Double-talk robust fast converging algorithms for network echo cancellation

Abstract

Abstract-There is a need for echo cancelers for echo paths with long impulse responses ( 64 ms). This in turn creates a need for more rapidly converging algorithms in order to meet the specifications for network echo cancelers. Faster convergence, however, in general implies a higher sensitivity to near-end disturbances, especially "double-talk." Recently, a fast converging algorithm has been proposed called proportionate normalized least mean squares (PNLMS) algorithm. This algorithm exploits the sparseness of the echo path and has the advantage that no detection of active coefficients is needed. In this paper we propose a method for making the PNLMS algorithm more robust against double-talk. The slower divergence rate of these algorithms in combination with a standard Geigel double-talk detector improves the performance of a network echo canceler considerably during double-talk. The principle is based on a scaled nonlinearity which is applied to the residual error signal. This results in the robust PNLMS algorithm which diverges much slower than PNLMS and standard NLMS. Tradeoff between convergence and divergence rate is easily adjusted with one parameter and the added complexity is about seven instructions per sample which is less than 0.3% of the total load of a PNLMS algorithm with 512 filter coefficients. A generalization of the robust PNLMS algorithm to a robust proportionate affine projection algorithm (APA) is also presented. It converges very fast, and unlike PNLMS, is not as dependent on the assumption of a sparse echo path response. The complexity of the robust proportionate APA of order two is roughly the same as that of PNLMS.Index Terms-Adaptive filter, double-talk detection, echo cancellation, normalized least mean squares (NLMS), proportionate normalized least mean squares (PNLMS), robustness.

Blind source separation and acoustic echo cancellation: A unified framework

“…Several notable methods have been proposed for this purpose. In [3], [4], [5], and [6] an extra component is employed in the AEC to help avoid filter divergence during DT. This component could be in the form of a DT detector or a step-size controller.…”

Frequency domain adaptive algorithm with nonlinear function of error-to-reference ratio for double-talk robust echo cancellation

“…We can choose a fixed S 1 to suit the specifications of the system in which the algorithm is to be implemented. In similar previous works [11,12], a nonlinearity is applied to the error scaled by a delicately tuned scaling variable. In contrast to this, our approach in Eq.…”

“…This algorithm achieves good performance without time-variant threshold estimators. Although only one [11] is addressed to the frequency domain among the above conventional algorithms [10][11][12][13][14], some of the others can also be made applicable. However, none of them inherently have signal level scalability, so their robustness is only achieved at the cost of peripheral estimators or detectors.…”

“…For example, error nonlinearities have been used to achieve robustness against double-talk [10][11][12]. In [10], the error signal limited by an infinite clipper is used for the adaptation.…”

“…It is also advised in [10] that the adaptation should be frozen whenever the reference input level is less than a predetermined level above the microphone signal, which indicates double-talk. In the other algorithms of this type [11,12], a variable scale factor is introduced to the error nonlinearity. Although a DTD is still required to control the scale factor, the adaptation is not completely frozen during double-talk.…”