JPS5842345A - Howling preventing device - Google Patents

Abstract

PURPOSE:To effectively prevent howling for a telephone set with speaker and a sufficiently assure the quality of communication signals, by estimating detour characteristics with an echo canceller for the cancellation of the detour signals. CONSTITUTION:At the initial operation of an echo canceller circuit 5, an audio switching circuit 4 interrupts an input signal from a microphone 1 over a prescribed time and gives a training signal from a signal generator 9 to the circuit 5 via a level adjuster 6 in place of the input signal. Thus, the circuit 5 detects the magnitude of change in amount of detoured signals at a hybrid circuit 3 due to the change in line conditions including the initial line connection. A level discriminator 8 detects a signal level control value from the magnitude in this change, drives level adjusters 6 and 7 selectively and controls the level of transmission or reception signals.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an anti-howry device for a telephone connected to a telephone line.

Recently, telephones with a hand 7 type of loudspeaker telephone system have been attracting attention. Jin no Ii's telephone is connected to the line by converting 2@49 through a trap hybrid circuit like conventional telephones, but at this time, the acoustic coupling between the transmitting end and the receiving end and the above-mentioned hybrid circuit There is a problem in that a howling margin is formed due to the rounding characteristic, and the howling margin is reduced. Therefore, in general, a device was devised to provide a level switch in the receiver circuit of a telephone to control the received loudspeaker output signal level by the received input signal level cape. Since the ζrogako 0 means acts as a chair switch, the above-mentioned switch and chio center can reject the transmitted signal or the received signal depending on the level of the transmitted and received signals during two-way communication. This prevents two-way communication, and also causes problems such as the loss of the owns signal. t) The beginning or end of the call signal may be cut off, leading to a decline in call quality.

Therefore, it has been proposed to prevent howling by canceling the signal transmission in the high and low circuits that form the howling rule.

That is, from the transmission signal, reception signal, and O correlation, the above-mentioned high score,
An echo canceller has been proposed that cancels the feedback signal in the high power circuit by estimating the feedback characteristic of the high power circuit and subtracting the pseudo dual transmission signal generated according to the estimated characteristic from the received signal. According to this system, there is no need to voice-switch the call signal depending on the level of the received signal, so the above-mentioned cutting at the beginning and end of a word is eliminated. However, in this echo canceller, the estimated characteristics determined by the correlation of the above signals are successively modified approximations for each of the above signals, so the initial setting and times #! When line characteristics change due to grafting and branching, the estimated characteristics will be disturbed. In other words, a large change occurs with respect to the steady value, and this causes a defective canceling signal to circulate, resulting in a problem of lowering the howling margin and deteriorating the call signal quality. Therefore, when making initial corrections or calls with such echo cancellers! 〉
In order to reduce the amount of howling and i-jingin at the edge, it is conceivable to cause an appropriate loss to the transmitting side or the receiving side. However, even if this loss is given, the loss is generally controlled by detecting the received signal level, so it is either a simple level change in the received signal or a change in the amount of looping caused by line fluctuations, resulting in a decrease in the ring margin. It is difficult to distinguish between howling and howling, which makes it difficult to effectively prevent howling and ensure signal quality.

The present invention has been made in consideration of these circumstances, and its purpose is to provide a simple and simple method that can effectively prevent the occurrence of howling in loudspeaker telephones and also ensure sufficient speech signal quality. The purpose of this invention is to provide a highly practical howling prevention device.

That is, in the present invention, the echo canceler estimates the echo characteristic and cancels the echo signal, and also calculates both sides of the above estimation from the correlation between the reception input signal or the cancellation residual signal and the plurality of sandal points of the transmission signal. The above objective was effectively achieved by obtaining an evaluation of the sequential variation Vt/# parameter for estimating the characteristics, determining the appropriate loss amount of the chair switch, and adjusting the level of the transmitting signal or receiving signal. -It is.

