CA1220823A - Adaptive filter update gain normalization - Google Patents
Adaptive filter update gain normalizationInfo
- Publication number
- CA1220823A CA1220823A CA000489848A CA489848A CA1220823A CA 1220823 A CA1220823 A CA 1220823A CA 000489848 A CA000489848 A CA 000489848A CA 489848 A CA489848 A CA 489848A CA 1220823 A CA1220823 A CA 1220823A
- Authority
- CA
- Canada
- Prior art keywords
- generating
- power estimate
- value
- gain
- normalized
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B3/00—Line transmission systems
- H04B3/02—Details
- H04B3/20—Reducing echo effects or singing; Opening or closing transmitting path; Conditioning for transmission in one direction or the other
- H04B3/23—Reducing echo effects or singing; Opening or closing transmitting path; Conditioning for transmission in one direction or the other using a replica of transmitted signal in the time domain, e.g. echo cancellers
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03H—IMPEDANCE NETWORKS, e.g. RESONANT CIRCUITS; RESONATORS
- H03H21/00—Adaptive networks
- H03H21/0012—Digital adaptive filters
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03H—IMPEDANCE NETWORKS, e.g. RESONANT CIRCUITS; RESONATORS
- H03H21/00—Adaptive networks
- H03H21/0012—Digital adaptive filters
- H03H2021/007—Computation saving measures; Accelerating measures
- H03H2021/0076—Measures relating to the convergence time
- H03H2021/0078—Measures relating to the convergence time varying the step size
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B3/00—Line transmission systems
- H04B3/02—Details
- H04B3/20—Reducing echo effects or singing; Opening or closing transmitting path; Conditioning for transmission in one direction or the other
- H04B3/23—Reducing echo effects or singing; Opening or closing transmitting path; Conditioning for transmission in one direction or the other using a replica of transmitted signal in the time domain, e.g. echo cancellers
- H04B3/238—Reducing echo effects or singing; Opening or closing transmitting path; Conditioning for transmission in one direction or the other using a replica of transmitted signal in the time domain, e.g. echo cancellers using initial training sequence
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Cable Transmission Systems, Equalization Of Radio And Reduction Of Echo (AREA)
- Filters That Use Time-Delay Elements (AREA)
Abstract
ADAPTIVE FILTER UPDATE
GAIN NORMALIZATION
Abstract Update gain normalization is employed in adaptive filters to control weighting of the filter impulse response updates in order to converge properly to a desired impulse response. Singing, i.e., oscillating, of the filter is overcome by adjusting the update gain when an incoming signal power estimate used to normalize the gain exceeds a prescribed threshold value. In one example, the normalized gain is adjusted to be a fixed value for power estimate values which exceed the threshold. In accordance with another aspect of the invention, a single normalized gain value is used to adjust the update gain in two adaptive filters employed as echo cancelers in a bidirectional voice frequency repeater.
GAIN NORMALIZATION
Abstract Update gain normalization is employed in adaptive filters to control weighting of the filter impulse response updates in order to converge properly to a desired impulse response. Singing, i.e., oscillating, of the filter is overcome by adjusting the update gain when an incoming signal power estimate used to normalize the gain exceeds a prescribed threshold value. In one example, the normalized gain is adjusted to be a fixed value for power estimate values which exceed the threshold. In accordance with another aspect of the invention, a single normalized gain value is used to adjust the update gain in two adaptive filters employed as echo cancelers in a bidirectional voice frequency repeater.
Description
2~
ADAPTIVE FILTER UPDATE
G.Z\IN NO~MALIZATION
Technical Field This invention relates to adaptive filters and, more particularly, to normali~ation of the filter update gain.
Background of the Invention Adaptive filters operate on a signal supplied thereto in accordance with a prescribed criterion to generate a desired output signal. Typically, the filters generate an impulse response characteristic in accordance with an algorithm which includes updating of the impulse response characteristic in response to an error signal. In this way the filter characteristic is optimized to yield the desired result.
It has been found advantageous to normalize the update gain of the adaptive ilter. The normalization tends to ma~e the performance of the filter insensitive to variations in received signal power. In one prior arrangement an average of the squares of input signal sample magnitudes is used to normalize the gain, as described in an article by Mr. D. L. Duttweiler entitled "A
Twelve-Channel Digital Echo Canceler", IEEE ~'ransactions on Communications~ Vol. COM-26, No. 5, May 1978, pp. 647-~53.
Another gain normalization arrangement employing a sum of the squares power estimate is disclosed in U.S.
patent 3,922,505 issued November 25, 1975. These prior arrangements attempt to protect against generating noisy, i.e., uncontrollably large, tap weights by controlling the update gain to be a "low" value for large incoming signal power estimate values.
~ lthough these prior arrangements may perform satisfactorily in some applications, problems arise in others. For example, situations can arise where the open-loop gain of the circuit including the adaptive filter becomes greater than unity. This causes "singing", i.e., ~2%e~ 3 oscillating, to occur in the circuit which, in turn, causes large power estimate values. In such instances, the low gain value generated by the prior gain normaliæation arrangements would slow down the correction of the oscillating condition.
Another problem may arise in applications where two adaptive filters are employed as echo cancelers in a bidirectional voice frequency repeater when talking in both directions of transmission occurs simultaneously~
This occurs, for e~ample, for full duplex data set transmission. When talking occurs in both directions of transmission, the echo cancelers are typically inhibited from updating their impulse response characteristics and proper estimates of the echo signals are not obtained.
15 ~ur-~v ~ ~he I nven~
. , The problems of prior adaptive filter update gain normalization arrangements resulting in singing, i.e., oscillating, and other undesirable results are overcome, in accordance with an aspect of the invention, by controllably adjusting the update gain when a normalization factor, eOg., power estimate of an incoming signal, used to normalize the gain exceeds a prescribed threshold.
In one example, not to be construed as limiting the scope of the invention, the normalized gain is controlled to be a fixed value for power estimates which exceed the prescribed threshold.
In another example, the normalized gain value is adjusted to be a value greater than that which would normally be generated for power estimate values which exceed the prescribed threshold value.
In accordance with another aspect of the invention, a single normalized gain value is used for updating the gain in both the adaptive filters employed as echo cancelers in a bidirectional voice frequency repeater.
,1 ....
- 2a ~
In accordance with an aspect of t:he invention there is provided in an adaptive transversal filter of the type including: means for generating a corresponding one of a plurality of amplitude coefficient signals, said plurality of amplitude coefficient signals forming an impulse response, each of said amplitude coefficient signal generating means including control].able gain means responsive to a gain control signal for acljusting gain used in updating said corresponding amplitude coefficient signal; means for generating said gain control signal including, means for generating a power estimate of a signal incoming to said adpative filter, and means supplied with said power estimate for generating a normalized gain value and for controllably increasing said gain value for power estimate values which are greater than a predeter-mined threshold value, said normalized gain value being said gain control signal.
Brief Description of the Drawing The invention will be more fully understood from , .j the following detailed description of an illustrative embodiment taken in connection with the appended figures in which:
FIG. 1 shows a transmission network employing an adaptive filter as an echo canceler which includes the invention;
FIG. 2 depicts in simplified form details of the echo canceler employed in transmission networks of FIG. 1 and FIG. 6;
FIG. 3 shows details of the power estimator circuit employed in the transmission networks of FIG. 1 and FIG. 6;
FIG. 4 depicts details of the adaptation constant calculator used in the transmission network of FIG. 1;
FIG. 5 is a graphical representation of a normalized gain characteristic, i.e., adaptation constant, illustrating an aspect of the invention;
FIG. 6 shows in simplified block diagram form a bidirectional transmission network employing aspects of the invention;
FIG. 7 depicts details of the adaptation constant calculator used in the embodiment of FIG. 6; and FIGS. 8 and 9 when connected AA AA and sB-BB form a flow chart useful in describing operation of the bidirectional transmission network of FIG. 6 including aspects of the invention.
