JPH0728244B2 - Echo canceller - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to an echo canceller that cancels near-end echo generated in a 2-wire to 4-wire conversion unit in a communication line such as satellite communication and voice packet communication. is there.

(Prior Art) In the current telephone line, a hybrid circuit for converting between a two-line line of a subscriber line and a four-line line of a transmission line is generally used. Due to the mismatch, the received signal leaks to the transmission line and an echo occurs. In a network with a large delay such as satellite communication, this echo interferes with communication, so
An echo canceller is inserted in the transmission line to eliminate the echo. In addition, in an integrated network that handles voice, image, and data in an integrated manner, voice is packetized for communication, which causes a large delay even in domestic lines. Therefore, it is necessary to use an echo canceller to eliminate echo in voice packet communication as well. 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, 12 is a subscriber line, 14 is a hybrid circuit (indicated by symbol H in the figure), 16 is a digital band analog converter (indicated by symbol A / D in the figure), 18
Is an analog-to-digital converter (denoted by A / D in the figure), 20 is a subtractor, 22 is an adaptive digital filter (denoted by ADF in the figure), and 24 is a double-talk detector (DTD in the figure). The adaptive digital filter 22, the subtractor 20 and the double talk detector 24 constitute an echo canceller 26.

The operation of this conventional circuit will be briefly described.

The received signal x (k) from the far end (where k is the sampling time) is converted into an analog signal x (t) by the D / A converter 16.
(However, t is time), is input to the hybrid circuit (H) 14, and is transmitted to the telephone 10 through the subscriber line 12 as a two-line signal. At this time, due to the impedance mismatch of the hybrid circuit (H) 14, the received signal x (t) leaks to the transmission path and is sent out as the echo signal y (t). On the other hand, the near-end input signal n (t) from the telephone 10 is converted into a 4-wire type signal by the hybrid circuit (H) 14 and sent to the transmission path. That is,
The transmission signal s (t) is the sum of the echo signal y (t) and the near-end input signal n (t), and s (t) = y (t) + n (t).
Is expressed as

The echo canceller 26 receives the received signal x (k) and the transmitted signal s converted into a digital signal by the A / D converter 18.
Input (k). The adaptive digital filter 22 of the echo canceller 26 determines the impulse response of the system (hereinafter referred to as an echo path) from the D / A converter 16 through the hybrid circuit (H) 14 to the A / D converter 18 by the learning identification method. Sequential estimation is performed, a pseudo echo signal (t) is generated from the estimated impulse response and the received signal, and the subtracter 20 subtracts this from the transmission input signal s (k) to cancel the echo. Therefore, the subtractor 20 functions as an echo canceling circuit, and the output of the subtractor 20 is the residual signal e (k) = s.
(K)-(k) is transmitted.

However, in the learning identification method, the received signal sequence X _k = (x (k), x (k−1),
, X (k−N + 1)) and the residual signal e (k) are used, the estimated impulse response is disturbed when the estimation is continued in the double talk state in which the near-end signal n (t) is large. Will end up. The double-talk detector 24 detects the double-talk state and outputs the estimation stop signal INH to stop the estimation of the adaptive digital filter 22 and prevent it. As this double talk detection method, for example, the received signal x
Power ratio between (k) and residual signal e (k) When the power ratio A is equal to or smaller than a predetermined threshold value A _T , that is, when the condition of A ≦ A _T (2) is satisfied, it is determined that the double talk state is established. Where A _T is the double talk detection threshold. This method has high double-talk detection sensitivity and can detect even a small level of the near-end signal n (k).

However, even when the echo path is switched to another system, the power of the residual signal e (k) increases, so that (2)
The expression is satisfied, and it is determined that the double talk is actually generated even though the double talk state is not established. As a result, the estimated stop signal INH described above is output from the double talk detector 24 to the adaptive digital filter 22. Will stop the estimation operation in.