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a schematic configuration diagram of a telephone terminal incorporating an embodiment of the present invention, in which 1 is a microphone as a transmitting end, 2 is a speaker as a receiving end, and 3 is a telephone terminal with a 41s-
This figure shows a hybrid circuit that converts the 29. A transmission signal (transmission signal) inputted through the microphone 1 is guided to the level adjuster (through the audio switching circuit 4). The signal is input to the hybrid circuit S, and is sent out from this hybrid circuit 3 to the line.The received input signal input from the line via the hybrid circuit 3 is input to the echo canceller circuit 5, and is sent to the line from the hybrid circuit 3. Then, the received signal is level-adjusted via a level adjuster 7, and then amplified and outputted by a speaker 2.These level adjusters 6.1 The insertion loss amount is variably set by a level determiner 8 which operates in response to the correlation value output of the echo canceller circuit 5, and the level of the transmitted signal or the received signal is adjusted individually.

The signal generator 9 cooperates with the audio switching circuit 4 to generate a training signal in place of the transmission signal when the echo canceller circuit 5 trains the two-way characteristic estimation process.

That is, during the initial operation of the echo canceller circuit 5, the input signal tm from the microphone 1 is cut off for a predetermined period of time in the audio switching circuit 4Fi, and instead of this, the training signal from the signal generator 9 is sent through the level adjuster 6. It leads to the echo canceller circuit 5. The echo canceller circuit δ uses this training signal tube to operate the echo canceller. The convergence operation at this time can be calculated using, for example, the steepest descent method, in which the input signal sequence is xt! (1y
)＊＊＊(C(Echo ΔSinnolus characteristic hl ”
h(1?) 'Error output te of canceled signal
, then JT is the sangringta j++
j j h, =h, +Δh.

Represented by a turtle. The second term component in this equation becomes the amount of correction of the pseudo echo characteristics for each Xanzor. What this formula means is a kind of correlation calculation consisting of the input signal sequence and the error output, so if you calculate the average value, it is not the level change of the received signal, but the initial value! It becomes possible to determine the magnitude of the change in the amount of signal looping in the hybrid circuit 3 due to a change in the line state including the I connection. Moreover,
At the time of initial operation or when recirculating by a call branch, the above characteristic -2 sufficiently approximates the two-way characteristic [7], so the 1 difference output becomes large, and the above correction amount also shows a large change. become. The level determiner 8 determines the correction level from the cumulative value over the interval of the correction cover.

That is, the signal rail control value is determined and the level adjuster 6.
7 is selectively driven, and the level of the transmitted or received signal is controlled by setting the above control value. still,
During training at the initial stage of line connection, the signal level on the receiving side is adjusted, and the above corrections are made to the reception state after training and during calls, or the adhesion is only monitored without making any corrections. In this state, signal level adjustment is performed on the transmitting side.

In this way, the echo canceller circuit 5 estimates the characteristics of both ground and cancels the two transmitted components, and also performs level adjustment according to the correlation between the transmitted signal and the received signal or error output, and eliminates the howling. will be effectively prevented.

Now, this device that performs such howling prevention processing,
For example, it is constructed as shown in FIG. 2 and FIG. 3.

In Fig. 2, an input signal x1, which is a transmission signal or a training signal, is guided to a hybrid circuit 3 via level adjustment @gt, and is also connected to a shift register (SR) with n stages.
77,, J Hull 11. , and is taken out over a plurality of sample sections I'14) as tap outputs of each stage. On the other hand, - the received input signal y1 including the received component of the training signal via the hybrid circuit 3 and the both-ground components of the transmitted signal is subtracted using the estimated value y of the pseudo echo signal, which will be described later, as a subtraction signal. The signal is input to the device 12. This lion reducer 1
In step 2, the estimated value y1 is subtracted from the received signal y, and a signal in which both the ground components have been canceled is obtained as an output. The output signal of this subtracter 12 is
The signal is guided to the receiving end via the level adjuster 7.