Detailed Descri~tion FIG. 1 shows in simplified block diagram form a transmission network including an aspect of the invention.
The embodiment of FIG. 1 is intended to couple via hybri~ 101 a bidirectional transmission path or facility 102 to a circuit arrangement including incoming unidirectional transmission path or facility l03 and outgoing unidirectional transmission path or Eacllity lO~.
Such incoming and outgoing circuit arrangements are, for example, four-wire transmission facility or switching machine terminations.
~22~23 Accordingly, incoming signals are supplied via unidirectional transmission path 103 to power estimator 105, an X input oE echo canceler 106 and hybrid 10'1 and, hence, to bidirectional transmission path 102. Outgoing signals frorn bidirectional path 102 are supplied via hybrid 101 to the ~ input of echo canceler 106. An output signal from the E output of echo canceler 106 is supplied to outgoing unidirectional transmission path 10~. Power estimator 105 generates power estimate X(k) which is supplied to adaptation constant calculator 107. ~daptation constant calculator 107 generates a normalized gain value K which has a characteristic in relationship to power estimate X(k), in accordance with an aspect of the invention, so that normalized gain value K is controllably adjusted to have a prescribed value when the value of power estimate X(k) exceeds a predetermined threshold value. Normalized gain value K is supplied to echo canceler 106 where it is used to adjust the normalized update gain for all the taps therein.
It is noted that large values of power estimate X(k) only occur when the transmission network is in a so-called "singing" condition, i.e., oscillating, and, conse~uently, prior gain normalizatlon arrangements would cause a low value of normalized gain value K to be generated which slows down the correction of the oscillating condition. We have determined that as X(k) becomes large in value, the singing condition can be avoided or its length of time minimized by either boosting the normalized gain value to be a value greater than that which would normally be generated for the power estimate values which exceed a prescribed threshold value or setting it at a Ei~ed value. ~\ .speciEic normcllizecl gain characteristic, not to be construed as limiting the scope of the invention, is shown in FIG. 5.
FIG. 2 shows in simplified block diagram form details of an adaptive filter which is advanta~Jeously employed Eor echo canceler 106. Echo canceler 106 is broadly similar to echo cancelers disclosed in U. S.
patents 3,49~,999 and 3,500,000. Also see an article entitled l'A Single-Chip VLSI Echo Canceler'l, by D. L. Duttweiler and Y. S. Chen, The Bell ~ Technical Journal, Vol. 59, No. 2, February 1980, pages 149-159, and an article entitled "Bell's Echo-Killer Chip", IEEE
, October 1980, pages 34-37. In this embodiment echo canceler 106 also includes an arrangement Eor controllably adjusting the normalized update gain of each tap in the adapti~e filter employed in the echo canceler.
Briefly, canceler 106 includes an adjustable signal processor, i.e., adaptive filter having a closed loop error control system which is self-adapting in that it automatically tracks signal variation in an outgoing ~ath.
More specifically, canceler 106 employs echo estimator 201 including an adaptive transversal filter arrangement for synthesizing a linear approximation of the echo, i.e., an echo estimate.
To this end, far end incoming signal X(~c) is usually supplied from a far end talking party over a first transmission path, e.g., path 103 (FIG. 1~, to a first input, namely input X, of echo canceler 106 and therein to an input of echo estisnator 201. Far end signal X(k) may be, for example, a digitally sampled speech signal, where k is an integer identifying the sampling interval. Howe~er, because of an impedance mismatch, for example, in hybrid 101 (FIG. 1), a portion of the hybrid input signal is reflected to the far end signal source as an echo. The echo is supplied from an output of hybrid 101 to the Y input of canceler 109 and therein to a first input of combining networ~c 202~ A second input to cornbinin~
network 202 is a signal estisnate of the echo generated b~
echo estimator 201. The echo estimate is supplied via leacl 203 from an output of echo estimator 201 to the second input of combining network 202. Combining network 202 generates error si~nal ~(k) corresponding to the al~ebraic difference between the echo estimate and the Y input to the echo canceler including the undesirable echo. Error signal E(k) is supp]ied via output E to path 104 (FIG. 1) and to adjustment networks 204-O throu~h 20~-(N-1) in estimator 201.
Estimator 201 includes a so-called tapped delay line comprised of delay units 205-1 through 205-N for realizing desired delays at ~he taps corresponding to convenient Nyquist intervals. Therefore, delayed replicas X(k-1) through ~(k-N) of incoming far end signal X(k) are generated at the corresponding taps. The signal at each tap position, namely X(k-1) through X(k-N) as well as X(k), is adjusted in response to error signal E(k). More particularly, si~nals X(k) through X~k-N) are individually weighted in response to E(k) via acorresponding one of adjustment networks 204-O t~rough 204-N, respectively. Adjustment networks 204-O through 204-N
each include multipliers 207 and 208, and feedback circuit 209. Feedback circuit 209 adjusts the tap weight to a desired value in a manner which will be apparent to those skilled in the art and explained in the above-noted references. In this adjustment, it is important to adjust the individual loop gains K(k) via controllable amplifier 210 to result in a stable system and, in accordance with the invention, to eliminate or minimize "singing", i.e., oscillation. This is realized by adjusting the loop gain K with the value normali~ed in accordance with the characteristic shown in FIG. 5 and generated by the adaptation constant calculator. The weighted replicas of X(k) Erom adjustment networks 204-O
through 204 (N-1) are summed via summing network 211 to generate the echo estimate signal approximating the echo to be canceled. The echo estimate is suppliecl via lead 203 to the second input o~ combinir1g network 202. In one~
embodiment, the number (N) of taps in echo canceler 106 is ~4.
E`IG. 3 shows in simplified block dia~ram for~n .
~;i~;263~23 details of power estimator 105 employed in the embodiments of FIG. 1 and FIG. 6. Power estimator 105 is essentially a low pass filter having a characteristic (1 ) X(k) = ~ ¦ X~k)~ )X(k~
X(k) is, for examplel a digital signal of sampling interval k. Accordingly, X(k) is supplied to rectifier 301 to generate ¦Xtk)¦. Then, ¦X(k)¦ is amplified by factor a in amplifier 302 and ~ ¦X(k)¦ is supplied to one input of adder 303. An output from adder 303 is X(k).
X(k) is supplied via delay unit 304 and amplifier 305 to a second input of adder 303. Delay unit 304 delays X(k~ by a sampling interval to yield X(k-1) and amplifier 305 amplifies X(k-1) to generate (1-~)X(k-1) which is supplied to the second input of adder 303.
FIG . 4 depicts in simplified block diagram form details of adaptation constant calculator 107 employed in the embodiment of FIG. 1. Accordingly, X(k) is supplied to unit 401 which generates LOG(X(k) ) . In turn, unit 402 generates K~LOG(X(k))] and yields K
in accordance with the characteristic shown in FIG. 5.
This is realizable by using a table look-up. ~hat is to say, K[LOG(X(k))] is the address in a read only memory (ROM) for the value of K. Alternatively, the value of X(k) can be used directly to address a RO~ to generate a corresponding value of K.