In an actual telephone line, the echo path is changed during the switching operation of the telephone from the on-hook state to the off-hook state or vice versa, or during the mute operation, and the echo cannot be canceled.

Therefore, conventionally, as disclosed in Japanese Patent Application Laid-Open No. 61-56526, the threshold value _AT is set to a value smaller than the threshold value at the previous sampling time by a constant value δ _u and a non-double threshold value when the double talk is determined. by setting such a value larger by a predetermined value [delta] _D than the threshold value at the previous sampling time during a talk spurt determination, approach to respond to changes in the echo path is used. That is, the threshold A _T in each case is defined as A _T (k + 1) = A _T (k) −δ _u during double talk determination A _T (k + 1) = A _T (k) + δ _D during non-double talk determination .

According to established how such threshold A _T, to the change in the echo path, when it is determined that the double talk threshold A for the value of _T decreases, a certain time passes and the power ratio A threshold A _T
(A> _AT ), and therefore the condition of the above-mentioned equation (2) is not satisfied, and thus estimation is started. Where δ
_Since the value of _u is very rarely lasting for more than 3 seconds in the actual double-talk state, conventionally, it is determined that A> A _T about 3 seconds after the determination of the double-talk state.

(Problems to be Solved by the Invention) However, in the above-described conventional method, since the change in the echo path is detected by the time of the double talk determination, it takes at least about 3 seconds from the change in the echo path until the change is detected. . Furthermore, if the safety during actual double talk is emphasized, the time for detection must be made sufficiently long, and there is a problem in that the followability of the echo canceller to changes in the echo path is greatly lost.

The present invention was made in order to eliminate the problem that the followability of echo path changes deteriorates when the double talk detection sensitivity of the echo canceller is increased, and therefore the object of the present invention is to improve the double talk detection sensitivity. The object of the present invention is to provide an echo canceller having good followability with respect to echo path changes without lowering the noise.

(Means for Solving the Problem) In order to achieve this object, according to the present invention, an adaptive digital filter for receiving a received signal to generate a pseudo echo signal and an echo of the received signal for the pseudo echo signal are provided. Echo canceling circuit that sends out a residual signal in which the echo signal is canceled by subtracting it from the added transmitting signal, and the double talk state is detected from the power ratio of this residual signal and the received signal to estimate the adaptive digital filter described above. In an echo canceller having a double-talk detector that stops its operation, an adaptive digital filter whose estimation operation is stopped when a double-talk detector detects a double-talk state and a change in echo path is detected while the estimation operation is stopped. The estimation start signal for restarting the estimation operation of The echo path change detector is provided with a calculation circuit for calculating the power ratio between the reception signal and the transmission signal and calculating the average value and the dispersion value of the power ratio, and the dispersion value is the first. Is smaller than the threshold value and the average value is larger than the second threshold value, it is determined that there is an echo path change, and the comparison determination circuit that outputs the above-mentioned estimation start signal is provided.

(Operation) According to the above-described configuration of the present invention, when the double-talk detector detects that the double-talk state is present, the echo path change detection detects the double-talk state. An estimation start signal for restarting the estimation operation of the adaptive digital filter stopped by the detection of the double talk state is output from the echo path change detector to the adaptive digital filter.

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, a variance value and an average value of the calculated power ratio per N sampling time are calculated, and subsequently, in the comparison / determination circuit, the variance value is constantly compared with the first threshold value. And compare the average value with a second threshold. When it is determined that the variance is smaller than the first threshold value and the average value is larger than the second threshold value when the estimation operation is stopped by the double-talk state detection, the comparison / determination circuit sends the adaptive digital filter to The estimation start signal of is output.

As described above, in the present invention, it is monitored whether or not the double talk state detection is caused by the change in the echo path,
The time until the estimated start signal is output due to the change of the echo path does not require a waiting time of 3 seconds like that of a tree, so that it is substantially only the processing time in the echo path change detection circuit. The echo canceller of the present invention can re-start the adaptive digital filter in about 1/10 or less of the conventional time, and remarkably follows an echo path change.

