Telephonic communications – Echo cancellation or suppression – Using digital signal processing
Reexamination Certificate
1999-07-09
2004-08-03
Nguyen, Duc (Department: 2643)
Telephonic communications
Echo cancellation or suppression
Using digital signal processing
C379S406010, C379S406050, C379S406080
Reexamination Certificate
active
06771772
ABSTRACT:
BACKGROUND OF THE INVENTION
In the field of telephone communication, there are echoes of electrical origin. The echo is generally the result of an impedance mismatch at various 2-wire or 4-wire junctions of tandem links forming a telephone connection, or an imperfect decoupling between the 4-wire receive and send ports of 2-to-4-wire junctions, defining thereby the balancing attenuation, or these two phenomena combined. The trouble caused by an echo becomes more important as the echo level is higher and its delay is longer.
FIG. 6
shows schematically an environment in which telephone network echoes arise, and in which the present invention is deployed. A signal is transmitted from “office #1”
610
to “office #2”
640
via telephone lines spanning between them. “Office #1”
610
transmits signals over a 2-wire link
615
to “central office #1”
620
. “Office #2”
640
communicates via a 2-wire link
635
to “central office #2”
630
. “Central office #1”
620
communicates via a 4-wire link
625
to “central office #2”
630
. The 4-wire link
625
has a pair of transmit lines
623
and a pair of receive lines
627
.
“Central office #1”
620
has a 2-to-4-wire junction
625
a
, and “central office #2”
630
has a 2-to-4-wire junction
625
b
. Again, impedance mismatch or imperfect decoupling in the system between “office #1”
610
and “office #2”
640
creates echoes.
To obviate the electrical echo in circuits having a long propagation time, two types of equipment are the subject of CCITT Recommendations: the echo suppressors dealt with in Recommendation G. 164, CCITT Yellow Book edition 1981, vol. III. 1, p. 154 ff., and the echo cancellers or compensators described in Recommendation G.165, CCITT Yellow Book edition 1981, vol. III.1, p. 182 ff. The echo cancellers constitute an improvement with respect to echo suppressors because the echo cancellers act less abruptly and, for example, in the case of a speech signal, they render it possible to reduce the level of the echo signal at the receiving end, while retaining the far-end speech information signal at its normal level.
There are several techniques that can be used to remove network echoes that arise from unbalanced near-end and far-end hybrids. A baseline solution for a cancellation of these echoes involves the combined use of adaptive filtering, non-linear processing of the residual echo, and double-talk detection. See “
ITU Recommendation
”, G.165 cancellers, March, 1993.
The echo path is usually modeled as a finite impulse response (FIR) system having a time-span (of up to 32 msec) and a bulk delay, which is due to the switching equipment and the internal processing delay of the system. If the bulk delay is unknown, then the time-span of the adaptive filter used for echo cancellation must cover both the bulk delay and the echo-span.
SUMMARY OF THE INVENTION
The problem with the combined use of adaptive filtering, non-linear processing, and double-talk detection is that the computational complexity associated with such an implementation is beyond desirable practical limits where the bulk delay is unknown. Also, the set of adaptive coefficients corresponding to delays less than the bulk-delay decreases the signal-to-noise ratio (SNR) at the echo canceller output due to what is known as a gradient noise associated with a gradient-based adaptive method. See S. Haykin, “
Adaptive Filter Theory
”, Prentice-Hall, 3rd Edition, 1996. Finally, convergence is much slower due to a higher number of adaptive taps in such a system.
The present invention is generally directed to an echo canceller having the precise estimate of the bulk-delay and a sufficient number of adaptive coefficients to model the echo path. A baseline solution of the present invention is to use the echo canceller itself to estimate the bulk-delay. After achieving a reasonable level of echo cancellation, the bulk-delay and the echo-span are estimated from the adaptive coefficients, followed by an enhanced mode of operation using the estimated parameters. In the alternative, for longer bulk delays, another embodiment of the present invention uses a subband adaptive system to estimate the bulk-delay and echo-span.
According to one aspect of the present invention, in an echo canceller in a telephone line, an estimator apparatus is used for estimating an echo signal on a telephone line carrying a proximal (sometimes known as a “near-end” signal) signal and an echo signal. The estimator apparatus has a subband adaptive echo canceller that has a set of subband adaptive coefficients. The subband adaptive echo canceller reports the set of subband adaptive coefficients after achieving a reasonable level of convergence.
In one embodiment of the present invention apparatus, the subband adaptive echo canceller comprises a set of subband decomposition filters and a set of adaptive filters. The set of subband decomposition filters decompose a proximal signal and an echo signal into corresponding sets of subband proximal and echo signals. The set of adaptive filters receive the sets of subband proximal and echo signals and adjusts the set of subband adaptive coefficients in response thereto. In the preferred embodiment, the decomposition filters are power-symmetric, infinite impulse response filter banks, and the subband adaptive filters comprise a normalized least mean-square adaptive filter to update the adaptive taps. In another embodiment, other subband processing methods can be used along with other adaptive algorithms. However, other subband decomposition techniques involving finite impulse response filters or wavelets or any other transform-domain approaches can be used.
The estimator apparatus further comprises an estimator receiving the set of subband adaptive coefficients and determines the echo signal bulk-delay as a function of the subband adaptive coefficients. The estimator has a gradient noise reducing unit. The estimator also has an echo path impulse response peak detector to locate the peak of the echo path impulse response. Furthermore, the estimator comprises a peak-to-average ratio measuring unit, a peak-to-average ratio threshold detector, and an echo-span calculation unit. In one embodiment, the echo-span calculation unit allocates more coefficients to the echo tail than to the echo signal preceding the peak delay. In the preferred embodiment, the echo-span calculation unit performs averaging around the bulk-delay peak in order to calculate the optimal number of coefficients to be used.
According to another aspect of the present invention, an echo canceller in a telephone line performs a method for canceling an undesired echo path impulse response having a bulk-delay, peak delay, echo-span, and echo tail in a telephone network carrying a proximal signal and an echo signal. The method comprises the steps of determining a set of adaptive coefficients as a function of the proximal signal and the echo signal and reporting the adaptive coefficients. The present invention method further comprises the steps of calculating the bulk-delay as a function of the adaptive coefficients and reporting the bulk-delay. Furthermore, the present invention method comprises the steps of calculating the echo-span as a function of the adaptive coefficients and reporting the echo-span.
The step of determining the adaptive coefficients in the present invention method comprises the steps of subdividing the proximal signal and the echo signal into sets of subbands, and processing the subdivided proximal signal and the subdivided echo signal, thereby determining the adaptive coefficients. The step of subdividing the proximal signal and the echo signal includes the step of using polyphase power-symmetric infinite impulse response filter banks. The step of processing comprises the step of subband adaptive filtering, which comprises the step of minimizing the normalized least mean-square of the echo signals in each subband.
The step of calculating the bulk-delay includes the step of reducing gradient noise while enhancing the response aroun
Hamilton Brook Smith & Reynolds P.C.
Nguyen Duc
LandOfFree
Method and apparatus for fast network bulk delay and span... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and apparatus for fast network bulk delay and span..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for fast network bulk delay and span... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3294369