FIG. 5 shows a normalized gain characteristic which may be employed, in accordance with an aspect of the invention, to minimize l'singing", i.e., oscillating, in the embodiments of FIGs. 1 and 6. Heretofore, as X(k) or, in this example, LOG [~(k)] increased, the normalized gain K would continue to decrease. ~rhi.s 3S resulted in problems because large values of X(k) typically occur only when the circuit is singing, i.e~, oscillating, and the "low" value of the normalizecl gain -- 8 ~
slows down the recovery from instability and osci:LlatingO
As shown in FIG. 5, the normali~ed qain K for 'llarge"
values of X(k) i5 boosted so that it remains, in this example, at a Eixed predetermined valueO It is again noted that the gain boost for large values of X(~) does not affect the normal voice frequency signals because their average power levels are not as large as those resulting during singing intervals. Typical values of normalized gain K for XMAX(k) in dBmo are as follows:
MAX( ) K
> ~ O
> -50.83 > -44.54 .5 > -38.65 .25 > -35 .125 > -30.67 .0625 > -27.64 .03125 > -23.81 .015625 > -21.02 .0078125 > -17.4 .0078125 > -14.72 .0078125 > -11.2 .0078125 > -8.57 .007~125 > -5.09 .OG78125 > -2.48 .0078125 > .98 .0078125 > 3.174 .0078125 FIG. 6 shows in simplified block diagram form details of an automati.cally adaptive bidirectional voice fre~uency repeater employing aspects o~ the invention. rrhe repeater shown in FIG, 6 is intende(l to enhance ~oice frequency signals being transmitted on a 2-wire transmission path or facility.
Accordingly, shown in E`IG. 6 are hybrid ~2~%3 9 _ circuits 601 and 602 which are adapted ~or coupling signals to and from the repeater to bidirectional transmission paths A and B, respectively. Hybrids 601 and 602 may be any of the hybrid arrangements known in the art for coupling signals from a bidirectional 2-wire transmission path or facility to incoming and outgoing unidirectional transmission paths or facilities. In one example~ hybrids 601 and 602 are of the electronic type.
An incoming signal output from hybrid 601 is supplied to an encoder input of CODEC 603. In this example, the encoder of CODEC 603 converts the analog input signal into an 8-bit ~-law PC~ digital signal, in well-known fashion. The 8-bit ~-law PC~ digital signal from CODEC 603 is supplied to a Y input of echo canceler 604.
Output E from echo canceler 604 is supplied via equalizer 605 and gain unit 606 to an X input of echo canceler 607, to a decoder input of CODEC 608 and to power estimator 609. The decoder of CODEC 608 converts the 8-bit ~-law PCM output from echo canceler 604 into an analog output signal, in well~known fashlon. The analog output from CODEC 608 is supplied to hybrid 602 and, in turn, to bidirectional 2-wire facility B.
An analog signal from bidirectional 2-wire facility B is supplied via hybrid 602 to an encoder input 25 of CODEC 60a. In turn, CODEC 608 converts the analog input signal, in this example, into a digital ~-law PCM
signal which is supplied to the Y input of echo canceler 607. Output E from echo canceler 607 is supplied via e~ualizer 610 and gain unit 611 to the X input of echo 30 canceler 60~, to a decoder input of CODEC 603 and to power estimator 612. CODEC 603 converts the 8-bit ~-law output E
from echo canceler 607 into an analocl sigr)cll, in well-know fashion. The analog signal from C~UEC 603 is supplied via hybrid 601 to bidirectional 2-wire facility A~
Echo canceler 60~ is employed to cancel the echo signal or reflected signal resulting in the signal transmission from bidirectional 2-wire facilit~ A because ~2~ 3 of the signal incoming to the repeater from bidirectional 2-wire facility B. Similarl~, echo canceler 607 i5 employed to cancel the echo or re~lected signal resulting in the si~nal transmission from bidirectional 2-wire facility B because of the signal incoming to the repeater from bidirectional facility A. Details Oe an echo canceler which may be employed for cancelers 604 and 607 are shown in FIG. 2 and have been described above.
In this example, equalizers 119 and 124 each include an eighth order finite impulse response filter of a type known in the art.
Power estimators 609 and 612 are identical to power estimator 105, details of which are shown in FIG. 3 and have been described above. Accordingly, power estimator 609 generates power estimate XA(k) of XA(k), in accordance with equation (1), which is supplied to adaptation constant calculator 613. Similarly, power estimator 612 generates power estimate XB(k) of XB(k), in accordance with equation (1), which is also supplied to adaptation constant calculator 613.
Adaptation constant calculator 613 generates nortnalized gain value K in a manner described below which is supplied to both of echo cancelers 604 and 607 to set the update gain for all the taps in each of the transversal filters used therein. In accordance with an aspect of the invention, adaptation constant calculator 613 selects between power estimate XB(~) and XA(k) in accordance with a prescribed criterion for generating normalized gain value K which is then used to adjust the update gain in each oE echo cancelers 604 and 607.
Consequently, each of echo cancelers 604 and 607 is allowed to adapt to a clesired impulse response characteristic even when there is "talking" in both directions oE transmission.
Heretofore, talkin~ in each direction o~ transmlssion would 3S have caused both echo cancelers to be inhibited frotn being updated.
Use of the same update gain value K for both directions of transmission also helps when singing, i.e., oscillating, is occurring to stabilize the repeater. That is, if one of the power estimates XA(k) or XB(k) is large relative to normally expected voice 5 band signals, an unstable system is indlicated and the update gain is controllably adjusted, as described above in relationship to the embodiment of FIG~ 1~ for both echo cancelers 604 and 607 and the repeater is rapidly stabilized. That is to say, the normalized gain value K is 10 either boosted or set at a fixed value for values of XMAX(k) which exceed a predetermined threshold.
FIG. 7 shows in block diagram form details of adaptation constant calculator 613 used in the embodiment of FIGo 60 Accordingly, power estimate XA(k) is 15 supplied to one input of comparator 701 and to input A of digital selector 702 while power estimate XB(k) is supplied to a second input of comparator 701 and to input B of selector 702. Comparator 701 generates an output representative of a logical 1 or true signal when 20 XA(k) > XB(k). This output from comparator 701 is supplied to the select input of digital selector 702. Selector 702 is operative to normally supply gB(k) to its output except when XA(k) is selected in response to the true signal from 25 comparator 7Q1. Simply s~ated, the output from selector 702 is the maximum of XA(k) and XB(k)' i.e., XMAx(k).
The ~lAX(k) signal is supplied to log unit 703 which generates its logarithm, namely, 30 LOG[XM~x(k)]. The LOG[XMAx(k)]
value is used as an address to access a look-up table to generate normalized update gain value K, in this example, in accordance with the characteristic shown in FIG. 5. q'he look-up table is stored, for example, in a read only memory 35 (ROM).
Although the embodiments of the present invention have been described in simplified block diagram form, ~2Z0~23 preferred embodiments are realized by appropriately programming a digital signal processor to obtain the equalizer, gain unit, power estimator, adaptation constant calculator and echo canceler functions. One such digital signal processor (DSP) unit is manufactured by American Telephone & Telegraph Co. and is described in several articles in The Bell System _~c ~ l, Vol. 60, No. 7, Part 2, dated September 1981. A prior known repeater employing a digital signal processor is broadly described in an article entitled, "Digital Signal Solves Hybrid Balance Puzzle", Tele~hone Engineer & ~ , August 1, 1983, pages 39-46.
Accordingly, the equalizer, gain unit, power estimator, adaptation constant calculator and echo canceler functions of the bidirectional repeater of FIG. 6 are realized by programming a DSP unit. Operation of the repeater, in accordance with aspects of the invention, is described in conjunction with the digital program flow chart of FIGS. 8 and 9. FIGS. ~ and 9 when connected AA-AA and BB-BB form a flow chart of a program routine for realizing the above-noted repeater functions. Thus, the DSP is initialized in well-known fashion via 801.
Operational block 802 performs the equalization and gain functions on error signal eB(k) from echo canceler 607 to form the A side signal XA(k). Then, operational block 803 forms the power estimate XA(k) of XA(k).