(Embodiment) An embodiment of the echo canceller of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram for explaining the echo canceller, and FIG. 3 is a block diagram showing a circuit configuration of an echo path detector provided in the echo canceller. The first
In the figure, the same components as those shown in FIG. 2 are designated by the same reference numerals, and the detailed description thereof will be omitted unless otherwise specified.

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

The echo path change detector 40 takes in the received signal x (k) and the transmitted signal s (k) and monitors the change in the echo path. When the change in the echo path is detected while the estimation operation is stopped, The estimation start signal for restarting the estimation operation of the adaptive digital filter 22 is output to the adaptive digital filter 22.

Next, the echo path change detector 40 will be described in detail.

In the present invention, the echo path change detector 40 is provided with a ratio calculation circuit 50 for calculating the power ratio of the reception signal x (k) and the transmission signal s (k), and an average calculation for calculating the average value m _AE of the power ratio. A calculation circuit 35 including a circuit 60, a dispersion calculation circuit 70 for calculating the dispersion value σ ² _AE of this power ratio, and a switch SW.
And a comparison / determination circuit 80 that determines the echo path change using the outputs of the average value and the variance value.

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

First, the transmission signal s (k) and the reception signal x (k) are input to the power ratio calculation circuit 50, and the power calculation circuits 51 and 52 calculate the average power of these signals s (k) and x (k) over a short period of time. S
After calculating (k) and X (k), they are calculated according to the following equation (3).

Therefore, the average powers S (k) and X (k) are converted into logarithmic domain values L _s (k) and L _x (k) by the logarithmic exchange circuits (log ₁₀ ) 53 and 54. In the subtractor 55, the difference between these signals (L _x (k) −L _s (k))
The power ratio A _E (k) in the above equation (3) is calculated and output by calculating In the power calculation circuits 51 and 52 described above, the average powers S (k) and X (k) are expressed by the following equation X (k) = ( ^1-2-5 ) * (k-1) as well known in the art. +2 ^-5 X (k) ² (4,
1) S (k) = ( ^1-2−5 ) S (k−1) + ²⁻⁵ S (k) ² (4,
2) can be obtained by

In this embodiment, this power ratio A _E (k) is switched to the switch S
It is supplied to the average calculation circuit 60 and the dispersion calculation circuit 70 via W respectively. The switch SW is configured to be closed by the estimated stop signal INH while the double talk detector 24 detects the double talk state and outputs the estimated stop signal INH. It may be a switch of any form of expression. While the switch SW is closed, therefore, while the double talk detector 24 detects the double talk state and outputs the estimated stop signal INH, the power in the average calculation circuit 60 and the dispersion calculation circuit 70 is reduced. The average value m _AE and variance value σ ² _AE of N sample times of the ratio A _E (k)
Respectively Ask according to. Here, N is a value corresponding to a time of about 100 msec to 200 msec, for example, 800 at 8 kHz sampling.
The value is about 1600 samples.

In this embodiment, in the average calculating circuit 60, first, the accumulator (ACC) 61 calculates the accumulated value of the power ratio A _E (k),
By the pulse T of the period N, the sum of the accumulated value of the Nth sample, that is, the power ratio A _E (k) of the Nth sample section The value of is set in the latch circuit (D) 62, and at the same time, the contents of the accumulator 61 are cleared. Then, the accumulated value set in the latch circuit 62 is sent to the multiplier 63, where it is multiplied by the constant "1 / N" to obtain the average value m _AE of the power ratio, which is compared as the output of the average calculation circuit 60. Judgment circuit 80 and distributed calculation circuit 70
And send it to. On the other hand, in the distributed calculation circuit 70, the above-mentioned switch
While the SW is closed, first the power ratio A in the squaring circuit 71.
_The square value A _E (k) ² of _E (k) is obtained, and this is sent to the accumulator 72, and the sum of the square value A _E (k) ² of N sample intervals is sent to the accumulator 72. Calculate the value of.

