JPS58125920A - Interference controller - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention] Regarding.

Echoes occur in communication systems when an electrical signal in the receive signal path encounters an imperfectly matched impedance in the hybrid and is partially reflected back through the transmit signal path to the transmitting source. As a result of this, the reflected signal or echo will be heard at the far end of the transmission path some time after the original signal has been transmitted. As the distance M between the speaker and the listener increases, the time it takes for the echo to reach the speaker increases, with the result that the echo becomes, at least subjectively, more annoying to the speaker.

Attempts to control echo are therefore widespread. The echo control device disclosed in U.S. Pat. No. 4,005,277 includes a device operated by audio signals known as an echo suppressor. Typically, an echo suppressor is used to suppress the echoes that would otherwise be returned to the speaker.
It includes some form of luck damping that is implemented in response to levels. 7) echo, which is the round trip transmission time between the signal source and the return point; such a system is usually satisfactory for terrestrial communication channels where the delay is not long; However, with a communication channel via a satellite link, the transmission delay is much longer, the echoes are more harsh, and the conversation is interrupted by cutting the return path when both speakers are talking, i.e., double-talking. may confuse.

In contrast, other echo control devices, known as echo cancellers, do not cut the output path, but rather synthesize signals that are copies of the echo signal's impulse response and algebraically subtract the estimated value from the output signal. to obtain a signal with echoes cancelled. For example, the most common echo canceller, disclosed in U.S. Pat. No. 3,499,999, uses a tapped delay line with an adjustable multiplier in an adaptive feedforward path called a transversal filter. are synthesized. The multiplier is automatically adjusted by a control signal e (e) derived from the difference between the echo and replica signals. Since the impulse response of the echo path can be quite long, the transversal filter of the replica signal Accurate synthesis requires a large number of taps and associated multipliers, making the device complex and expensive.

The canceller models the voice assuming that the voice signal is an echo signal, and derives an estimated value of the impulse response of the echo path separately from the impulse response of the true echo path.
The near-end talker's call also becomes an obstacle for the echo canceller. To overcome this problem, near-end audio detectors are now used that inhibit the echo canceller's ability to form an echo path impulse response model when the detector assumes that a near-end audio signal is present. . An example of an echo canceller device including an active near-end audio detector is disclosed in US Pat. No. 4,129,753-1;

In this, the near-end audio detector includes a memory or other active circuitry, where the memory records the magnitudes of a number of past echo signals. In operation, the detector detects the transmitted signal. The magnitude of the current signal on the path is compared with the magnitude of the past echo signal stored in memory, and this comparison indicates that the current signal is considered to be near-end voice and not an echo signal. The moss provides an inhibit signal to the echo path modeling process.

This device solves the problem associated with other conventional devices of falsely indicating a near-end call when the signal is actually a large echo signal, but the near-end call detector used Not only does it require active circuitry, but it is also extremely large-scale, since the accuracy of the device is directly proportional to the amount of memory used (the number of past echo magnitudes kept).

In accordance with a feature of the invention, an interference controller for essentially canceling the interfering signal of the actual return signal propagating along the transmit signal path of a near-end transceiver, in operation, is subtracted from the returned signal to generate an interference cancellation signal, an interference copy signal is generated in response to the interference canceled signal and the input signal in the receive path of the transceiver, and other interference copy signals are subtracted from the return signal. the other interference is canceled out by the input signal, the other interference mirror signal is generated in response to the other interference canceled signal, and the coordination of the generation of the interference mirror signal An interference controller is provided that is inhibited in response to a predetermined relationship between interference canceled signals.

According to another feature of the invention, a detector for inhibiting adjustment of the interference controller in the presence of a near-end signal in the transmit signal path of the near-end transceiver is configured to detect the actual return signal along the transmit signal path. means for subtracting the detector interference copy signal from the propagation of the signal to generate a signal that cancels the detector interference, and the input signal propagated along the receive signal path of the transceiver and the detector interference canceling signal; processing means for receiving an input signal F-jD M d Z) and accumulating delayed sample values of increasing delay to form combinations of M values; ↑Means for generating copy number 18 by combining the Id signals collected in 11 and canceling out the interference of the detector, and receiving the signal canceling out the interference of the detector and the human signal, so that the delay of the human signal gradually increases. control means for storing a plurality of sample values and providing a signal for inhibiting adjustment of the interference controller according to a predetermined relationship between the detector interference cancellation signal and the plurality of increasingly delayed sample values; .