On the other hand, during the operation of estimating the pseudo echo path characteristics, the subtracter J2
The canceled residual signal generated as the output of Correlation values are being determined. These correlation values are input to the cumulative adders 141*14x to 141 at each sample timing, and are accumulated over several sandal intervals. As a result, the W horizontal adder 14 (r14z to 1411K stores the corrected tag gain value for each sample representing the in-AJ las characteristic of the flaw echo path. The value is multiplied by @I#1*ll~16!IK, and the input signal over the multiple sample intervals described above!,
After being multiplied by the tag output of each, the square sum circuit 16
is input to and subjected to convolution operation. The estimated value 9N of the pseudo echo signal according to the tag gain value of the input signal XB is obtained by the convolution operation by the summation circuit 16,
Ru. An estimated value 1i of the pseudo signal is generated from the input signal X according to the estimated value of the recirculation characteristic of the hybrid circuit 8 indicated by the tag gain value, and this is given to the subtraction Ijll JK.

Then, when the tap gain value converges and the interference characteristic is estimated based on this, a pseudo echo signal 9K is generated according to this characteristic and cancels the interference signal contained in the received input signal yi. will be held. In addition to modifying the tag gain only during the training period, it is also possible to modify the tag gain during the call period, during which the characteristics of the pseudo echo path are estimated and the cancellation of the -C and C transmission signals is performed according to the estimated characteristics. *X of the above tag gain,
, -) or υ estimation characteristics may of course be modified. Also, instead of the residual signal e, the received input signal F
It goes without saying that it is possible to estimate the spidering characteristic (pseudo echo path characteristic) from the correlation between 1 and the human input signal Xi in a plurality of sample sections. In addition, without using the above-mentioned special training signal as an input signal for tap correction, it is also possible to directly use the Tsuzume audio signal to estimate the rounding characteristic Ω and perform a thousand corrections. Of course it's a good thing.

Now, the correlation value between the tag output of the input signal This shows that. Therefore, +13M~ for these multipliers 13
Each output of 13n is led to the absolute value summation path 77, and after the summation of the absolute values of the correlation values is determined, cumulative addition number 1
1, the data are accumulated over multiple sample sections and averaged. As a result, an approximate value of the correlation value indicating the amount of correction is obtained. The approximate correlation value indicating the amount of correction is guided to the discrimination circuit 1#, and is used as a signal indicating the signal level control value corresponding to the value, that is, the gain α (0 to 1)! are applied to the multiplexers xo and xi, respectively. These multiplexers 20.21 input a fixed level setting value (control value) called gain rlJ to the other terminal, and adjust the gain, t9, etc. according to the level determination result of the con/crater 2j. The level control value is selectively extracted and applied to each level adjustment elks*r.

The converter/4 converter 22 inputs the transmission input signal GA and the reception input signal 1, and determines the magnitude relationship between their levels.

Therefore, for the transmitted signal level X, the received signal level y
is small, it is determined that this is a transmitting state, and the multi-level adjustment unit 20.21 applies a level control value of the welfare gain rlJ to the level adjuster 6 on the transmitting side, and at the same time adjusts the level on the receiving side. The level control value corresponding to the gain α is given to the unit 7.

This will impart an appropriate loss to the received signal. In addition, even if the transmitting signal level and the receiving signal level are the same, if the above correction amount changes, the gain α is given to the level adjuster 1 to add a loss to the received signal and increase the knowing margin. ensure that On the other hand, if the transmitted signal level is lower than the received signal level, it is determined that a call is currently being received, and a control operation is performed to cause a loss to the transmitted signal. By the way, Confrey! 22 level comparison,
This may be done by directly comparing the signal levels, but it goes without saying that the levels may be compared by adding appropriate weights such as 0.5x and y, for example.

Further, it is natural that the number of addition samples in the cumulative addition circuit 18 is shorter than the delay of the loop and shorter than the convergence time of the estimated characteristic (tap gain).

In this way, the rotation characteristic is estimated, the rotation signal is canceled, and the above estimated characteristic is corrected,
By adding loss to the transmitted and received signals in accordance with the amount of correction of the estimated characteristic value, the output of large cancellation residuals at the initial operation of the device and when the line is turned during a call can be effectively suppressed, resulting in howling. Mijin will be secured.