Operational block 804 causes the generation of echo canceler 604 error signal eA(k). This is achieved by employing a known echo canceler algorithm, for example, the one disclosed in the article entitled "A Single-Chip VLSI
Echo Canceler" cited above. Operational block 805 causes the equalization and gain functions to be performed on the side A error signal eA(k) to form side B output signal XB(k). Then, operational block 806 causes power estimate XB(k) to be formed of side B output signal XB(k). Operational block 807 causes echo canceler 607 error signal output eg(k) to be computed. This 2~ 3 computation is realized in the same fashion as Eor eA(b) in operational block 804. Operational block 810 causes MAX(k) MAX of XA(k), XB(b) to be found. Then, operational block 809 converts XMAX(k) to logarithmic form. Operational block 810 generates an address for the adaptat:ion constant, iOe~, normalized gain value K using LOG(XMAX(k)).
Operational block 811 causes echo canceler 604 to be updated using adaptation constant K. Similarly, operational block 812 causes echo canceler 607 to be updated also using adaptation constant K.
ADAPTIVE FILTER UPDATE
G.Z\IN NO~MALIZATION
Technical Field This invention relates to adaptive filters and, more particularly, to normali~ation of the filter update gain.
Background of the Invention Adaptive filters operate on a signal supplied thereto in accordance with a prescribed criterion to generate a desired output signal. Typically, the filters generate an impulse response characteristic in accordance with an algorithm which includes updating of the impulse response characteristic in response to an error signal. In this way the filter characteristic is optimized to yield the desired result.
It has been found advantageous to normalize the update gain of the adaptive ilter. The normalization tends to ma~e the performance of the filter insensitive to variations in received signal power. In one prior arrangement an average of the squares of input signal sample magnitudes is used to normalize the gain, as described in an article by Mr. D. L. Duttweiler entitled "A
Twelve-Channel Digital Echo Canceler", IEEE ~'ransactions on Communications~ Vol. COM-26, No. 5, May 1978, pp. 647-~53.
Another gain normalization arrangement employing a sum of the squares power estimate is disclosed in U.S.
patent 3,922,505 issued November 25, 1975. These prior arrangements attempt to protect against generating noisy, i.e., uncontrollably large, tap weights by controlling the update gain to be a "low" value for large incoming signal power estimate values.
~ lthough these prior arrangements may perform satisfactorily in some applications, problems arise in others. For example, situations can arise where the open-loop gain of the circuit including the adaptive filter becomes greater than unity. This causes "singing", i.e., ~2%e~ 3 oscillating, to occur in the circuit which, in turn, causes large power estimate values. In such instances, the low gain value generated by the prior gain normaliæation arrangements would slow down the correction of the oscillating condition.
Another problem may arise in applications where two adaptive filters are employed as echo cancelers in a bidirectional voice frequency repeater when talking in both directions of transmission occurs simultaneously~
This occurs, for e~ample, for full duplex data set transmission. When talking occurs in both directions of transmission, the echo cancelers are typically inhibited from updating their impulse response characteristics and proper estimates of the echo signals are not obtained.
15 ~ur-~v ~ ~he I nven~
. , The problems of prior adaptive filter update gain normalization arrangements resulting in singing, i.e., oscillating, and other undesirable results are overcome, in accordance with an aspect of the invention, by controllably adjusting the update gain when a normalization factor, eOg., power estimate of an incoming signal, used to normalize the gain exceeds a prescribed threshold.
In one example, not to be construed as limiting the scope of the invention, the normalized gain is controlled to be a fixed value for power estimates which exceed the prescribed threshold.
In another example, the normalized gain value is adjusted to be a value greater than that which would normally be generated for power estimate values which exceed the prescribed threshold value.
In accordance with another aspect of the invention, a single normalized gain value is used for updating the gain in both the adaptive filters employed as echo cancelers in a bidirectional voice frequency repeater.
,1 ....
- 2a ~
In accordance with an aspect of t:he invention there is provided in an adaptive transversal filter of the type including: means for generating a corresponding one of a plurality of amplitude coefficient signals, said plurality of amplitude coefficient signals forming an impulse response, each of said amplitude coefficient signal generating means including control].able gain means responsive to a gain control signal for acljusting gain used in updating said corresponding amplitude coefficient signal; means for generating said gain control signal including, means for generating a power estimate of a signal incoming to said adpative filter, and means supplied with said power estimate for generating a normalized gain value and for controllably increasing said gain value for power estimate values which are greater than a predeter-mined threshold value, said normalized gain value being said gain control signal.
Brief Description of the Drawing The invention will be more fully understood from , .j the following detailed description of an illustrative embodiment taken in connection with the appended figures in which:
FIG. 1 shows a transmission network employing an adaptive filter as an echo canceler which includes the invention;
FIG. 2 depicts in simplified form details of the echo canceler employed in transmission networks of FIG. 1 and FIG. 6;
FIG. 3 shows details of the power estimator circuit employed in the transmission networks of FIG. 1 and FIG. 6;
FIG. 4 depicts details of the adaptation constant calculator used in the transmission network of FIG. 1;
FIG. 5 is a graphical representation of a normalized gain characteristic, i.e., adaptation constant, illustrating an aspect of the invention;
FIG. 6 shows in simplified block diagram form a bidirectional transmission network employing aspects of the invention;
FIG. 7 depicts details of the adaptation constant calculator used in the embodiment of FIG. 6; and FIGS. 8 and 9 when connected AA AA and sB-BB form a flow chart useful in describing operation of the bidirectional transmission network of FIG. 6 including aspects of the invention.
Detailed Descri~tion FIG. 1 shows in simplified block diagram form a transmission network including an aspect of the invention.
The embodiment of FIG. 1 is intended to couple via hybri~ 101 a bidirectional transmission path or facility 102 to a circuit arrangement including incoming unidirectional transmission path or facility l03 and outgoing unidirectional transmission path or Eacllity lO~.
Such incoming and outgoing circuit arrangements are, for example, four-wire transmission facility or switching machine terminations.
~22~23 Accordingly, incoming signals are supplied via unidirectional transmission path 103 to power estimator 105, an X input oE echo canceler 106 and hybrid 10'1 and, hence, to bidirectional transmission path 102. Outgoing signals frorn bidirectional path 102 are supplied via hybrid 101 to the ~ input of echo canceler 106. An output signal from the E output of echo canceler 106 is supplied to outgoing unidirectional transmission path 10~. Power estimator 105 generates power estimate X(k) which is supplied to adaptation constant calculator 107. ~daptation constant calculator 107 generates a normalized gain value K which has a characteristic in relationship to power estimate X(k), in accordance with an aspect of the invention, so that normalized gain value K is controllably adjusted to have a prescribed value when the value of power estimate X(k) exceeds a predetermined threshold value. Normalized gain value K is supplied to echo canceler 106 where it is used to adjust the normalized update gain for all the taps therein.
It is noted that large values of power estimate X(k) only occur when the transmission network is in a so-called "singing" condition, i.e., oscillating, and, conse~uently, prior gain normalizatlon arrangements would cause a low value of normalized gain value K to be generated which slows down the correction of the oscillating condition. We have determined that as X(k) becomes large in value, the singing condition can be avoided or its length of time minimized by either boosting the normalized gain value to be a value greater than that which would normally be generated for the power estimate values which exceed a prescribed threshold value or setting it at a Ei~ed value. ~\ .speciEic normcllizecl gain characteristic, not to be construed as limiting the scope of the invention, is shown in FIG. 5.