In accordance with one embodiment of the present invention, a near-end speech detector is included that has improved near-end speech detector performance and prevents the high probability of false detection of near-end speech of prior art devices. An echo canceller is provided.

In accordance with a feature of the invention, the near-end voice is reduced by incorporating a second echo canceller into the near-end voice detector to form a correlated near-end voice detector before the active echo canceller. A near-end speech detector is formed that is capable of distinguishing echo signals with large amplitudes.

The echo canceller included in the near-end audio detector is located in front of the prior art audio detector and receives as input a signal propagating along the transmit signal path. The near-end audio detector transforms this signal and then provides it to the input of a prior art audio detector.

An embodiment of the present invention will be described below with reference to the accompanying drawings. In the accompanying drawings, like numbers indicate like parts.

Referring now to FIG. 1, the figure essentially shows a single transmitting terminal connecting a near-end transceiver 10 to a receive signal path 12 and a transmit signal path 14 through a hybrid network 16. ing. Hybrid network 16 includes a balanced network (not shown) for impedance matching purposes. Ideally, all signals propagating through receive signal path 12 would be fed through high frit 16 to near-end transceiver 10, and signals from transceiver 10 would be fed only through hybrid 16 to transmit signal path 14. Can be placed. However, since impedance mismatches are unavoidable in the actual signal paths 12 and 14 connected to the hybrid 16, some of the signal energy in the receive path 12 will appear in the transmit path 14, resulting in some type of echo suppression or cancellation device. If not present, it is returned as an echo to the far-end transceiver (not shown). It is therefore necessary to cancel the echo signals appearing on the transmit signal path 14 using an echo cancellation device such as that shown in FIG. 1, which device includes a near-end audio detector.

As shown in FIG. 1, echo canceller 18 includes a transversal filter 18, a combinational circuit 20, and a near-end speech detector 22. Transversal filter 1
8 is also called a tapped delay line in the prior art,
It receives the signal x(k) arriving at the near-end transceiver 10 through the receive signal path 12, stores its successive samples, processes these samples, and produces as output an echo replica signal ? An estimate of the echo signal on the transmit signal path 14 is formed, called (k).

The combinational circuit 20 is arranged on the transmit signal path 14 and converts the echo replica signal 9(k) produced by the transversal filter 18 into the actual return echo signal y ( k) to form a difference signal defined as the echo cancellation signal e(k).

The echo cancellation signal e(k) then propagates along the remainder of the transmit signal path and is received by a far-end transceiver (not shown). This same echo cancellation signal is fed back to the transversal filter 18 to alter the estimate of the echo signal, which is the echo copy signal ↑(k) formed by the transversal filter 18. supply the necessary information. Vi, in the ideal case? (k) can be accurately modeled, and the echo cancellation signal e (k) is equal to 0V (-, the far-end transceiver receives no echo. However, in reality, there is some echo A cancellation signal will always be returned to the far end.

The near-end sound occurring on the transmit signal path 14 is stronger than the echo and acts as undesirable noise as far as the transparsal filter 18 is concerned. If the operation of the transversal filter 18 is not continued even when near-end audio is present, the echo copy signal 9(k) will be different from the actual return echo signal y(k), and the near-end It ends up being a model of the voice at the edge.

To solve this problem, near-end speech detector 22 is used to inhibit continuous updating of transversal filter 18 when near-end speech signals are present. As shown, the near-end audio detector 22 receives the input signal x(
k) and the actual returned echo signal y(k), and using an algorithm related to these two signals, a transversal filter is applied when a near-end audio signal is considered to be present. prohibit.

An example of a detection algorithm that can be implemented with the near-end speech detector 22, which is well known to those skilled in the art, is y(k)>1/2 max(x(k), x(k-1)
, -, x (k-(N-1) ) ) (1). Here, according to 20), the near-end speech detector 22 uses a transversal filter similar to the transversal filter 18 for generating the sequence (x(k), -x(k (N 1))) and a signal y(k) is individually compared Vζ with each of the stored values of the input signal and an inhibit signal is generated if y(k) is greater than half the maximum of the stored values in the series. When the inequality in equation (1) is satisfied, the near-end audio detector 22 sends a prohibition signal to the transversal filter 1B.
8 until the inequality in equation (1) is no longer satisfied.
The operation of the transversal filter 18i1 is stopped.