By the way, in the above explanation, the multiplication 11hll*+1h~
13.0 The sum of the absolute values of the correlation values which is the output is calculated by the total value total profit circuit JFK, as shown by the broken line in Fig. 2.
The square summation circuit 11 may be used to obtain the summation of the squared values of the respective correlation values, and this may be used as the evaluation value. This can be said to perform average power processing, whereas the former is peak detection processing. Therefore, it is preferable to perform I&nt using the sum of absolute values when detecting a sudden change in the amount of correction, and to use the sum of squares when detecting a quasi-stationary change.

In addition, when the amount of abrasion of a tag having a large value, which is the main signal component of the pseudo machining impulse characteristic, that is, the rounding estimation characteristic, is large, the residual signal and the input signal as shown in Fig. 2. When we calculate the correlation value and estimate its characteristics, we find that the main tag position is summation circuit 1'1.23 summation sun 7"
It may come off from the A-slot. If such a problem occurs, it becomes very difficult to adjust the level, calculate the level control value given to the devices 6 and 1, and detect the level. Therefore, when the delay time is approximately known, the multiplication-sum processing described above may be performed at a position corresponding to the delayed sound. For example, when the main tag exists near the llfth tag, its [
It is preferable to perform multiplication and summation processing using the correlation value of E to obtain the level control value α.

As described above, as shown in FIG. 2, an appropriate loss to be imparted to the transmitted and received signals is determined from the correlation value between the transmitted signal and the received signal, and by providing this loss, a howling margin is secured.

However, this kind of signal processing is equivalent to using a pseudo correlation value between the transmitted signal and the received signal (residual signal), so in the case of strict reliability processing, 31st
The apparatus may be configured as shown in 1. That is, the 2nd H.K.
In the configuration shown, the value of the residual signal .
Since the signal is multiplied by the xs O tap output, if the main signal component of the impulse characteristic of the echo path has a delay, it becomes difficult to determine this because the detected output level of the sum of the nine multiplications described above is small. Therefore, as shown in FIG. 3, shift registers 121m11g to 11. A shift register 4 having an appropriate number of delay stages may be provided in the preceding stage to compensate for the delay of the main signal component.

In addition, with the mounting shown in FIG. 3, the residual signal e t - one-stage shift register J j
J m t - is configured to be distributed over a plurality of null sections as a tag output corresponding to the tap output of the input signal x1. Then, the multiplier 11 and 133
〜IJIl, for each of these corresponding cod f (input signal zt) and residual signal e(1?
), and calculate the correlation value with [cumulative adder 1 as a correction value].
f, 14H-14. is giving to Then, the summation process to obtain the signal level control value is performed by the multipliers I Jl , J
J, ~13. Since each correlation value has already been obtained by I2 in multiple sample intervals, the sum of these O absolute values is obtained by the absolute value summation circuit 11, and the sum value is directly sent to the discrimination circuit 111. It is configured to give.

In such a configuration, it is sufficient to calculate the correction amount of the main signal component as an evaluation value, so the summation process can be performed over several tags.
If the main signal component 0 group delay of the echo path characteristic is in the second half of fC), there is no need to repeat the process from the beginning of f. It is only necessary to perform the summation process of the correlation values.

As described above, as shown in FIGS. 2 and 3, according to the present device, the gain adjustment for the loss imparted to the transmitted signal or the received signal is performed after the training operation is performed at the initial stage.
The discrimination circuit 1# determines the oiioas for the above learning characteristic.
This will be done by monitoring changes. When the deviation between the estimated characteristic and the rotation characteristic becomes large, this is detected and the level adjuster 6. The gain in IK is lowered, that is, the insertion loss is increased to ensure a howling margin.

In this regard, when learning identification, which is the process of estimating the wrap-around characteristic, is performed using a normal signal such as a transmitting signal, the convergence operation due to correction of the estimated value becomes slow, so the multiplication sum It is desirable to increase the gain adjustment sensitivity of the level adjusters g and y based on the level control value determined by .