FIG. 2 shows in simplified block diagram form details of an adaptive filter which is advanta~Jeously employed Eor echo canceler 106. Echo canceler 106 is broadly similar to echo cancelers disclosed in U. S.
patents 3,49~,999 and 3,500,000. Also see an article entitled l'A Single-Chip VLSI Echo Canceler'l, by D. L. Duttweiler and Y. S. Chen, The Bell ~ Technical Journal, Vol. 59, No. 2, February 1980, pages 149-159, and an article entitled "Bell's Echo-Killer Chip", IEEE
, October 1980, pages 34-37. In this embodiment echo canceler 106 also includes an arrangement Eor controllably adjusting the normalized update gain of each tap in the adapti~e filter employed in the echo canceler.
Briefly, canceler 106 includes an adjustable signal processor, i.e., adaptive filter having a closed loop error control system which is self-adapting in that it automatically tracks signal variation in an outgoing ~ath.
More specifically, canceler 106 employs echo estimator 201 including an adaptive transversal filter arrangement for synthesizing a linear approximation of the echo, i.e., an echo estimate.
To this end, far end incoming signal X(~c) is usually supplied from a far end talking party over a first transmission path, e.g., path 103 (FIG. 1~, to a first input, namely input X, of echo canceler 106 and therein to an input of echo estisnator 201. Far end signal X(k) may be, for example, a digitally sampled speech signal, where k is an integer identifying the sampling interval. Howe~er, because of an impedance mismatch, for example, in hybrid 101 (FIG. 1), a portion of the hybrid input signal is reflected to the far end signal source as an echo. The echo is supplied from an output of hybrid 101 to the Y input of canceler 109 and therein to a first input of combining networ~c 202~ A second input to cornbinin~
network 202 is a signal estisnate of the echo generated b~
echo estimator 201. The echo estimate is supplied via leacl 203 from an output of echo estimator 201 to the second input of combining network 202. Combining network 202 generates error si~nal ~(k) corresponding to the al~ebraic difference between the echo estimate and the Y input to the echo canceler including the undesirable echo. Error signal E(k) is supp]ied via output E to path 104 (FIG. 1) and to adjustment networks 204-O throu~h 20~-(N-1) in estimator 201.
Estimator 201 includes a so-called tapped delay line comprised of delay units 205-1 through 205-N for realizing desired delays at ~he taps corresponding to convenient Nyquist intervals. Therefore, delayed replicas X(k-1) through ~(k-N) of incoming far end signal X(k) are generated at the corresponding taps. The signal at each tap position, namely X(k-1) through X(k-N) as well as X(k), is adjusted in response to error signal E(k). More particularly, si~nals X(k) through X~k-N) are individually weighted in response to E(k) via acorresponding one of adjustment networks 204-O t~rough 204-N, respectively. Adjustment networks 204-O through 204-N
each include multipliers 207 and 208, and feedback circuit 209. Feedback circuit 209 adjusts the tap weight to a desired value in a manner which will be apparent to those skilled in the art and explained in the above-noted references. In this adjustment, it is important to adjust the individual loop gains K(k) via controllable amplifier 210 to result in a stable system and, in accordance with the invention, to eliminate or minimize "singing", i.e., oscillation. This is realized by adjusting the loop gain K with the value normali~ed in accordance with the characteristic shown in FIG. 5 and generated by the adaptation constant calculator. The weighted replicas of X(k) Erom adjustment networks 204-O
through 204 (N-1) are summed via summing network 211 to generate the echo estimate signal approximating the echo to be canceled. The echo estimate is suppliecl via lead 203 to the second input o~ combinir1g network 202. In one~
embodiment, the number (N) of taps in echo canceler 106 is ~4.
E`IG. 3 shows in simplified block dia~ram for~n .
~;i~;263~23 details of power estimator 105 employed in the embodiments of FIG. 1 and FIG. 6. Power estimator 105 is essentially a low pass filter having a characteristic (1 ) X(k) = ~ ¦ X~k)~ )X(k~
X(k) is, for examplel a digital signal of sampling interval k. Accordingly, X(k) is supplied to rectifier 301 to generate ¦Xtk)¦. Then, ¦X(k)¦ is amplified by factor a in amplifier 302 and ~ ¦X(k)¦ is supplied to one input of adder 303. An output from adder 303 is X(k).
X(k) is supplied via delay unit 304 and amplifier 305 to a second input of adder 303. Delay unit 304 delays X(k~ by a sampling interval to yield X(k-1) and amplifier 305 amplifies X(k-1) to generate (1-~)X(k-1) which is supplied to the second input of adder 303.
FIG . 4 depicts in simplified block diagram form details of adaptation constant calculator 107 employed in the embodiment of FIG. 1. Accordingly, X(k) is supplied to unit 401 which generates LOG(X(k) ) . In turn, unit 402 generates K~LOG(X(k))] and yields K
in accordance with the characteristic shown in FIG. 5.
This is realizable by using a table look-up. ~hat is to say, K[LOG(X(k))] is the address in a read only memory (ROM) for the value of K. Alternatively, the value of X(k) can be used directly to address a RO~ to generate a corresponding value of K.
FIG. 5 shows a normalized gain characteristic which may be employed, in accordance with an aspect of the invention, to minimize l'singing", i.e., oscillating, in the embodiments of FIGs. 1 and 6. Heretofore, as X(k) or, in this example, LOG [~(k)] increased, the normalized gain K would continue to decrease. ~rhi.s 3S resulted in problems because large values of X(k) typically occur only when the circuit is singing, i.e~, oscillating, and the "low" value of the normalizecl gain -- 8 ~
slows down the recovery from instability and osci:LlatingO
As shown in FIG. 5, the normali~ed qain K for 'llarge"
values of X(k) i5 boosted so that it remains, in this example, at a Eixed predetermined valueO It is again noted that the gain boost for large values of X(~) does not affect the normal voice frequency signals because their average power levels are not as large as those resulting during singing intervals. Typical values of normalized gain K for XMAX(k) in dBmo are as follows:
MAX( ) K
> ~ O
> -50.83 > -44.54 .5 > -38.65 .25 > -35 .125 > -30.67 .0625 > -27.64 .03125 > -23.81 .015625 > -21.02 .0078125 > -17.4 .0078125 > -14.72 .0078125 > -11.2 .0078125 > -8.57 .007~125 > -5.09 .OG78125 > -2.48 .0078125 > .98 .0078125 > 3.174 .0078125 FIG. 6 shows in simplified block diagram form details of an automati.cally adaptive bidirectional voice fre~uency repeater employing aspects o~ the invention. rrhe repeater shown in FIG, 6 is intende(l to enhance ~oice frequency signals being transmitted on a 2-wire transmission path or facility.
Accordingly, shown in E`IG. 6 are hybrid ~2~%3 9 _ circuits 601 and 602 which are adapted ~or coupling signals to and from the repeater to bidirectional transmission paths A and B, respectively. Hybrids 601 and 602 may be any of the hybrid arrangements known in the art for coupling signals from a bidirectional 2-wire transmission path or facility to incoming and outgoing unidirectional transmission paths or facilities. In one example~ hybrids 601 and 602 are of the electronic type.
An incoming signal output from hybrid 601 is supplied to an encoder input of CODEC 603. In this example, the encoder of CODEC 603 converts the analog input signal into an 8-bit ~-law PC~ digital signal, in well-known fashion. The 8-bit ~-law PC~ digital signal from CODEC 603 is supplied to a Y input of echo canceler 604.
Output E from echo canceler 604 is supplied via equalizer 605 and gain unit 606 to an X input of echo canceler 607, to a decoder input of CODEC 608 and to power estimator 609. The decoder of CODEC 608 converts the 8-bit ~-law PCM output from echo canceler 604 into an analog output signal, in well~known fashlon. The analog output from CODEC 608 is supplied to hybrid 602 and, in turn, to bidirectional 2-wire facility B.
