merging together the echo path and a nonlinear coefficient.

18. A handheld portable communication device, comprising:

at least one processor; and

at least one memory including computer program code,

the at least one memory and the computer program code configured to, with the at least one processor, cause the apparatus to perform at least the following;

perform an echo suppression process on the handheld portable communication device by performing at least the following;

select, based on an error signal, an expansion basis from a plurality of basis function expansions for operating the estimator;

convert a sample based input signal into a frequency domain multichannel signal using the selected expansion basis, where the frequency domain multichannel signal has a number of portions and where the number of portions is a total number of channels or an expansion order of the selected expansion basis;

estimate posteriors of at least two unknown state quantities;

estimate model parameters comprising at least one noise covariance as a function of the estimated posteriors of the at least two unknown state quantities, the input signal and a microphone signal from a microphone;

compute at least one gain factor for adapting the posteriors of the at least two unknown state quantities in dependence on at least one of the estimated model parameters, the input signal, and the estimated posteriors of the at least two unknown state quantities;

estimate an acoustic echo signal from the multichannel input signal and the adapted posteriors of the at least two unknown state quantities, wherein the acoustic echo signal comprises an estimate of an acoustic echo caused at least by the far-end signal being played by a loudspeaker and resultant output picked up by a microphone; and

generate the error signal by subtracting the estimated echo signal from a frequency-domain version of the microphone signal;

suppress residual linear and nonlinear error caused at least by the acoustic echo by filtering the error signal using a weighting function that depends at least on the error signal, the input signal and directly or indirectly on the adapted posteriors of the unknown state quantities;

to direct the filtered error signal for transmission to a far-end;

in which the handheld portable communication device further comprises;

a radio receiver which provides the far-end signal from which the microphone signal derives; and

a radio transmitter for sending the filtered error signal to a source from which the radio receiver received the far-end signal.

19. The handheld portable communication device according to claim 18, in which the at least two unknown state quantities comprise mean and covariance of the echo path, and mean and covariance of nonlinear coefficients.

20. The handheld portable communication device according to claim 18, in which the handheld portable communication device comprises a multichannel system and the at least two unknown state quantities comprise channel wise mean and multichannel covariance of a compound quantity formed by merging together the echo path and a nonlinear coefficient.