JPH02260716A - Echo canceller - Google Patents

Abstract

PURPOSE:To improve traceability for the change of an echo path by outputting an estimation start signal by judging that the echo path is changed when a variance is smaller than a first threshold value, and the variance is larger than a second threshold value smaller than the first threshold value, and also, a mean value is larger than a third threshold value. CONSTITUTION:When the variance delta<2>AE is smaller than the threshold value (first threshold value TH1) to identify the double talk of voice, and simultaneously it is larger than the threshold value (second threshold value TH2) to identify the double talk of a modem signal smaller than the TH1, and also, the mean value mAE is larger than a certain threshold value (third threshold value TH3), the detection of the double talk stage by a double talk detector 24 is judged as the change of the echo path, not the double talk actually. In such a way, the fact that TH2<delta<2>AE<TH1 and mAE<TH3 are satisfied by using the first threshold value TH1, the second threshold value TH2, and the third threshold value TH3 set in advance can be judged as the change of the echo path.

Description

DETAILED DESCRIPTION OF THE INVENTION (Industrial Application Field) This invention relates to an echo canceller that cancels near-end echoes generated in a 2-wire to 4'a converter in communication lines such as satellite communication and voice packet communication. It is.

(Prior art) Current telephone lines generally use a hybrid circuit that converts between a two-wire subscriber line and a four-wire transmission line. Mismatching causes the received signal to leak into the transmission line, causing an echo. In networks with large delays such as satellite communications, this echo becomes a hindrance to communication, so an echo canceller is inserted into the transmission line to cancel the echo. In addition, in integrated networks that handle voice, images, and data in an integrated manner, voice is packetized for communication, which causes large delays even on domestic lines, so it is necessary to use an echo canceller to cancel echoes even in voice packet communication. There is.

FIG. 2 is a block diagram showing a conventional circuit configuration when an echo canceller is applied to a telephone line.

In the figure, 10 is a telephone, 1.2 is a subscriber line, and 14
is a hybrid circuit (indicated by symbol H in the figure), 16
is a digital-to-analog converter (indicated by the symbol D/A in the figure), 18 is an analog-to-digital converter (indicated by the symbol A/D in the figure), 20 is a subtracter, and 22 is an adaptive digital filter ( (indicated by the symbol ADF in the figure) and 24 is a double talk detector (indicated by DTD in the figure)
and this adaptive digital filter 22, subtracter 2o
The double talk detector 24 constitutes an echo canceller 26.

The operation of this conventional circuit will be briefly explained.

The received signal x(k) (where k is the sampling time) from the far end is converted into an analog signal x(t) (where t is the time) by the D/A converter 16, and the hybrid circuit (H)+4 and is transmitted to the telephone set 1o through the subscriber line 12 as a two-wire signal. At this time, due to the impedance mismatching of the hybrid circuit (H)+4, the received signal x(t) leaks into the transmission path and is sent out as an echo signal y(t).
The near-end input signal n(t) from 0 is input to the hybrid circuit (
1-1) It is converted into a 4-wire signal by 14 and sent to the transmission path. That is, the transmitted signal s (t) is the sum of the echo signal y (t) and the near-end input signal n (t), and is expressed as 5(
t) = y(t) + n(t).

The echo canceller 26 receives the received signal x(k) and the A/
A transmission signal 5(k) converted into a digital signal by the D converter 18 is input. The adaptive digital filter 22 of the echo canceller 26 uses a learning identification method to detect D/
System from A converter 16 to A/D converter 18 via hybrid circuit (H)+4 (hereinafter referred to as echo path)
A pseudo echo signal y(t) is generated from the estimated impulse response and the received signal, and this is subtracted from the transmitted input signal 5(k) by the subtracter 20 to cancel the echo. . Therefore, this subtractor 2
0 functions as an echo cancellation circuit, and the output of this subtracter 20 is sent out as a residual signal e(k) = 5(k) - y(k).

However, in the learning identification method, the received signal sequence is used to estimate the echo path impulse response: X = = (x(k), x(k
-1) , -1x(k-N◆l) ) and the residual signal e(
k), if the estimation is continued in 1f in a double talk state where the near-end signal n(t) is large, the estimated impulse response will be disturbed.The double talk detector 24 detects the double talk state. , outputs an estimation stop signal INH to stop the estimation of the adaptive digital filter 22, thereby preventing this. As this double talk detection method, for example, the power ratio of the received signal x(k) and the residual signal e(k) is determined, and when this power ratio A is less than a predetermined threshold @AT, that is, A≦A When the following conditions are satisfied, it is determined that a double talk state is present. Here, AT is a double talk detection threshold. This method has high double talk detection sensitivity and the near-end signal n(k)
can be detected even at small levels.

