Telephonic communications – Echo cancellation or suppression – Using digital signal processing
Reexamination Certificate
2000-10-25
2003-08-26
Isen, Forester W. (Department: 2644)
Telephonic communications
Echo cancellation or suppression
Using digital signal processing
C379S406010
Reexamination Certificate
active
06611594
ABSTRACT:
FIELD OF THE INVENTION
The invention pertains to network echo cancellation. The invention is particularly suitable for use for hybrid echo cancellation at a two wire to four wire interface in a communications network such as a telephone network.
BACKGROUND OF THE INVENTION
In communications networks, it is possible for an echo of an upstream signal to be coupled onto a downstream signal (directions arbitrary), which echo either corrupts data or decreases the quality of data (e.g., increased noise). For instance, in telephone communications networks, the individual customers usually couple into the main telephone network through a two wire, analog connection in which transmissions in both directions are carried on the same pair of wires (tip and ring). However, the central portion of the network typically is a four wire, digital system in which communications in the upstream and downstream directions are carried on separate wire pairs (i.e., two wires in the receive direction and two wires in the transmit direction).
The interface between the two wire and the four wire portions of the network are a source of echo. The echos typically result because the impedance of the 2-wire facility is imperfectly balanced in the 4-to-2 wire junction causing the incoming signal to be partially reflected over an outgoing path to the source of incoming signals.
The circuitry interfacing the two wire and four wire portions of the network is commonly termed a line interface card. The standard components of a line interface card are well known to persons of skill in the related arts. One of those components is a hybrid cancellation circuit, the function of which is to cancel echo signals in order to improve the quality of service. In short, a hybrid cancellation circuit generates an attenuated version of the original receive signal (that is the source of the echo) and subtracts it from the signal lines that carry the transmit signal onto which the echo has been imposed. These hybrid cancellation circuits utilize an adaptive filter to operate on a supplied signal in a prescribed manner such that a desired output signal is generated.
Typically, adaptive filters generate a transfer function according to an algorithm that includes updating the transfer function characteristic in response to an error signal. In this manner, the filter characteristic is optimized to produce a desired result.
When used in an echo cancellation circuit, an adaptive filter is used to generate an echo path estimate that is updated in response to an error signal. Adaptive echo cancelers have been employed to mitigate the echoes by adjusting the transfer function (impulse response) characteristic of an adaptive filter to generate an estimate of the reflected signal or echo and, then, subtracting it from the outgoing signal. The filter impulse response characteristic and, hence, the echo estimate is updated in response to the outgoing signal for more closely approximating the echo to be cancelled.
Various designs and algorithms for adaptive filters for echo cancellation are well known. Although prior art arrangements of adaptive filters perform satisfactorily in some applications, often it is impossible to simultaneously achieve both sufficiently fast response to changing echo paths and sufficiently high steady-state estimation quality. Consequently a continuing need is to achieve more rapid response to changing conditions while at the same time maintaining adequate steady-state estimation quality.
FIG. 1
is a block diagram illustrating an exemplary network echo cancellation circuit
10
. Input line
12
carries the far-end signal x
n
(i.e., the receive direction signal relative to the transceiver having the echo cancellation circuit). Input line
14
carries the near-end signal w
n
as well as background noise v
n
. The far-end signal x
n
is the source of potential echo.
Block
16
represents the true echo path h
ep
which is coupled into the near-end signal on line
14
as represented by summation node
18
. In accordance with an echo cancellation scheme, the far-end signal x
n
is also input into an echo cancellation circuit
20
which generates an estimated echo path h
n
. The estimated signal path h
n
is generated using a digital adaptive filter algorithm as illustrated by block
124
. The estimated echo signal is then subtracted at subtraction circuit
26
from the signal path y
n
, which includes the true echo h
ep
, the near-end signal w
n
, and the background noise v
n
.
FIG. 1
also illustrates an additional feature known as a double-talk detector
22
which is often incorporated into echo cancellation schemes. A double-talk detector receives at its inputs the incoming far-end signal, x
n
, and the outgoing near-end signal w
n
, including background noise, v
n
, and echo and detects when significant signals exists simultaneously in both directions. When such a condition is detected, the double-talk detector
22
controls the adaptive filter so as to prevent it from adapting to the double-talk situation, which would otherwise temporarily decrease the quality of echo cancellation once the double-talk condition ceases.
Other circumstances in which network echo cancellation may be necessary or at least desirable are numerous and would be well known to persons of skill in the related arts.
Even further, acoustic echo cancellation systems are known in the prior art such as in connection with audio teleconferencing equipment. Particularly, when using a speaker phone, there is a path between the telephone speaker and the telephone microphone through which an echo can be introduced. Particularly, the microphone in a room picks up the sound created in the room. Part of the sound created in the room includes the sound coming from the opposite end of the telephone call that emanates from the telephone speaker. This creates a feedback loop that can result in distortion and, in the worst cases, howling instability.
Many digital adaptive algorithms for echo cancellation are known in the prior art, including Least Mean Squares (LMS) algorithms, Normalized Least Mean Squares (NLMS) algorithms, and Proportionate Normalized Least Mean Squares (PNLMS) among others. Generally, the more effective the particular algorithm at cancelling echo signals, the greater the processing power required to implement the algorithm.
Accordingly, it is an object of the present invention to provide an adaptive algorithm particularly adapted for echo cancellation that is both effective in terms of echo cancellation and efficient in terms of minimizing the complexity of the calculations.
SUMMARY OF THE INVENTION
The invention is a method and apparatus for performing echo cancellation utilizing an efficient and effective adaptive algorithm. The invention is particularly useful in connection with network echo cancellation but is more broadly applicable to any situation where an adaptive estimate of a signal must be generated in real-time. For instance, the invention also may be applied to acoustic echo cancellation such as in teleconferencing systems.
The invention includes a signed regressor proportionate normalized least mean squares algorithm for generating an echo cancellation signal to be subtracted from the echo containing signal. The algorithm is a step-size adaptive algorithm with lower complexity and smaller numerical dynamic range compared to prior art PNLMS algorithms. The present invention achieves such goals by using the signs of the regressor rather than the regressor value itself in the gradient calculation. The resultant convergence rate of the algorithm is about the same as for prior art PNLMS algorithms, but with much lower processing power requirements.
Also disclosed is a double-talk robust version of the algorithm with superior performance particularly when exposed to high level near-end signals.
REFERENCES:
patent: 5951626 (1999-09-01), Duttweiler
Takahashi et al., A New Normalized Signed Regressor LMS Algorithm, Singapore ICCS/ISITA '92, pp.: 1181-1185 vol. 3.
Benesty Jacob
Gaensler Thomas Fritz
Agere Systems Inc.
Harold Jefferey
Isen Forester W.
Synnestvedt & Lechner LLP
LandOfFree
Robust signed regressor PNLMS method and apparatus for... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Robust signed regressor PNLMS method and apparatus for..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Robust signed regressor PNLMS method and apparatus for... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3117932