An analog signal from bidirectional 2-wire facility B is supplied via hybrid 602 to an encoder input 25 of CODEC 60a. In turn, CODEC 608 converts the analog input signal, in this example, into a digital ~-law PCM
signal which is supplied to the Y input of echo canceler 607. Output E from echo canceler 607 is supplied via e~ualizer 610 and gain unit 611 to the X input of echo 30 canceler 60~, to a decoder input of CODEC 603 and to power estimator 612. CODEC 603 converts the 8-bit ~-law output E
from echo canceler 607 into an analocl sigr)cll, in well-know fashion. The analog signal from C~UEC 603 is supplied via hybrid 601 to bidirectional 2-wire facility A~
Echo canceler 60~ is employed to cancel the echo signal or reflected signal resulting in the signal transmission from bidirectional 2-wire facilit~ A because ~2~ 3 of the signal incoming to the repeater from bidirectional 2-wire facility B. Similarl~, echo canceler 607 i5 employed to cancel the echo or re~lected signal resulting in the si~nal transmission from bidirectional 2-wire facility B because of the signal incoming to the repeater from bidirectional facility A. Details Oe an echo canceler which may be employed for cancelers 604 and 607 are shown in FIG. 2 and have been described above.
In this example, equalizers 119 and 124 each include an eighth order finite impulse response filter of a type known in the art.
Power estimators 609 and 612 are identical to power estimator 105, details of which are shown in FIG. 3 and have been described above. Accordingly, power estimator 609 generates power estimate XA(k) of XA(k), in accordance with equation (1), which is supplied to adaptation constant calculator 613. Similarly, power estimator 612 generates power estimate XB(k) of XB(k), in accordance with equation (1), which is also supplied to adaptation constant calculator 613.
Adaptation constant calculator 613 generates nortnalized gain value K in a manner described below which is supplied to both of echo cancelers 604 and 607 to set the update gain for all the taps in each of the transversal filters used therein. In accordance with an aspect of the invention, adaptation constant calculator 613 selects between power estimate XB(~) and XA(k) in accordance with a prescribed criterion for generating normalized gain value K which is then used to adjust the update gain in each oE echo cancelers 604 and 607.
Consequently, each of echo cancelers 604 and 607 is allowed to adapt to a clesired impulse response characteristic even when there is "talking" in both directions oE transmission.
Heretofore, talkin~ in each direction o~ transmlssion would 3S have caused both echo cancelers to be inhibited frotn being updated.
Use of the same update gain value K for both directions of transmission also helps when singing, i.e., oscillating, is occurring to stabilize the repeater. That is, if one of the power estimates XA(k) or XB(k) is large relative to normally expected voice 5 band signals, an unstable system is indlicated and the update gain is controllably adjusted, as described above in relationship to the embodiment of FIG~ 1~ for both echo cancelers 604 and 607 and the repeater is rapidly stabilized. That is to say, the normalized gain value K is 10 either boosted or set at a fixed value for values of XMAX(k) which exceed a predetermined threshold.
FIG. 7 shows in block diagram form details of adaptation constant calculator 613 used in the embodiment of FIGo 60 Accordingly, power estimate XA(k) is 15 supplied to one input of comparator 701 and to input A of digital selector 702 while power estimate XB(k) is supplied to a second input of comparator 701 and to input B of selector 702. Comparator 701 generates an output representative of a logical 1 or true signal when 20 XA(k) > XB(k). This output from comparator 701 is supplied to the select input of digital selector 702. Selector 702 is operative to normally supply gB(k) to its output except when XA(k) is selected in response to the true signal from 25 comparator 7Q1. Simply s~ated, the output from selector 702 is the maximum of XA(k) and XB(k)' i.e., XMAx(k).
The ~lAX(k) signal is supplied to log unit 703 which generates its logarithm, namely, 30 LOG[XM~x(k)]. The LOG[XMAx(k)]
value is used as an address to access a look-up table to generate normalized update gain value K, in this example, in accordance with the characteristic shown in FIG. 5. q'he look-up table is stored, for example, in a read only memory 35 (ROM).
Although the embodiments of the present invention have been described in simplified block diagram form, ~2Z0~23 preferred embodiments are realized by appropriately programming a digital signal processor to obtain the equalizer, gain unit, power estimator, adaptation constant calculator and echo canceler functions. One such digital signal processor (DSP) unit is manufactured by American Telephone & Telegraph Co. and is described in several articles in The Bell System _~c ~ l, Vol. 60, No. 7, Part 2, dated September 1981. A prior known repeater employing a digital signal processor is broadly described in an article entitled, "Digital Signal Solves Hybrid Balance Puzzle", Tele~hone Engineer & ~ , August 1, 1983, pages 39-46.
Accordingly, the equalizer, gain unit, power estimator, adaptation constant calculator and echo canceler functions of the bidirectional repeater of FIG. 6 are realized by programming a DSP unit. Operation of the repeater, in accordance with aspects of the invention, is described in conjunction with the digital program flow chart of FIGS. 8 and 9. FIGS. ~ and 9 when connected AA-AA and BB-BB form a flow chart of a program routine for realizing the above-noted repeater functions. Thus, the DSP is initialized in well-known fashion via 801.
Operational block 802 performs the equalization and gain functions on error signal eB(k) from echo canceler 607 to form the A side signal XA(k). Then, operational block 803 forms the power estimate XA(k) of XA(k).
Operational block 804 causes the generation of echo canceler 604 error signal eA(k). This is achieved by employing a known echo canceler algorithm, for example, the one disclosed in the article entitled "A Single-Chip VLSI
Echo Canceler" cited above. Operational block 805 causes the equalization and gain functions to be performed on the side A error signal eA(k) to form side B output signal XB(k). Then, operational block 806 causes power estimate XB(k) to be formed of side B output signal XB(k). Operational block 807 causes echo canceler 607 error signal output eg(k) to be computed. This 2~ 3 computation is realized in the same fashion as Eor eA(b) in operational block 804. Operational block 810 causes MAX(k) MAX of XA(k), XB(b) to be found. Then, operational block 809 converts XMAX(k) to logarithmic form. Operational block 810 generates an address for the adaptat:ion constant, iOe~, normalized gain value K using LOG(XMAX(k)).
Operational block 811 causes echo canceler 604 to be updated using adaptation constant K. Similarly, operational block 812 causes echo canceler 607 to be updated also using adaptation constant K.
Claims (19)
1. In an adaptive transversal filter of the type including:
means for generating a corresponding one of a plurality of amplitude coefficient signals, said plurality of amplitude coefficient signals forming an impulse response, each of said amplitude coefficient signal generating means including controllable gain means responsive to a gain control signal for adjusting gain used in updating said corresponding amplitude coefficient signal;
means for generating said gain control signal including, means for generating a power estimate of a signal incoming to said adpative filter, and means supplied with said power estimate for generating a normalized gain value and for controllably increasing said gain value for power estimate values which are greater than a predetermined threshold value, said normalized gain value being said gain control signal.
means for generating a corresponding one of a plurality of amplitude coefficient signals, said plurality of amplitude coefficient signals forming an impulse response, each of said amplitude coefficient signal generating means including controllable gain means responsive to a gain control signal for adjusting gain used in updating said corresponding amplitude coefficient signal;
means for generating said gain control signal including, means for generating a power estimate of a signal incoming to said adpative filter, and means supplied with said power estimate for generating a normalized gain value and for controllably increasing said gain value for power estimate values which are greater than a predetermined threshold value, said normalized gain value being said gain control signal.
2. The inventon as defined in claim l wherein said gain value is increased to be larger than that which would normally be generated for said power estimate values which exceed said predetermined threshold.
3. The invention as defined in claim 2 wherein said power estimate generating means comprises a low pass filter.
4. The invention as defined in claim 3 wherein said normalized gain value generating means includes means for storing predetermined normalized gain values cor-responding to values of said power estimate and means responsive to said power estimate value for addressing said storing means to output a corresponding normalized gain value.