However, even when the echo path switches to another system, the power of the residual signal e(k) increases, so (2)
The formula holds true, and it is determined that double talk is occurring even though it is not actually a double talk state, and as a result, the above-mentioned estimated stop signal INH! is output from the double talk detector 24. The resultant signal is output to the adaptive digital filter 22, and the estimation operation in this filter 22 is stopped.

In an actual telephone line, the echo path changes when the telephone switches from an on-hook state to an off-hook state or vice versa, or when a mute operation is performed, making it impossible to eliminate echoes.

Conventionally, as disclosed in Japanese Unexamined Patent Publication No. 61-56526, the threshold AT is set to a constant i! than the threshold at the previous sampling time when double talk is determined. i6.
A method is used to accommodate changes in the echo path by setting the value to be a value as small as 16 degrees and, when determining non-double talk, to a value that is a constant 16 degrees larger than the threshold at the previous sampling time. In other words, the a value Av in each case is: ■Double talk judgment AtCk◆I) = A v(k)-5u■Non-double talk judgment AT(k◆1)=A y(k)+5゜It is determined that

If we follow this method of determining the threshold AT, if a change in the echo bus is determined to be double talk, the threshold AT will be
As the value of , decreases, the power ratio AtJ<
It becomes larger than fl@Ay (A>AT), so the condition of the above-mentioned equation (2) is no longer satisfied, and therefore estimation is started. Here, the value of 6u is determined so that A > A at the end of about 3 seconds after a double talk state is determined, since it is extremely rare for an actual double talk state to last for more than 3 seconds. ing.

(Problem to be Solved by the Invention) However, in the conventional method described above, changes in the echo path are detected based on the double talk judgment time, so it takes at least about 3 seconds from the time the echo path changes until it is detected. .

If emphasis is placed on safety during actual double talk, the detection time must be made sufficiently long, which poses a problem in that the echo canceller's ability to follow changes in the echo path is significantly lost.

Furthermore, in recent years, not only voice communication but also data communication using modems is carried out over telephone lines. However, since the double talk state continues for a long time during data communication, the conventional method cannot be applied to data communication, and its use is limited to voice communication.

This invention was made in order to eliminate the above-mentioned problem that increasing the double talk detection sensitivity of the echo canceller deteriorates the followability of echo path changes and the problem that it cannot be applied to modem signals. Therefore, it is an object of the present invention to provide an echo canceller that has good ability to follow echo path changes without lowering the double talk detection sensitivity, and can also be applied to modem signals.

(Means for Solving the Problems) In order to achieve this object, the present invention provides an adaptive digital filter that inputs a received signal and generates a pseudo echo signal, and an adaptive digital filter that generates a pseudo echo signal by inputting a received signal. an echo cancellation circuit that sends out a residual signal with the echo signal canceled by subtracting it from the transmitted signal, and a double talk state is detected from the power ratio of this residual signal and the received signal to estimate the above-mentioned adaptive digital filter. In an echo canceller having a double talk detector that stops the operation, the adaptive digital filter stops the estimation operation when the double talk detector detects a double talk state and detects a change in the echo path while the estimation operation is stopped. An echo path change detector is provided which outputs an estimation start signal for restarting the estimation operation to the above-mentioned adaptive digital filter. a calculation circuit that calculates an average value and a variance value of the ratio; and a calculation circuit that calculates an average value and a variance value of the ratio; The present invention is characterized in that it has a comparison/determination circuit that determines that there is an echo path change when the value is large and outputs the above-mentioned estimation start signal.

(Operation) According to the configuration of the present invention described above, when the double talk detector detects a double talk state, the echo path change detector detects that if the detection of the double talk state is due to a change in the echo path, the echo path change detector detects the double talk state. An estimation start signal is output from the echo path change detector to the adaptive digital filter to restart the estimation operation of the adaptive digital filter that was stopped due to the detection of this double talk state.

In order to output this estimation start signal, the power ratio between the power of the received signal and the power of the transmitted signal is always calculated in the calculation circuit regardless of the presence or absence of the double talk state.
Then, when the estimation operation is stopped, the variance value and average value per N sampling time of this calculated power ratio are calculated, and then, in the comparison judgment circuit, this variance value is always applied to the first threshold value and the second threshold value. The average value is compared to a threshold value and the average value is compared to a third threshold value. Then, when the estimation operation is stopped due to double talk state detection, if it is determined that this variance is smaller than the first threshold value, larger than the second threshold value, and the average value is larger than the third threshold value, this comparison judgment circuit The above-mentioned estimation start signal is output from the adaptive digital filter to the adaptive digital filter.