When the inequality in equation (1) is no longer satisfied, near-end speech detector 22 sends an energizing signal to transversal filter 18 to restart the transversal filter 18 update process. Coefficient 1/ of equation (1)
2 follows the assumption that there is at least 6 dB loss in the signal going from the receive signal path 12 to the transmit signal path 14 through the hybrid 16. However, when the initial return echo signal y(k) is large, the near-end voice detector 22
A problem remains with this device, since operation of the transversal filter 18 may be erroneously inhibited by this.

Referring to diagram M2, transversal filter 18 receives and stores samples of input signal x(k), and then synthesizes echo copy signal 9(k). combinational circuit 20
then subtracts this signal from the return echo signal y(k) to form an echo cancellation signal, which is then applied to transmit signal path 1.
4 to a far-end transceiver (not shown). As shown in FIG. 2, near-end speech detector 24 includes a prior art near-end speech detector 22 as well as a transversal filter 26 and a combinational circuit 28. As shown in FIG. Transversal filter 26 operates similarly to transversal filter 18 described above, receiving successive samples of input signal x(k) and synthesizing therefrom a voice detector echo transcript signal yd(k). do. Although the combinational circuit 28 is not included in the transmit signal path 14 and does not affect the signal passing through the transmit signal path 14, this voice detector mirror signal? d(k) is subtracted from the actual return echo signal y(k) on the transmit signal path 14 leaking through the hybrid 16 to obtain the echo-cancelled signal ed(k) of the audio detector.
form. Echo cancellation signal ea (k) of the audio detector
is then fed back to the transversal filter 26, providing the transversal filter 26 with the information necessary to improve the echo path estimate. The audio detector echo cancellation signal ed(k) formed by the combinational circuit 28 is also provided as an input to the prior art near-end audio detector 22, which is also responsive to the input signal x(k). It has become. Prior art near-end audio detector 22Vi
Identify near-end audio and inhibit transversal filter 18 according to the following equation:

ed(k)) 1/2 max(x(k), x(k-
1),...,X (k-(N-1)) )(2) Specifically, the echo cancellation signal ea(k
) is less than half of the largest individually delayed value of the input signal stored in the prior art near-end speech detector 22, near-end speech is considered to be present and the transversal filter 18 will be prohibited from continuing updating for a period of time when the inequality in equation (2) is satisfied.

Formula (1)1. ' (2), the present invention differs from the prior art in that the near-end voice detector 22 of the prior art uses the echo canceled signal ea (k ) is used. Therefore, by using ed(k), the possibility that the near-end voice detector 24 mistakes a large echo signal as near-end voice and erroneously inhibits the transversal filter 1B is reduced compared to the prior art using signal y(k). This results in a reduction compared to near-end voice detectors.

A detailed view of one embodiment of the invention is shown in FIGS. 3 and 4, arranged as in FIG. This figure shows the structures of the tapped delay lines of the transversal filter 1B and the transversal filter 26 of the near-end voice detector. As shown in Figure 3, the transversal filter 18 Vi30 t -30N 1 (
N-1) delay elements and N I'li! d multiplier 3
2132 N and N tap weight generators 34134
N and an accumulator 36. The signal that has passed through the reception path 12 is applied to a transversal filter '18, propagated through (N-1) delay elements 30 t 30 N-□, and is related to the current input signal value x (k). As shown in Figure 3, the values (X(k), x(k-1),
..., x(k (N 1))), forming a sequence of length N.

As previously discussed in connection with FIGS. 1 and 2, echo cancellation is used to aid in updating the echo copy signal 9(k) in order to improve modeling of the actual return signal y(k). The resulting signal e (k) is transversal filter 1
8 will be fed back. However, the echo-cancelled signal e (k) itself is not suitable for improving the echo replica signal 9(k). Therefore, the echo-canceled signal e(k) is passed to the N-tap weight generator 34134 which functions to individually form the N-tap weight values.
pass through. The values of these weights are the N sampled and delayed input signal values x(k) to x(k-(
N-1)'), which when summed in an accumulator 36 form an echo replica signal '>(k). As illustrated in detail in connection with the tap weight generator 341, the exemplary tap weight values ho (k
) from the echo-canceled signal e (k), multiplies the echo-canceled signal e (k) by its associated input signal sample value x (k) in a multiplier 351, and integrates the resulting composite signal. Circuit network 37! Tap weight value on average. (k) such that its polarity and magnitude indicate the necessary correction value required based on its input signal sample value x(k)Ic. Generator 34
Each of these tap weight values ho(k) −↑ (k) generated by r 34 N is generated by a respective associated delayed input signal value x (k ) by a respective associated −1 multiplier 321 32 XN-□(k
) is multiplied by More specifically, the tap weight value h
o (k) is the multiplier 32! Input sample xCk by
), and hl(k) is multiplied by x(
k1), and the tap weight value 1N-
□(k) is input sample x(k(
N-1)).