5. The invention as defined in claim 4 wherein said transversal filter is employed in an echo canceler.
6. In an adaptive transversal filter of the type including:
means for generating a corresponding one of a plurality of amplitude coefficient signals, said plurality of amplitude coefficient signals forming an impulse response, each of said amplitude coefficient signal generating means including controllable gain means responsive to a gain control signal for adjusting gain used in updating said corresponding amplitude coefficient signal;
means for generating said gain control signal including, means for generating a power estimate of a signal incoming to said adaptive filter, and means supplied with said power estimate for generating a normalized gain value and for limiting said normalized gain value to a predetermined value for power estimate values which are greater than a predetermined threshold value, said normalized gain value being said gain control signal.
means for generating a corresponding one of a plurality of amplitude coefficient signals, said plurality of amplitude coefficient signals forming an impulse response, each of said amplitude coefficient signal generating means including controllable gain means responsive to a gain control signal for adjusting gain used in updating said corresponding amplitude coefficient signal;
means for generating said gain control signal including, means for generating a power estimate of a signal incoming to said adaptive filter, and means supplied with said power estimate for generating a normalized gain value and for limiting said normalized gain value to a predetermined value for power estimate values which are greater than a predetermined threshold value, said normalized gain value being said gain control signal.
7. The invention as defined in claim 6 wherein said normalized gain value is limited to a fixed value for power estimate values which are greater than said threshold value.
8. The invention as defined in claim 7 wherein said power estimate generating means comprises a low pass filter.
g. The invention as defined in claim 7 wherein said normalized gain value generating and limiting means includes means for storing predetermined normalized gain values and means responsive to said power estimate value for addressing said storing means to output a corresponding normalized gain value.
10. The invention as defined in claim 9 wherein said transversal filter is employed in an echo canceler.
11. In a bidirectional repeater of the type including a first echo canceler adapted for canceling echo signals resulting in a first direction of transmission and a second echo canceler adapted for canceling echo signals resulting in a second direction of transmission each of said echo cancelers including an adaptive transversal filter having a plurality of means each for generating a corresponding one of a plurality of amplitude coefficient signals, said plurality of amplitude coefficient signals forming an impulse response, each of said amplitude coefficient signal generating means including controllable gain means responsive to a gain control signal for adjusting gain used in updating said corresponding amplitude coefficient signal;
means for generating a single gain control signal which is supplied to both said first and second echo cancelers including, first means for generating a first power estimate of a signal incoming to said first echo canceler, second means for generating a second power estimate of a signal incoming to said second echo canceler, and means supplied with said first and second power estimate values for generating a single normalized gain value and for limiting said normalized gain value to a predetermined value for power estimate values which are greater than a predetermined threshold value, said normalized gain value being said gain control signal.
means for generating a single gain control signal which is supplied to both said first and second echo cancelers including, first means for generating a first power estimate of a signal incoming to said first echo canceler, second means for generating a second power estimate of a signal incoming to said second echo canceler, and means supplied with said first and second power estimate values for generating a single normalized gain value and for limiting said normalized gain value to a predetermined value for power estimate values which are greater than a predetermined threshold value, said normalized gain value being said gain control signal.
12. The invention as defined in claim 11 wherein said means for generating said normalized gain value includes means for selecting either said first power estimate or said second power estimate in accordance with a prescribed criterion for generating said normalized gain value.
13. The invention as defined in claim 12 wherein said prescribed selecting criterion is to select the greater of the first and second power estimates
14. The invention as defined in claim 13 wherein said normalized gain value is limited to a fixed value for selected power estimate values greater than said predetermined threshold value.
15. The invention as defined in claim 14 wherein said normalized gain value generating and limiting means further includes means for storing predetermined normalized gain values corresponding to predetermined power estimate values and means responsive to said selected power estimate values for addressing said storing means to output a corresponding normalized gain value.
16. The invention as defined in claim 15 wherein each of said first and second power estimate generating means comprises a low pass filter.
17. In a bidirectional repeater of the type including a first echo canceler adapted for canceling echo signals resulting in a first direction of transmission and a second echo canceler adapted for canceling echo signals resulting in a second direction of transmission, each of said echo cancelers including an adaptive transversal filter having a plurality of means each for generating a corresponding one of a plurality of amplitude coefficient signals, said plurality of amplitude coefficient signals forming an impulse response, each of said amplitude coefficient signal generating means including controllable gain means responsive to a gain control signal for adjusting gain used in updating said corresponding amplitude coefficient signal;
means for generating a single gain control signal which is supplied to both said first and second echo cancelers including, first means for generating a first power estimate of a signal incoming to said first echo canceler, second means for generating a second power estimate of a signal incoming to said second echo canceler, and means supplied with said first and second power estimate values for generating a single normalized gain value and for increasing said normalized gain value for power estimate values which are greater than a predetermined threshold value, said normalized gain value being said gain control signal.
means for generating a single gain control signal which is supplied to both said first and second echo cancelers including, first means for generating a first power estimate of a signal incoming to said first echo canceler, second means for generating a second power estimate of a signal incoming to said second echo canceler, and means supplied with said first and second power estimate values for generating a single normalized gain value and for increasing said normalized gain value for power estimate values which are greater than a predetermined threshold value, said normalized gain value being said gain control signal.
18. The invention as defined in claim 17 wherein said means for generating and increasing said normalized gain value includes means for selecting the greater value of said first and second power estimate values for generating said single normalized gain value.