In this way, in the present invention, it is monitored whether or not the double talk state detection is caused by a change in the echo path.
The time from when a change in the echo path occurs to when an estimation start signal is output does not require a waiting time of 3 seconds as in the conventional case, and is essentially only the processing time in the echo path change detection circuit. The echo canceller of
The adaptive digital filter can be restarted in about 1/10 of the time required by the conventional method, and has extremely good followability to changes in the echo path.

Additionally, it simultaneously monitors whether the transmitted and received signals are modem signals, and can also be applied to modem signals.

(Embodiments) Hereinafter, embodiments of the echo canceller of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram for explaining this echo canceller, and FIG. 3 is a block diagram showing the circuit configuration of an echo path detector provided in this echo canceller. Furthermore, the first
In the figures, the same components as those shown in FIG. 2 are designated by the same numbers, and detailed explanation thereof will be omitted unless specifically mentioned.

As shown in FIG. 1, an echo canceller 30 according to the present invention is added to the conventional configuration conditions of an adaptive digital filter (ADF) 22, a double talk detector (DTD) 24, and a subtracter 20 as an echo cancellation circuit. An echo path change detector 40 is provided.

This echo path change detector 40 takes in the received signal x(k) and the transmitted signal 5(k) to monitor changes in the echo path, and when it detects a change in the echo path while the estimation operation has stopped, The configuration is such that an estimation start signal for restarting the estimation operation of the adaptive digital filter 22 is output to the adaptive digital filter 22.

Next, this echo path change detector 40 will be explained in detail.

In the present invention, the echo path change detector 40 includes a power ratio calculating circuit 50 that calculates the power ratio of the received signal x(k) and the transmitted signal S(k), and an average value m of the power ratio.
A calculation circuit 35 consisting of an average calculation circuit 60 that calculates AEa, a variance calculation circuit 70 that calculates the variance value σ2AE of this power ratio, and a switch SW, and the outputs of these average values and variance values are used to determine echo path changes. A comparison and determination circuit 80 is also included.

The operation of the echo path change detector having such a configuration will be explained with reference to FIGS. 1 and 3.

First, the transmission signal 5(k) and the reception signal x(k) are input to the power ratio calculation circuit 50, and the power ratio calculation circuit 51.52
After calculating the average powers 5(k) and X(k) of these signals 5(k) and x(k) over a short period of time, the following equation (3
).

Therefore, the average powers S (k) and X (k) are converted into values Ls (k) and Lx (k) in the logarithmic domain using a logarithmic conversion circuit (I209+) 53.54. The subtracter 55 obtains the power ratio A E(k)! of the above equation (3) by finding the difference (L x (k) - L 5 (k)) between these signals.
Calculate and output. Here, the power calculation circuit 51 mentioned above
．． In 52, the average power 5(k), X(k)
As is well known in the art, it can be determined by the following formula % formula %().

In this embodiment, this power ratio A t(k) is passed through the switch SW to the average calculation circuit 60 and the variance calculation circuit 7.
0 respectively. This switch SW is activated when the double talk detector 24 detects a double talk state and outputs the estimated stop signal IN)-1.
Electronic, optical, or
While this switch SW, which may be any type of mechanical switch, is open, the double talk detector 24 detects the double talk state and outputs the estimated stop signal lNl-1. In the meantime, the average calculation circuit 60 and the variance calculation circuit 70 calculate the average value mAE and the variance value 02AE ! of the power ratio A E(k) over N sample times. Find each according to . Here, N is 20 from 100m5eC
A value corresponding to a time of about 0m5ec, for example, 8kHz2
The value is approximately 800 to 1600 samples.

In this embodiment, the average calculation circuit 60 first calculates the cumulative value of the power ratio A E(k) using an accumulator (ACC:) 61, and then calculates the cumulative value of the power ratio A E(k) using the pulse T having the period N. Therefore, the power ratio A e(
The value of the sum Σ Aε(k) of
At the same time, the contents of the accumulator 61 are cleared. Then, it is sent to the cumulative Ja Mira calculator 63 set in the latch circuit 62, where it is multiplied by the constant 1/NJ to obtain an average power ratio of 41 m□.
It is sent to the comparison/judgment circuit 80 and the variance calculation circuit 70 as an output of 0.