Furthermore, if the WAm fluctuation is large, which means that the correction amount fluctuates with a large amplitude, a large amount of unnecessary error signals will be routed to the receiving side.

Therefore, such defects [also in the sense of suppressing]
The above-mentioned control of all profit adjustment (loss) has a remarkable effect. In addition, when adjusting the gain tiim of the level adjustment 41t #F, if the discriminant value t is given as an analog value, the level of the transmitted signal or received signal can be continuously varied, so the estimated characteristics can be improved especially with the call signal itself. This makes it possible to reduce deterioration in call signal quality when corrections are made.

As a guideline for the number of cumulative terms in the cumulative addition circuit 11, the number of cumulative terms in the cumulative addition circuit 11 should be significantly smaller than the above-mentioned delayed rainfall amount of Masai in Kurugu within the range where an integral effect occurs, or the delay amount of the loop should be made as large as possible. If you do this, you can expect stable operation.

As described above, according to the present invention, with a simple device configuration, the turning 114! In addition to canceling the feedback, when making corrections, a signal loss is given according to the amount of correction, so it is possible to suppress or reduce howling extremely effectively.
It has an excellent effect of being able to sufficiently cope with fluctuations due to line branches.

Note that the present invention is not limited to the above embodiments.

For example, the series of signal processing described above may be performed entirely digitally, or may be performed entirely analogously. Of course, it is also possible to perform signal processing in a digital/analog hybrid format for each function. Needless to say, even if correlation processing is performed using the reception input signal of the hybrid circuit 3 instead of the loss residual signal, the difference will be lost.

It is also possible to perform the correlation processing separately for estimating the transmission characteristics using the echo canceller and for generating the level control value. In short, the present invention can be implemented with various modifications without departing from the gist thereof.

[Brief explanation of the drawing]

The drawings show an embodiment of the present invention, and FIG. 1 is a block diagram showing a schematic structure of the armored body, and FIGS. 2 and 3 are diagrams showing the structure of a howling prevention processing circuit, respectively. DESCRIPTION OF SYMBOLS 1...Microphone, 2...Speaker, 3...Hybrid circuit, 4...Audio switching circuit, 5...Echo canceller circuit, 6.7...Level control value, 8...
...Level judge, 9...Signal generator s Jl@a
llB~J1n...Shift register, 12...Subtractor, J sl. 133-13a... Multiplier, 141.14-14.
...cumulative addition circuit, "1.15B-%-15,...
Multiplier, 16... Summation circuit, 17... Absolute value summation circuit, 18... Cumulative addition circuit, 9... Discrimination circuit, z
o, xi... multi-brevna, 22... con i sle -! , 23. Square summation circuit, 24...shift register, 7j1.2M. ~is,...shift register.

Claims (1)

[Claims] (Re) Estimating the loop characteristics in the hybrid circuit of the Marugame structure connected to the line via the hybrid circuit, and inputting nine pseudo signals generated according to the estimated characteristics to the receiving input of the hybrid circuit. an echo canceller circuit that cancels a loop component by subtracting it from the signal; means for determining a signal level control value from the sum of correlation values at a plurality of sample points between the transmitted signal and the received signal from the telephone; and the signal level. A howling prevention device comprising means for selectively adjusting the level of the transmitted signal or the received signal in accordance with a control value. (2) The selective level adjustment of the transmitted signal or the received signal is performed by The howling prevention device according to claim 1, which applies a loss based on the signal level control value to the side where the signal level is lower by comparing the level with the received input signal. (3) Howling prevention device between the transmitted signal and the received signal The howling prevention device according to claim 1, wherein the correlation value at a plurality of Nangle points is obtained by a process of estimating the feedback characteristic of the echo canceller circuit. (4) The estimation of the feedback characteristic is performed by: 2. The howling prevention device according to claim 1, wherein the howling prevention device performs the method by determining the correlation between the received input signal or the received cancellation residual signal from the hybrid circuit and the transmitted signal.