JPS59151544A - Adaptive type echo cancellor - Google Patents

Abstract

PURPOSE:To obtain a prescribed canceling characteristic to any receiving signal by providing a tap coefficient control circuit controlling a tap coefficient stored in a storage circuit depending on a tap coefficient stored by the storage circuit and an output signal of a correcting arithmetic circuit. CONSTITUTION:A pseudo echo line comprising a receiving signal storage circuit 105, tap coefficient storage circuit 107, and an integration arithmetic circuit 110 forms a pseudo echo signal comprising a receiving signal. A subtraction circuit 109 subtracts an echo signal from the pseudo echo signal and outputs a residual echo signal. A correcting amount arithmetic circuit 108 operates a tap coefficient correcting the amount of a transversal filter constituting the pseudo echo line from the receiving signal and the residual echo signal. An addition correcting circuit 204 corrects the tap coefficient. A tap coefficient correcting circuit 401 controls the tap coefficient stored by the circuit 107 depending on the tap coefficient stored by the circuit 107 and the output signal of the circuit 108.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an adaptive echo canceler for eliminating echo signals in a communication line.

2. Description of the Related Art Conventionally, echo prevention devices have been used as means for preventing communication interference caused by echo signals generated in long-distance communication lines. However, this device is not connected to the line.

Since the off operation is performed, there are drawbacks such as cutting off at the beginning of speech and generation of click noise.

Echo canceling devices have been put into practical use to solve the drawbacks of such echo blocking devices. This device measures the echo path characteristics, synthesizes a pseudo echo signal that approximates the echo signal based on the measured characteristics, and cancels the actual echo signal using the pseudo echo signal. Such an echo canceler does not have the drawbacks of an echo blocker because it does not turn on and off a single line.

As is well known, the adaptive echo cancellation device can be regarded as a type of system identification device. That is, a received signal is used as a signal for system identification, and a pseudo echo path that is almost the same as the echo path of the unknown system is formed (identified).

Here, the received signal needs to have a frequency band that is the same as or greater than that of the echo path to be identified. That is, identification cannot be performed for bands other than the band of the received signal used, and the identification becomes so-called undefined.

In an actual echo cancellation device, when a normal voice signal or white noise is used as an identification signal, stable identification can be performed because it contains the spectrum of the same band as the nine echo paths. However, when applied to a communication line, a signal with a very narrow band compared to the band of nine echo paths may be applied as an identification signal for a long time. That is, continuous sine waves, modem signals, etc. may be transmitted. When such a narrowband signal is used as a received signal, the band of the narrowband signal is identified, but the other bands are undefined. Sunawa Juku.

A correction loop is formed for the band of the narrowband signal and identification is possible, but a correction loop is not formed for other bands and identification is not possible.

On the other hand, in actual echo cancellation devices, there are limitations due to the hardware configuration, and noise (disturbance) may be mixed into the echo signal in the echo path, so it is difficult to identify the pseudo echo path. A calculation error occurs in the process. This calculation error will not cause any major problems if the identification signal includes the spectrum of the entire echo path band.

but.

When the narrowband signal is applied, the following problems occur.

That is, in a frequency band outside the narrow band where there is no signal, a correct transfer function cannot be estimated, and it becomes undefined. Therefore, if estimation is continued for a long time, calculation errors will gradually accumulate and may exceed the dynamic range of the storage circuit that stores the coefficients of the transversal filter that constitutes the pseudo echo path. The impulse model generated in this manner differs from the actual echo path characteristics even in the narrow band, making it impossible to cancel the nine echoes.

SUMMARY OF THE INVENTION An object of the present invention is to provide an adaptive echo cancellation device that ensures predetermined cancellation characteristics for all received input signals and at the same time provides stable operation.

The present invention provides a pseudo echo path that creates a pseudo echo signal from a received signal, a subtraction circuit that takes the difference between one echo signal, and the pseudo echo signal and outputs a residual echo signal.

a correction amount calculating circuit that calculates and outputs a tap coefficient correction amount of a transversal filter that constitutes the pseudo echo path from the received signal and the residual echo signal;

2. A signal stored in a storage circuit for tap coefficients of the transversal filter in the pseudo-echo path.

an adaptive echo canceller comprising: an addition correction circuit that adds a tap coefficient and an output of the correction amount calculation circuit to correct a tap coefficient stored in the storage circuit; a tap coefficient side that controls the tap coefficients stored in the storage circuit depending on the output signal of the correction amount calculation circuit;

It is characterized by having both circuits.