On the other hand, in the distributed calculation circuit 70, while the above-mentioned switch SW is closed, the power ratio A E (
k), and calculates the square value A t(k)'' of
The accumulated value is then the squared value of the N sample interval (k)
Calculate the value of the sum of 2 (>, , A b k) 2). Then, as in the case described above, this accumulated value is set in the latch circuit (D) 73 by a pulse T of period N, and at the same time, the contents of the accumulator 72 are cleared.

Then, the accumulated value set in the latch circuit 73 is sent to the multiplier 74, where it is multiplied by a constant 11/NJ and sent to the root mean square of the power ratio A, (k) of the N sample section. The subtracter 76 receives the average power ratio mAtM squared value (mAE') obtained by squaring the power ratio output from the average calculation circuit 60 described above. Then, consider the average @m□ and the variance value σ2AE of the power ratio in the root mean square value. Since the transmitted signal 5(k) is the sum of the near-end input signal n(k) and the echo signal y(k), (
mAE'), the dispersion value difference is sent to the comparison/judgment circuit 80.

Using this average value m□ and the variance value σ2AE, the comparison and determination circuit 80 determines whether the estimation stop is due to a change in the echo path, and if it is determined that the estimation stop is due to a change in the echo path, the estimation start signal ADP ! It is output to the adaptive digital filter 22.

Next, a method for determining an echo path change in this comparison and determination circuit will be described.

Now, consider what the power ratio A t(k) will be when the impulse response of the echo path = ah, and then calculate 5(k) = n during non-double talk and double talk.
(k) + y(k)=n(k)+H”X(k)
(7) (where T represents a transposition matrix). Average power 5(k) of the transmitted signal 5(k) given by equation (7)! When you find it, n (k) and x (k)
is uncorrelated, so when the echo path delay is small,
Using short poem n average power X (k) of equation (4,1),
Approximately 5(k) =N(k) +1H12X(k)
(8) Therefore, from equations (3) and (8), the power ratio A of the received signal x(k) and the transmitted signal 5(k)
ε(k) is expressed as follows.

Therefore, in non-double talk, the average power N (k) of the near-end input is a very small value, so the power ratio A1(k) in equation (9) is given by equation (10). Power ratio A E (
k) reflects the attenuation in the echo path (see FIG. 1), and in a telephone line, it is a nearly constant positive value for a given echo path.

That is, the power ratio A 1(k) calculated by equation (5)
The average value m□ is a positive value, and the variance value O''AE calculated by equation (6) is a small value.

On the other hand, during double talk, ■ When the received input signal x (k) and the near-end input signal n (k) exist simultaneously, ■ When x (k) and n (k) are audio signals, x (k) and n
Since (k) is uncorrelated, A expressed by equation (9)
The temporal change in E(k) is large, that is, the variance value σ2□ in equation (6) becomes a large value.

■When x(k) and n(k) are modem signals, in the medium and low speed modems that are commonly used, x(k) and n(
Since k) is a signal obtained by phase modulating or frequency modulating a sine wave, the amplitude change is extremely small, that is, (9)
The temporal change in A E(k) expressed by the equation is very small, and the variance value σ2Aε of equation (6) is a much smaller value than the value when the audio signal is not double-talked (described above). Become.

Furthermore, (2) if only the near-end input exists, x(k) will be a very small value, and therefore Equation (9) A t(k) will be a large negative value. That is, the average value mAE in equation (5) is a negative value.

Considering the above points, the variance value σ2A! is smaller than the threshold for identifying double talk in the voice (this is referred to as the first threshold THI), and at the same time is smaller than the threshold for identifying double talk in the modem signal (this is referred to as the second threshold TH)
2), and if the average value mAE is greater than a certain threshold (this is taken as the third threshold TH3), the double talk detector 24 (FIG. 1) detects the double talk state 18: It is possible to set in advance the first threshold THI, the second threshold TH2, and the third threshold TH3, which can be used to detect a change in the echo path rather than double talk.

Therefore, the first threshold TH1, the second threshold TH2, and the third threshold set in advance in this way! ! Using TH3, when TH2<02AE<TH1 and mAE>TH3 are satisfied, it is possible to determine that there is a change in the echo path.

Therefore, in this embodiment, this comparison/judgment circuit 80 is combined with a comparison circuit 81 between the third threshold value TH3 and the average value mAE, a comparison circuit 82 between the first lower threshold value H1 and the variance value σ2AE, and a comparison circuit 82 between the third threshold value TH3 and the average value mAE, and the second threshold value TH2. and the dispersion value G''AE, and an AND (A N D) circuit 84 between the results of these comparison circuits 81, 82 and 83, and the pulse T. In the comparison/judgment circuit 80 of the configuration,
In the comparison circuit 81, the average value mAE and the third threshold value T) (
3 and outputs logic "1" if mAE>TH2. In addition, in the comparison circuit 82, to the variance value σ,
is compared with the first threshold TH1, and σ2AE<THl
If so, a logic "1" is output. Also, in the comparison circuit 83, the variance value σ2A1. is compared with the second threshold Tl-12, and σ2A! >TH2, outputs logic "1".

