Agere Systems Inc.

Lucent Technologies Inc.

Tomas Fritz Gaensler

Jacob Benesty

Steven Leslie Gay

Man Mohan Sondhi

A robust adaptive filter for use in a network echo canceller or other digital signal processing application utilizes a coefficient vector update device that, through the application of fast converging algorithms to a fast impulse response filter yields fast convergence of the adaptive filter's characteristics with the avoidance of divergence due to the onset of double talk. Robustness is also provided, via an adaptive scale non-linearity device which applies an adaptive scale non-linearity to the filter algorithms fed to the fast impulse response filter by the coefficient vector update device, so that the samples of an echo signal to be cancelled which are taken during the onset of double talk can be handled in such a manner that after the double talk detector causes adaptation to cease, the initial, potentially disturbing samples do not cause significant divergence in the filter system.

Adaptive filter for network echo cancellation

A method and apparatus for performing double-talk detection in an acoustic echo canceller in which a detection statistic is advantageously computed based on an estimate of a cross-correlation between the far-end signal and the return signal which has been normalized with use of an estimate of a covariance matrix of the far-end signal. The estimate of the cross-correlation between the far-end signal and the return signal may be further normalized with use of either an estimate of a variance of the return signal or an estimate of a covariance matrix of the return signal. In certain illustrative embodiments of the invention, one or more of these quantities may be estimated based on signal samples sampled over a predetermined time window. And in another illustrative embodiment of the present invention, the coefficients of the adaptive filter employed in the acoustic echo canceller itself are advantageously used to compute the detection statistic. These computations may be performed in either the time domain or the frequency domain. The detection statistic so computed is compared with a predetermined threshold, which threshold may be advantageously fixed at a value close to one, in order to determine whether or not double-talk has occurred.

Method and apparatus for performing double-talk detection in acoustic echo cancellation

A method is presented for dynamic resource allocation in a speech signal echo canceler enabling more efficient echo cancellation and as a result the ability for an Integrated Circuit to handle additional channels than heretofore possible. This is accomplished by applying one or more of three efficiency enhancing strategies. First, no update of coefficients is computed or convolution performed if the power level of the far end speech signal is below a given threshold. Second, convolution is limited to the set of active taps (i.e., taps that account for most of the power in the echo). Third, new coefficients are computed only when the power of the error signal is greater than a given threshold. Lastly, the set of active coefficients is periodically updated. These strategies release computational resources from unnecessary computations and divert them to other channels that may be active.

Dynamic allocation of resources for echo cancellation

An adaptive filter suitable for network echo cancellation and other applications contains a coefficient vector update device for feeding coefficient vector updates to a finite impulse response filter in accordance with fast converging algorithms. A double talk detector is included for causing filter adaptation to cease in the presence of double talk in the system being echo cancelled. The coefficient vector update device utilizes a proportional affine projection algorithm to provide fast convergence of the filter system and improved performance over other filter devices utilizing different fast converging algorithms.

Adaptive filter utilizing proportional affine projection algorithm

An adaptive filter (and a corresponding adaptive filtering method) for use in an echo cancellation environment, comprising a finite impulse response filter having a vector of adaptive coefficients, and an adaptive coefficient vector update module which modifies the vector of adaptive coefficients in response to an error signal and based on a predetermined probability density function for said error signal. In accordance with several illustrative embodiments of the present invention, the probability density function is advantageously non-Gaussian, and has certain advantageous characteristics which provide for robustness in the adaptive algorithm based thereupon.

Growing community of inventors

Chatham, NJ, United States of America

Tomas Fritz Gaensler