Next, a description will be given with reference to the drawings.

FIG. 1 shows a conventional adaptive echo canceler in block diagram form.

The received signal Xj at time j, which is reversely input to the received signal input terminal 101, is input to the received signal storage circuit 105 and simultaneously sent to the echo path through the received signal output terminal 102. The transmission signal Y at time j input from the echo path to the transmission signal input terminal 103 is sent to the subtraction circuit 109, and is sent to the pseudo echo path 1.
The difference from the pseudo echo signal Qj- created in step 06 is taken. The output signal ej of the subtraction circuit 109 is sent to the transmitting signal output terminal 1.
04 and sent out to the transmission path, and simultaneously sent to a correction amount calculation circuit 108, which will be described later.

A pseudo-reverberation path 106 consisting of a transversal filter convolves the contents of the received signal storage circuit 105 that stores the received signal, the storage circuit 107 that stores the tap coefficients of the transversal filter, and these first circuits 105 and 107. It is composed of an integral calculation circuit 110 that performs calculations. Then, the output signal of the integral operation Δ circuit 110 becomes the pseudo echo signal Yj.

The output signal ej of the subtraction circuit 109 and the reception signal from the storage circuit 105 are applied to a correction amount calculation circuit 108, and the correction amount of the tap coefficient of the transversal filter is calculated. Furthermore, the output of the correction amount calculation circuit 1073 is applied to the addition correction circuit 204 for correcting the tap 0 coefficient.

FIG. 2 is a block diagram of the adaptive echo canceler shown in FIG. 1, in which the learning identification method is used as a modification algorithm.

In FIG. 2, the correction amount calculation circuit 1o8 is a multiplier 201.
, 203. Divider 202. It is constituted by a square accumulation circuit 205 and calculates the following equation (1).

Also, based on equation (2), the addition correction circuit 204 corrects the coefficient.

Here, i represents the first tap, and α is the amount of correction? , N
indicates the number of taps of the transversal filter.

If we follow the learning identification correction algorithm above.

When a normal voice signal, white noise, or the like is used as the identification signal, sufficiently stable identification can be performed because it contains the spectrum of the same band as the two echo paths. However, when a signal having a very narrow band compared to the echo path band, such as a continuous sine wave or a modem signal, is transmitted as an identification signal, calculation errors may cause unstable operation.

In contrast, two major features of the present invention are that the correction amount calculation circuit 1
We focused on the fact that the calculation errors that occurred in 08 were very small in the initial stage, and we made sure that calculation errors did not accumulate in the same direction, so that we could provide stable operation for all identification signals. Figure 3 shows a basic embodiment of the invention.

This device differs from the configuration shown in FIG. 2 in that the contents of the tap coefficients of the transversal filter depend on both the storage signal of the storage circuit 107 that newly stores the tap coefficients and the output signal of the correction amount calculation circuit 108. A tap coefficient control circuit 401 is provided at the output of the memory circuit 107 for controlling the circuit so as not to run out of control in one direction.

FIG. 4 is a first diagram specifically showing the tap coefficient control circuit 401.
This is an example.

The polarity signal sgnJ (j-1) of the tap coefficient h1 (j-D stored in the storage circuit 107 and the polarity signal sgnAJ (j-1) of the output signal Δh, (j-1) of the correction amount calculation circuit 108
-1) and are applied to the exclusive OR (goo) 503. And if the polarities of these signals are the same, then K
A selection circuit 50 that selects l and 2 if the polarity is different.
2. It still includes a multiplier 501 that multiplies the output of this selection circuit 502 by the tap coefficient hi (j-1) stored in the storage circuit 107.

Here, Kl and 2 are constants and 2>Kl.