The N outputs of multiplier 32132N are then provided as separate inputs to accumulator 36, which is the N product ho(k)
x(k) to h N-□(k) x (k-(N=1
)) to produce the echo copy signal y(k) as the output of the transversal filter 18. As mentioned above in connection with FIG. 2, the echo copy signal? (k) is subtracted from the actual return signal y (k) by a combinational circuit 20, yielding a sequentially updated value of the echo cancellation signal e (k) as the output of the echo canceller, which is then sent to the end.

A detailed embodiment of the near-end audio detector 24 is shown in FIG. In the figure, the transversal filter 26 consists of the same components as the transversal filter 18, and (M-
1) delay elements 40t 4G, M multipliers-1 421-42, M tap weighting generators 441-4;
4 M (where only the multiplier 41! and the tap weight generator 44+, including the integrator network 431, are shown in detail in FIG. 4; the others have been omitted to simplify the diagram) and It includes an accumulator 46. In operation, the input signal x (k) propagated through the receiving signal path is introduced into the transversal filter 26, and the delay elements 401-40M
, and in relation to the current value x(k) of the input signal, a sequence of length M (X(k), x(k−1), −
..., x(k-CM-1))). Tap weight generator 44! -44 is a near-end voice detector combination circuit, 28
The echo cancellation signal of the voice detector generated by ed (k)
In response, the tap weight generator 34! Generate the M voice detector tap weight values jo(k) through JM-□(k) in the same manner as described above in connection with . Each of the speech detector tap weight values is then multiplied by the associated delayed value of the input signal in its associated multiplier 421-42.

In more detail, tap 1 of the voice detector
(k) is multiplied by the input sample value x (k) in multiplier 421, and j+(k) is multiplied by x(k-1) and multiplier 422
The tap weight value j (k) of the voice detector is multiplied by the multiplier 42M and the delayed input value +vi-1 x (k-(Fvj-1)).

The M outputs of multiplier 42 x -42M are then provided as separate inputs to accumulator 46, which contains the M products JO(
k) A transversal filter 26 which forms the sum of x(k) to JM1(k)x(k-(muvi-1)) and corresponds to the near-end voice detector echo transcript signal yd(k). generates an output signal. As previously discussed in connection with FIG. 2, the voice detector echo transcript signal yd(k) is subtracted from the actual return echo signal y(k) in a combinational circuit 28 and the near-end voice detector 22 of the prior art The signal given as an input, i.e. the echo cancellation signal of the audio detector ed
(k) is generated.

As previously discussed in connection with FIG. 2, the near-end voice detector's echo-cancelled signal ed(k) is fed back to the transversal filter 26, and the near-end voice detector 22 is Given. In conjunction with the transversal filter 26, the audio detector echo cancellation signal ed(k) is transmitted to each tap weight generator 44! -44M
The tap weight value ↑o(k) J
(k) is updated continuously so that the echo transcript signal y of the voice detector is
Let d(k) continuously converge to the actual return echo signal y(k). As mentioned above in connection with Figure 2,
The voice detector echo cancellation signal ed(k) is the input of the prior art near-end voice detector, and in conjunction with equation (2), the near-end voice detector 24 distinguishes between large echoes and near-end voice. Be able to distinguish between. After this, the near-end audio detector 2
4 can reduce the probability that the operation of the transversal filter 18 is erroneously detected by lowering the false detection rate compared to the prior art device.