19. The invention as defined in claim 18 wherein said gain value is increased to be a value larger than that which would normally have been generated for said power estimate values which are greater than said predetermined threshold value.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US654,410 | 1984-09-26 | ||
US06/654,410 US4591669A (en) | 1984-09-26 | 1984-09-26 | Adaptive filter update gain normalization |
Publications (1)
Publication Number | Publication Date |
---|---|
CA1220823A true CA1220823A (en) | 1987-04-21 |
Family
ID=24624729
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CA000489848A Expired CA1220823A (en) | 1984-09-26 | 1985-08-30 | Adaptive filter update gain normalization |
Country Status (5)
Country | Link |
---|---|
US (1) | US4591669A (en) |
EP (1) | EP0176312B1 (en) |
JP (1) | JPH0671220B2 (en) |
CA (1) | CA1220823A (en) |
DE (1) | DE3570561D1 (en) |
Families Citing this family (35)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0167677B1 (en) * | 1984-07-13 | 1989-03-08 | BELL TELEPHONE MANUFACTURING COMPANY Naamloze Vennootschap | Signal processing arrangement |
JPS62104324A (en) * | 1985-10-31 | 1987-05-14 | Toshiba Corp | Adaptive automatic equalizer |
US4697261A (en) * | 1986-09-05 | 1987-09-29 | M/A-Com Government Systems, Inc. | Linear predictive echo canceller integrated with RELP vocoder |
US4845746A (en) * | 1987-06-23 | 1989-07-04 | Rockwell International Corporation | Echo canceller with relative feedback control |
FR2618277B1 (en) * | 1987-07-16 | 1989-12-15 | Centre Nat Rech Scient | ADAPTIVE PREDICATOR. |
JPH02280422A (en) * | 1989-04-20 | 1990-11-16 | Nec Corp | Training method for echo canceller for voice conference equipment |
NL8901247A (en) * | 1989-05-19 | 1990-12-17 | Philips Nv | ADAPTIVE TIME DISCREET TRANSVERSAL FILTER. |
EP0420269A3 (en) * | 1989-09-28 | 1992-12-30 | Fujitsu Limited | Adaptive digital filter including low-pass filter |
FR2671221B1 (en) * | 1990-12-27 | 1995-03-31 | Alcatel Radiotelephone | DEVICE FOR CANCELING ACOUSTIC NOISE. |
JP2842026B2 (en) * | 1991-02-20 | 1998-12-24 | 日本電気株式会社 | Adaptive filter coefficient control method and apparatus |
EP0715407B1 (en) * | 1991-02-20 | 2002-06-05 | Nec Corporation | Method and apparatus for controlling coefficients of adaptive filter |
EP0515761A1 (en) * | 1991-05-31 | 1992-12-02 | International Business Machines Corporation | Adaptive equalization system and method for equalizing a signal into a DCE |
DE69207509T2 (en) * | 1991-08-20 | 1996-08-22 | Philips Electronics Nv | Telecommunication system with a line for digital traffic with an arrangement for branching the digital duplex traffic, and arrangement for use in the telecommunication system |
US5268927A (en) * | 1992-10-06 | 1993-12-07 | Mayflower Communications Company, Inc. | Digital adaptive transversal filter for spread spectrum receivers |
US5412686A (en) * | 1993-09-17 | 1995-05-02 | Motorola Inc. | Method and apparatus for power estimation in a communication system |
FI935834A (en) * | 1993-12-23 | 1995-06-24 | Nokia Telecommunications Oy | A method for adapting to an echo point in an echo canceller |
JP2947093B2 (en) * | 1994-11-02 | 1999-09-13 | 日本電気株式会社 | Method and apparatus for system identification with adaptive filters |
JP3381112B2 (en) * | 1995-03-09 | 2003-02-24 | ソニー株式会社 | Echo canceler |
US5596600A (en) * | 1995-04-06 | 1997-01-21 | Mayflower Communications Company, Inc. | Standalone canceller of narrow band interference for spread spectrum receivers |
FR2738695B1 (en) * | 1995-09-13 | 1997-11-14 | France Telecom | ADAPTIVE IDENTIFICATION METHOD AND DEVICE AND ADAPTIVE ECHO CANCELER INCLUDING SUCH A DEVICE |
US5734715A (en) * | 1995-09-13 | 1998-03-31 | France Telecom | Process and device for adaptive identification and adaptive echo canceller relating thereto |
SE505152C2 (en) * | 1995-10-11 | 1997-07-07 | Ericsson Telefon Ab L M | Adaptive echo extinguishing procedure |
US5668865A (en) * | 1996-02-26 | 1997-09-16 | Lucent Technologies Inc. | Echo canceler E-side speech detector |
JP2924762B2 (en) * | 1996-02-28 | 1999-07-26 | 日本電気株式会社 | Adaptive filter and adaptation method thereof |
DE69630784T2 (en) * | 1996-09-24 | 2004-09-30 | Hewlett-Packard Co. (N.D.Ges.D.Staates Delaware), Palo Alto | Data processing equipment and method |
US6185299B1 (en) * | 1997-10-31 | 2001-02-06 | International Business Machines Corporation | Adaptive echo cancellation device in a voice communication system |
US6570985B1 (en) | 1998-01-09 | 2003-05-27 | Ericsson Inc. | Echo canceler adaptive filter optimization |
US6249147B1 (en) | 1999-03-09 | 2001-06-19 | Fujitsu, Ltd. | Method and apparatus for high speed on-chip signal propagation |
US7512149B2 (en) * | 2003-04-23 | 2009-03-31 | At & T Intellectual Property Ii, L.P. | Bit and power allocation scheme for full-duplex transmission with echo cancellation in multicarrier-based modems |
US6813352B1 (en) | 1999-09-10 | 2004-11-02 | Lucent Technologies Inc. | Quadrature filter augmentation of echo canceler basis functions |
DE10030123A1 (en) * | 2000-06-20 | 2002-01-03 | Infineon Technologies Ag | Circuit arrangement for analog echo cancellation |
GB2372653A (en) * | 2000-09-29 | 2002-08-28 | Univ Bristol | Adaptive filters |
US7035397B2 (en) * | 2001-09-14 | 2006-04-25 | Agere Systems Inc. | System and method for updating filter coefficients and echo canceller including same |
US7151803B1 (en) | 2002-04-01 | 2006-12-19 | At&T Corp. | Multiuser allocation method for maximizing transmission capacity |
JP2007514358A (en) * | 2003-12-10 | 2007-05-31 | コーニンクレッカ フィリップス エレクトロニクス エヌ ヴィ | Echo canceller with serial configuration of adaptive filters with individual update control mechanisms |
Family Cites Families (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS579258B1 (en) * | 1971-05-15 | 1982-02-20 | ||
GB1434239A (en) * | 1972-08-10 | 1976-05-05 | Siemens Ag | Echo cancellers |
US4129753A (en) * | 1977-12-09 | 1978-12-12 | Bell Telephone Laboratories, Incorporated | Echo canceller using feedback to improve speech detector performance |
US4468641A (en) * | 1982-06-28 | 1984-08-28 | At&T Bell Laboratories | Adaptive filter update normalization |
US4468640A (en) * | 1982-06-28 | 1984-08-28 | At&T Bell Laboratories | Adaptive filter update normalization |
US4467441A (en) * | 1982-10-08 | 1984-08-21 | At&T Bell Laboratories | Adaptive filter including controlled tap coefficient leakage |
-
1984
- 1984-09-26 US US06/654,410 patent/US4591669A/en not_active Expired - Lifetime
-
1985
- 1985-08-30 CA CA000489848A patent/CA1220823A/en not_active Expired
- 1985-09-17 DE DE8585306609T patent/DE3570561D1/en not_active Expired
- 1985-09-17 EP EP85306609A patent/EP0176312B1/en not_active Expired
- 1985-09-26 JP JP60211205A patent/JPH0671220B2/en not_active Expired - Fee Related
Also Published As
Publication number | Publication date |
---|---|
DE3570561D1 (en) | 1989-06-29 |
JPS6184927A (en) | 1986-04-30 |
JPH0671220B2 (en) | 1994-09-07 |
EP0176312B1 (en) | 1989-05-24 |
US4591669A (en) | 1986-05-27 |
EP0176312A1 (en) | 1986-04-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CA1220823A (en) | Adaptive filter update gain normalization | |
US4998241A (en) | Echo canceller | |
US4894820A (en) | Double-talk detection in an echo canceller | |
US4636586A (en) | Speakerphone with adaptive cancellation of room echoes | |
US5612996A (en) | Loop gain processing system for speakerphone applications | |
US5796725A (en) | Echo canceller capable of cancelling an echo signal at a high speed | |
US5313498A (en) | Method and arrangement of echo elimination in digital telecommunications system | |
JP2853455B2 (en) | Echo canceller | |
US4845746A (en) | Echo canceller with relative feedback control | |
CA1169135A (en) | Adaptive filter including a far end energy discriminator | |
EP0114855B1 (en) | Adaptive filter update normalization | |
US4467441A (en) | Adaptive filter including controlled tap coefficient leakage | |
CN111277718B (en) | Echo cancellation system and method thereof | |
US4628157A (en) | Bidirectional adaptive voice frequency repeater | |
US4584441A (en) | Bidirectional adaptive voice frequency repeater | |
US4468640A (en) | Adaptive filter update normalization | |
CA1168330A (en) | Echo canceler far end energy discriminator | |
US6959167B1 (en) | Noise-level adaptive residual echo suppressor | |
WO1982003144A1 (en) | Energy band discriminator | |
US7062038B1 (en) | System and method of using two coefficient banks in an adaptive echo canceller | |
US20030076844A1 (en) | Method and system for filtering a signal and providing echo cancellation using an adaptive length filter | |
JP2551869B2 (en) | Echo canceller | |
JPH0946277A (en) | Echo canceller | |
JPH07154468A (en) | Echo eliminator |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
MKEX | Expiry |