In the AND circuit 84, the respective outputs of these comparison circuits 81, 82 and 83 are ANDed with the pulse T, and the output of this AND is outputted as an estimation start signal to the adaptive digital filter 22 (FIG. 1). and restart it. In this way, T H2< G ”AE <
When T H1 and mAl:〉TH3, it is determined that the cause of stopping the estimation operation is an echo path change,
Estimated start signal A[)P (logical rlJ)! Outputs one sample jig section.

It is clear that the present invention is not limited to the embodiments described above, and that many modifications and changes can be made.For example, the configuration of the echo path change detector explained in FIG. 3 is not the configuration of this embodiment. The configuration is such that the average value and variance value are calculated using the power ratio of the received signal and the transmitted signal, and these average value and variance value are compared with respective predetermined threshold values to output an estimation start signal. It would be good if it was.

Further, the configuration examples shown in block diagrams in FIGS. 1 and 3 show the components necessary for understanding the present invention. generator, generator of N-period pulses T, sampling period N3, first, second and third thresholds THI;
Memories such as TH2 and TH3, as well as other circuits and/or other conventional means that can be easily understood by those skilled in the art are omitted from illustration and explanation.

(Effects of the Invention) As is clear from the above description, according to the echo canceller of the present invention, the circuit that calculates the average and variance of the power ratio of the received signal and the transmitted signal and the double talk detector are in the double talk state. While the estimation operation is stopped, when the variance value is smaller than the first threshold value, at the same time it is smaller than the first threshold value, is larger than the second threshold value, and the average value is larger than the third threshold value. Since we have installed a circuit that determines that there is a change in the echo path, this can be determined in less than 1/10 of the time compared to conventional methods.Therefore, we can expect an improvement in the ability to follow changes in the echo path. It can also be applied.

Roughly speaking, if the constant delay of the echo path is taken into consideration when calculating the average power of the received signal, it can also be applied to an echo main canceler that cancels echoes due to echo paths with large delays.

[Brief explanation of drawings]

FIG. 1 is a block circuit diagram for explaining the echo canceller of the present invention, FIG. 2 is a block circuit diagram for explaining a conventional echo canceller, and FIG. 3 is an echo path change constituting the echo canceller of the present invention. FIG. 2 is a block circuit diagram showing a specific example of the configuration of a detector. 76...Subtractor 71.75...Squaring circuit 80...Comparison/judgment circuit 81.82.83...Comparison circuit 84...AND (A N D) circuit SW-... Sui Nchi. 10--Telephone, 12--Subscriber line 16
... Digital to analog (D/A) converter + 8-
- Analog to digital (A/D) converter 20...
Echo cancellation circuit (or subtractor) 22...Adaptive digital filter (ADF) 24-...Double talk detector (
DTD) 30--Echo canceller, 35--Calculation circuit 40--Echo path change detector 50--Power ratio calculation circuit, 51.52--Power calculation circuit 53.54--Logarithmic conversion circuit , 60 - Average calculation circuit 61.72 - Accumulator (ACC) 62.73 - Latch circuit (D)

Claims (1)

[Claims] (1) An adaptive digital filter that inputs the received signal and generates a pseudo-echo signal, and sends out a residual signal in which the echo signal is eliminated by subtracting the pseudo-echo signal from the transmitted signal to which the echo of the received signal is added. In an echo canceller having an echo cancellation circuit and a double talk detector that detects a double talk state from the power ratio of the residual signal and the received signal and stops the estimation operation of the adaptive digital filter, the double talk detector is configured to When a talk state is detected and a change in the echo path is detected while the estimation operation is stopped, an estimation start signal is output to the adaptive digital filter for restarting the estimation operation of the adaptive digital filter whose estimation operation has been stopped. an echo path change detector is provided, the echo path change detector includes a calculation circuit that calculates a power ratio of the received signal and the transmitted signal, and calculates an average value and a variance value of the power ratio; is smaller than a threshold value, the variance value is smaller than the first threshold value, larger than a second threshold value, and the average value is larger than a third threshold value, it is determined that there is an echo path change, and the estimation start signal is output. What is claimed is: 1. An echo canceller comprising: a comparison/judgment circuit;