With such a configuration, if the signal polarity of the storage circuit 107 and the signal polarity of the correction amount calculation circuit 108 are the same, the signal h1 obtained by multiplying the tap coefficient h1 (j-1) of the storage circuit 107 by the constant Kl (Apply j−+f to the adder circuit 204,
Conversely, if the polarities are different, the addition circuit 204 includes the storage circuit 10.
A signal obtained by multiplying a constant by 2 is applied to the output of 7. That is, the tag coefficients are controlled to reduce calculation errors.

This makes it possible to prevent the tag coefficients of the transversal filter from going out of control in one direction due to the accumulation of calculation errors.

FIG. 5 shows a second embodiment of the tag coefficient control circuit 401 of the present invention.

That is, the tag coefficients stored in the storage circuit 107 are (
The polarity signal sgnJ (j-1) of j-1), the output signal Δh of the correction amount calculation circuit 108, and the polarity signal sgnΔhI(j-1) of (j-1) are connected to an exclusive OR gate 50.
It leads to 6. Also, the output of the memory circuit 107 is divided into two branches jL, and one is set as a constant! A first multiplier 504 that multiplies a constant by 2 is connected to the other, and a second multiplier 505 that multiplies a constant by 2 is connected to the other. Furthermore, a selection circuit 507 is provided which selects one of the outputs of the multipliers 504 and 505 according to the output of the multiplier 506.

Similar to the previous embodiment, the selection circuit 507 operates in a manner similar to the previous embodiment.9 For example, if the signal polarity of the memory 1 circuit 1079 and the signal polarity of the correction amount calculation circuit 108 are the same, the addition circuit 204 multiplies the output signal of the memory circuit 107 by a constant. The output from multiplier 504 is applied. Of course.

K1<K2, and the effect is exactly the same as in the above embodiment.

As can be seen from the above description, even if a narrowband signal is applied as a reception signal for a long period of time, calculation errors will not be accumulated, and an adaptive echo cancellation device that provides sufficiently stable operation can be realized.

[Brief explanation of the drawing]

Figure 1 is a block diagram of a conventional adaptive echo canceler;
1, FIG. 3 is a block diagram of an adaptive echo canceller according to the present invention, and FIG. 4 is a block diagram of the correction amount calculation circuit of FIG. FIG. 5 is a diagram showing a second embodiment of the present invention. In the figure, 105... Reception signal storage circuit, 107... Tag coefficient storage circuit, 109... Subtraction circuit, 110... Integral calculation circuit 1, 201.20"3... Multiplier, 202...
. . . Divider, 204 . . . Adder, 205 . . . Square accumulation circuit. 1106... Pseudo echo path, 401... Tap coefficient control circuit, 501, 504, 505... Multiplier, 502
. 507...Selection circuit. -2:

Claims (1)

[Scope of Claims] 1. A pseudo-echo path for creating a pseudo-echo signal from a received signal; a subtraction circuit for taking the difference between the echo signal and the pseudo-echo signal and outputting a residual echo signal; a correction amount calculating circuit that calculates and outputs a tag coefficient correction amount of the transversal filter that constitutes the entire pseudo echo path from the residual echo signal;
4) Adaptive echo cancellation having an addition correction circuit that adds the tag coefficients stored in the storage circuit for letter tap coefficients and the output of the correction amount calculation circuit to correct all the tag coefficients stored in the storage circuit. Adaptive type apparatus characterized in that the apparatus includes a tap coefficient control circuit that controls the tag coefficients stored in the storage circuit depending on the tag coefficients stored in the storage circuit and the output signal of the correction amount calculation circuit. Echo canceller. 2. In the adaptive echo canceling device according to claim 1, the tap coefficient control circuit is configured to depend on the polarity of the tap coefficient stored in the storage circuit and the polarity of the output signal of the correction amount calculation circuit. and a multiplier that multiplies the constant selected by the selection circuit by the tap coefficient stored in the storage circuit. Device. 3. In the adaptive echo canceling device according to claim 1, the tap coefficient control circuit is configured to output a first signal to one of two signals obtained by branching the tap coefficient output stored in the storage circuit.
a first multiplier that multiplies the other by a constant; a second multiplier that multiplies the other by a second constant; the polarity of the tap coefficient stored in the storage circuit; Depends on. An adaptive echo cancellation device comprising: a circuit for selecting the output of either the first or second multiplier. Extras below