[Brief explanation of the drawing]

FIG. 1 is a block diagram illustrating an example of a prior art echo canceller device including a prior art near-end audio detector, and FIG. 2 is an echo canceller including an embodiment of the improved near-end audio detector of the present invention. Block Diagram Showing an Example of Apparatus FIGS. 3 and 4 are block diagrams showing a detailed embodiment of the echo canceller apparatus of FIG. 2, arranged as in FIG. 5. [Explanation of symbols of main parts] Name in claims Code Name in specification Interference copy from return signal 20 Subtraction circuit and means for subtracting the signal First processing means 1B Transversal filter detector 24 Near-end speech detector Second processing means 26 Transversal filter control means 22 Near-end speech detector M-1 delay elements 40 Tap weight generator with M delay elements
44 Tap weight generator M multipliers 42 Multiplier addition means 46 Accumulator Applicant Western Electric Company, Inc. Procedural Amendment (Method) March 1, 1980 Patent Office Commissioner ri Kazuo Sugi Case 1 Display of 1984 Patent Application No. 184799,
Relationship with the case of the person making the amendment Patent applicant R (shipment date: February 22, 1988) -7 Correspondence for amendment (1) Column ``Brief explanation of one drawing of the specification'' tiiCorrect content As shown in the attached sheet (])
Page 27 of the specification, lines 6 to 8 [Figures 3 and 4]
The figure is a block diagram. 3 and 4 are block diagrams showing detailed embodiments of the echo canceller device shown in FIG. 2, and FIG. 5 is a diagram showing the connection relationship between FIGS. 6 and 4. ” he corrected.

Claims (1)

Claims: 1. In an interference controller that essentially cancels the actual return signal propagated along the transmit signal path of a near-end transceiver, in operation an interference replica signal is subtracted from the return signal. an interference cancellation signal is generated in response to the interference cancellation signal and an input signal on a receiving path of the transceiver;
Another interferometric signal is subtracted from the return signal to generate an additional interference cancellation signal, and the other interferometric signal is generated in response to the other interference cancellation signal and the input signal, and the other interferometric signal is generated in response to the input signal and the other interference cancellation signal. An interference controller, characterized in that adjustment of the generation of interferometric signals according to a predetermined relationship between the signals is prohibited. 2. The controller according to claim 1, further comprising: means for subtracting the interference copy signal from the return signal to generate an interference cancellation signal; and means for receiving the input signal and the interference cancellation signal; first processing means for storing N sample values of increasing delay and combining the plurality of values with an interference cancellation signal to generate an interference replica signal; and subtracting another interference replica signal from the return signal. a detector including means for generating another interference cancellation signal, receiving the other interference cancellation signal and the input signal and storing M sample values with increasing delays of the input signal; a second processing means for generating another interference copy signal by combining the interference cancellation signal of the input signal with the input signal; An interference controller characterized in that it includes control means for storing and locking the processing means according to a predetermined relationship between another interference cancellation signal and a plurality of increasingly delayed sample values. 3. In the controller according to claim 2, the second processing means responds to the input signal so that the delay of the input signal becomes gradually thicker. (M-1) delay elements forming a transversal filter responsive to No. 8 and M taps each generating a respective one of the M tap weight values in response to the other interference cancellation signal. a weight generator, M multipliers each responsive to a respective one of the M increasingly delayed sample values and the M tap weight values to generate a respective product of these values; 4. An interference controller characterized in that it includes an addition means for adding products to generate another interference copy signal. 4. The controller according to claim 2 or 3, wherein characterized in that the control means is adapted to lock the first processing means when the signal is thicker than half of the largest sample value among the plurality of sample values of the human input signal with increasing delay; 5. An interference controller that inhibits adjustment of the interference controller in the presence of a near-end signal on the transmit signal path of the near-end transceiver, and which detects the actual return signal propagated through the transmit signal path. A means for subtracting the interfering signal of the detector from the signal to generate detector interference cancellation +14; receive,
processing means for storing a first plurality of M progressively delayed sample values of the input signal and for combining the M values with a detector interference cancellation signal to generate a detector interference copy signal; , receives the detector's interference cancellation signal and the input signal, stores a plurality of increasingly delayed samples of the input signal, and stores the detector's interference cancellation signal and the +M number of increasingly delayed samples of the input signal. and control means for providing a signal to inhibit adjustment of the interference controller according to a predetermined relationship between the sample values. 6. The detector according to claim 5, wherein the processing means forms a transversal filter that generates M sample values of increasing delay in response to an input signal. (M-1) delay elements, each of which generates each of the M +15 tap weight values in response to the detector interference cancellation signal; M multipliers that generate a product of these values in response to each of the sample values and each of the M tap weight values, and summing the M products to generate an interferogram signal. A detector characterized in that it includes means. 7. In the detector according to claim 5 or 6, the interference cancellation number 15 of the detector is less than half of the maximum sample value among the sample values of the input signal whose delay gradually increases. Detector characterized in that the control means is operative to provide a signal inhibiting adjustment of the interference controller when the interference controller is large.