Then, the accumulated value is set in the latch circuit (D) 73 by the pulse T of the cycle N and the contents of the accumulator 72 are cleared at the same time as in the case described above.

Then, the accumulated value set in this latch circuit 73 is multiplied by
Send to 74, where there is a constant "1 / N" multiplied by this and the mean square value of the power ratio A _E (k) in the N sample section Is calculated and then sent to the subtractor 76. The subtractor 76 has an average value m of the power ratio output from the average calculation circuit 60 described above.
_The squared value (m _AE ² ) of the average value obtained by squaring the _AE by the squaring circuit 75 is sent. And the squared value of the mean value (m _AE ² ), the variance value Is calculated and sent to the comparison / determination circuit 80.

Using the average value m _AE and the variance value σ ² _AE , the comparison and determination circuit 80 determines whether or not the estimation stop is due to a change in the echo path, and if it is determined to be due to the echo path change, the estimation start The signal ADP is output to the adaptive digital filter 22.

Next, a method of judging an echo path change in this comparison and judgment circuit will be described.

Now, let's consider what the power ratio A _E (k) will be when the impulse response of the echo path is set to 1H.Next, we will discuss the mean value m _AE and dispersion of the power ratio during non-double talk and double talk. Consider the value σ ² _AE . The transmitted signal s (k) is
Since it is the sum of the signal near-end input n (k) and the echo signal y (k), (However, ^T represents the transversion matrix). Transmission signal s (k) given by equation (7)
The average power S (k) of n is calculated as n (k) and x (k)
Is uncorrelated, Power is (4,1) when the echo path delay is small.
Since it can be approximated by the short-time average power X (k) of the equation, Becomes Therefore, from equations (3) and (8), the power ratio A _E (k) between the received signal x (k) and the transmitted signal s (k) is It is represented by.

Therefore, the average power N of the near-end input during non-double talk
Since (k) is a very small value, the power ratio A _E (k) in equation (9) is Becomes The power ratio A _E (k) given by Eq. (10) reflects the attenuation in the echo path (see Fig. 1), and in a telephone circuit, for one echo path, it has a substantially constant positive value. It is a value. That is, the average value m _AE of the power ratio A _E (k) calculated by the equation (5) is a positive value, and the variance value σ ² _AE calculated by the equation (6) is a small value.

On the other hand, at the time of double talk, if reception input and near-end input processing exist simultaneously, x
Since (k) and s (k) are uncorrelated, the temporal change of A _E (k) represented by the equation (9) is large. That is, (6)
The variance value σ ² _AE of the equation has a large value.

Also, if there is only the near-end input, x (k) since as a very small value (9) A _E (k) becomes a large negative value. That is, the average value m _{AE of the} equation (5) is a negative value.

Considering the above points, the variance value σ ² _AE is smaller than a certain threshold value (this is the first threshold value TH1), and the average value m _AE is a threshold value (this is the second threshold value TH2). If it is larger than the above, even if the double talk detector 24 (FIG. 1) detects the double talk state, it can be detected that the change is not actually the double talk but the echo path, and the first threshold TH1, The second threshold value TH2 can be preset respectively. Therefore, when using the first threshold TH1 and the second threshold TH2 set in advance as described above, when σ ² _AE <TH1 and m _AE > TH2 (11) are satisfied, it means that the echo path is changed. It is possible to judge.

