Dynamic allocation of resources for echo cancellation

Details Dynamic allocation of resources for echo cancellation Dynamic allocation of resources for echo cancellation

1999-11-02

2004-06-15

Harvey, Minsun Oh (Department: 2644)

Telephonic communications

Echo cancellation or suppression

Disable or inhibit function

C379S406080

Reexamination Certificate

active

06751314

ABSTRACT:

FIELD OF INVENTION
The invention relates generally to improved methods of echo cancellation. Specifically, the invention improves upon the efficiency of echo cancellation by providing a dynamic allocation of echo cancellation resources, thus allowing for increased capacity of multi-channel echo canceler integrated circuits (ICs).
BACKGROUND OF THE INVENTION
Referring to
FIG. 1
, a prior art network, such as a telephone network is shown wherein an echo y(n) is generated when a far end signal x(n) crosses a hybrid local loop telephone termination
100
. The path through termination
100
will hereinafter be referred to as the echo path, and its impulse response will be referred to as the echo path impulse response. ‘n’ is used herein to refer to the index of the samples of the signal.
Methods of echo cancellation in networks are well known. Generally, an echo canceler device
300
receives signal x(n) as an input and computes an estimate ŷ(n) of the echo, y(n). In general, the estimate ŷ(n) at time instant n is determined by multiplying the current and past L-1 samples by L filter coefficients, respectively. The output from the echo canceler
300
is input to an adder
200
which outputs the difference between the estimated echo ŷ(n) and the signal s(n) which consists of the echo y(n) together with the near end signal, v(n), and the near end background noise, w(n).
When the coefficients correspond to those of the impulse response of the true echo path, the echo is canceled. However, the echo path is initially unknown and in addition may change over time. Therefore the filter coefficients are iteratively adjusted to adapt to the echo path's current response. To this end, the output of the adder
200
is fed back to the processor running the adaptive algorithms
375
. In each sample period, new coefficients are computed, resulting in a new estimate of the echo path impulse response.
The DTD (an abbreviation for Double Talk Detector) shown in block
400
, detects when both the far and near end speech are present simultaneously. In this event, the signal y(n) is not strictly the echo of x(n) but rather the combination of the echo, a near end speech signal and some noise. In this event, new coefficients should not be computed for filter
350
since they would be wrong. DTD
400
detects the double talk condition and instructs the echo canceler to continue to compute an estimate of the echo but not to adapt the coefficients.
The estimate of the echo, ŷ(n), is generally derived as follows: Assume a sampling rate of 8 kHz and an echo path impulse response of duration 64 ms. Then the number of coefficients, L, that must be adapted, is 512. Echo canceler device
300
comprises a processor executing a filter
350
and an adaptive algorithm
375
. The aim of the adaptation algorithms is to make the coefficients of the filter
350
equal to the corresponding coefficients of the actual impulse response of the echo path. Filter
350
performs a convolution of x(n) with the estimated impulse response by multiplying each filter coefficient with its corresponding sample of the far end signal, x(n) (the current sample x(n) and the previous L-1 samples), and adding all the L products. The result is the echo estimate ŷ(n).
As currently designed, adaptive echo cancelers compute new coefficients and an echo estimate ŷ(n) at each sampling instant. Currently, network echo canceler integrated circuits (ICs) are designed to process, for example, 32 channels. Thus, in the above example, the echo canceler IC must update 512 coefficients and perform a 512 tap convolution for each of the 32 channels every 1/8000 of a second, or 125 &mgr;s. It is desirable to increase the efficiency of echo cancelers such that an IC can handle a larger number of channels.
SUMMARY OF THE INVENTION
The subject invention provides a new echo cancellation method for voice signals which increases the computational efficiency of present-day echo cancelers such that a single echo canceler IC can handle a larger number of voice channels. This is accomplished by determining the active taps of the impulse response (i.e., those taps that have coefficients with significant magnitudes) and updating the filter coefficients only for the active taps and only when the short term power of the error signal e(n), which is the difference between the estimated echo and the actual echo, is greater than a predetermined threshold. If the error is below the predetermined threshold the filter continues to perform a convolution with the active taps using previously determined coefficients. Furthermore, additional efficiency for line echo cancelers is realized by advantageously performing the coefficient updates and convolution in only those intervals where the short time power of x(n) is above a specified threshold.


REFERENCES:
patent: 5343522 (1994-08-01), Yatrou et al.
patent: 5631900 (1997-05-01), McCaslin et al.
patent: 5940455 (1999-08-01), Ikeda
patent: 5951626 (1999-09-01), Duttweiler
Steven L. Gay, “An Efficient, Fast Converging Adaptive Filter for Network Echo Cancellation,” in Conference Record of the Thirty-Second Asilomar Conference on Signals, Systems & C omputers (Michael B. Matthews ed., 1998).
S. Ikeda and A. Sugiyama, “A Fast Convergence Algorithm for Sparse—Tap Adaptive FIR Filters for an Unknown Number of Multiple Echoes” IEEE pp III-41-44 (1994).
K. Bullington and J. M. Fraser “Engineering aspects of TASI,” Bell System Tech. J., pp. 353-364 (1959).
S. Kawamura and M. Hatori, “A tap selection algorithm for adaptive filters,” IEEE proc. of ICASSP, pp. 2979-2982 (1986).
V. Madisetti, D. Messerschmitt, and N. Nordstrom, “Dynamically reduced complexity implementation of echo cancelers,” IEEE proc. of ICASSP, pp. 1313-1316 (1986).
J. Homer et al., “Quantifying the Effects of Dimension on the Convegence Rate of the LMS Adaptive FIR Estimator,” in IEEE Transactions on Signal Processing vol. 46, No. 10 (1998).
ITU-T Recommendation G. 168 “Transmission Systems and Media, Digital Systems and Networks—Digital Network Echo Cancellers” Section 3.4 (Apr. 1997).

Affiliated with

Also associated with

Rating

Dynamic allocation of resources for echo cancellation does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Dynamic allocation of resources for echo cancellation, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Dynamic allocation of resources for echo cancellation will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-3345219