Therefore, in this embodiment, the comparison / determination circuit 80 includes a comparison circuit 81 for comparing the second threshold value TH2 and the average value m _AE and a first threshold value TH1.
And a variance value σ ² _AE comparing circuit 82, and both comparing circuits 81 and 82
AND circuit 83 for each result of pulse and pulse T
And with. The comparison / determination circuit 80 having such a configuration
Then, in the comparison circuit 81, the average value m _AE is compared with the second threshold value TH2, and if m _AE > TH2, a logic “1” is output.
If the variance value σ ² _AE <TH1, the comparison circuit 82 outputs a logic “1”. In the AND circuit 83, the logical product of each output of these both comparison circuits 81 and 82 and the pulse T is taken, and the output of this logical product "1" is used as the estimation start signal to the adaptive digital filter 22 (FIG. 1). Print and restart it. Thus, when σ ² _AE <TH1 and m _AE > TH2, it is determined that the cause of stopping the estimation operation is the echo path change, and the estimation start signal ADP (logic "1"
Is output for one sampling period.

It is obvious that the present invention is not limited to the above-mentioned embodiments but can be modified or changed in many ways. For example, the configuration of the echo path change detector described in FIG. 3 does not have to be the configuration of this embodiment, and the average value and the variance value thereof are calculated using the power ratio of the reception signal and the transmission signal,
It suffices that the average value and the variance value are compared with respective predetermined threshold values to output the estimation start signal.

Further, although the components necessary for understanding the present invention are shown in the configuration examples shown in the block diagrams in FIGS. 1 and 3, the clock pulse necessary for operating the echo canceller of the present invention is shown. The generator, the generator of the N-period pulse T, the memory of the sampling period N, the first and second threshold values TH1 and TH2, etc., and other circuits and / or other conventional means that can be easily understood by those skilled in the art are shown in FIG. The description is omitted.

(Effect of the Invention) As is apparent from the above description, according to the echo canceller of the present invention, the circuit for calculating the average and variance of the power ratio of the received signal to the transmitted signal and the double-talk detector are in the double-talk state. Since a circuit for determining that there is a change in the echo path when the variance value is smaller than the first threshold value and the average value is larger than the second threshold value while the estimation operation is stopped is provided, The determination can be performed in about 1/10 or less time compared with, and therefore improvement in followability to changes in echo path can be expected.

Furthermore, if a constant delay of the echo path is taken into consideration when obtaining the average power of the received signal, the present invention can be applied to an echo canceller that cancels an echo due to an echo path with a large delay.

[Brief description of drawings]

FIG. 1 is a block circuit diagram for explaining an 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 forming the echo canceller of the present invention. It is a block circuit diagram which shows the specific structural example of a detector. 10 ... Telephone, 12 ... Subscriber line 16 ... Digital-to-analog (D / A) converter 18 ... Analog-to-digital (A / D) converter 20 ... Echo canceling 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) 63,74 …… Multiplier, 70 …… Dispersion calculation circuit 71 , 75 …… Square circuit, 76 …… Subtractor 80 …… Comparison judgment circuit, 81,82 …… Comparison circuit 83 …… AND (AND) circuit SW …… Switch.

Claims (1)

[Claims] 1. An adaptive digital filter for inputting a received signal to generate a pseudo echo signal, and a residual signal obtained by canceling the echo signal by subtracting the pseudo echo signal from a transmitted signal to which an echo of the received signal is added. An echo canceller having an echo canceling circuit for transmitting and a double talk detector for detecting a double talk state from the power ratio between the residual signal and the received signal to stop the estimation operation of the adaptive digital filter, Detects a double-talk state and detects a change in the echo path while the estimation operation is stopped, an estimation start signal for restarting the estimation operation of the adaptive digital filter whose estimation operation has stopped is applied to the adaptive digital filter. And an echo path change detector for outputting to the echo path change detector. Of a power ratio between a signal and a transmission signal, and a calculation circuit for calculating an average value and a variance value of the power ratio, the variance value being smaller than a first threshold value, and the average value being larger than a second threshold value. An echo canceller, comprising: a comparison / determination circuit that determines that there is an echo path change and outputs the estimation start signal.