JPS6315537A - Echo canceller - Google Patents

Abstract

PURPOSE:To eliminate the need for a training signal at the start of an echo canceller and to suppress echo or howling due to the fluctuation of an echo path by giving the loss of a voice switch of a degree not causing howling at the start of application of the echo canceller and controlling the canceller adaptively. CONSTITUTION:The voice switch 7 has a loss not causing howling at the start of the application of the echo canceller 5. An echo path loss arithmetic means 9 obtains the loss of the echo path including the echo canceller 5, an insertion loss arithmetic means 10 operates the insertion loss to be inserted, a low reduction means 13 decreases the stored loss in a range not exceeding the converging speed of the echo canceller, a maximum value selection means 12 compares the operated echo path loss with the decreased loss, stores lose which is larger in a loss storage means 11 as a loss and a margin addition means 14 provides howling margin. As the application of the echo canceller is advanced, the process above is repeated to decrease the insertion loss. When the detector 8 detects howling at the start of the fluctuation of the echo path, the storage means 11 replaces the loss into that at the start of application.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to an echo canceling device for controlling echoes in long-distance circuits such as satellite communications, teleconference systems, speakerphones, and the like.

2. Description of the Related Art In recent years, with the rapid increase in long-distance lines, especially satellite communications, and the practical use of teleconference systems, clear and natural voice transmission has been desired. Echo signals are a phenomenon in which a signal uttered by a speaker returns to the speaker due to reflection on the receiving side, etc., but in teleconference systems and speakerphones that include long-distance lines and acoustic systems, echoes due to reverberation occur. Deterioration of call quality occurs due to noise and howling. Recently, adaptive echo cancellers that cancel this echo are being put into practical use, but since these devices identify the echo path based on the far-end speaker's voice, they can be used at the beginning of a call or when the echo path fluctuates. Because the echo canceller does not cancel the echo signal sufficiently, a large echo may return or howling may occur.

Therefore, when using an echo canceller, the gain of the line must be lowered at the beginning of a call or when the echo path changes. Therefore, conventional methods have been used, such as combining an echo canceller and a voice switch, or using a training signal at the initial stage of adaptation.

As an example, NEC's Ito et al. have shown an audio transmission device for teleconferences (1981, Telecommunications Institute National Conference 691).

The conventional echo canceling device as described above will be described below with reference to the drawings.

FIG. 3 shows the configuration of a conventional echo canceling device.

In FIG. 3, 1 is a receiving side input terminal, 2 is a receiving side output terminal, 3 is a transmitting side input terminal, 4 is a transmitting side output terminal, 31
is an echo canceller that cancels the echo returning from the receiving side output terminal 2 to the transmitting side input terminal 3. 32 is an audio switch which compares the received input signal R1 and the transmitted output signal Si to determine whether to send or receive a call, and when transmitting, adds loss Loss to the receiving input, and when receiving, adds loss Loss to the transmitting output.
Add s. FIG. 6 shows the configuration of the audio switch. Reference numeral 33 denotes a call state determination means for determining the call state from the reception input signal level Lsi and the transmission output signal level Leo; when receiving a call, switch 34 is set to b side and switch 35 is set to
When transmitting, switch 34 to a side, switch 3 to side a.
5 to the b side - 36 is a receiving side loss means that adds loss Loss to the receiving input, and 37 adds loss Loss to the transmitted signal.
This is a transmission loss means that adds .

The operation of the echo canceling device configured as described above will be explained below.

First, in an initial state in which the echo canceller has not canceled any echo signals, a training signal is input to the reception input terminal 1. The training signal is supplied as a reception input R1 to the echo canceller 31 via the reception side input terminal 1 and the audio switch 32. Further, the signal Ri is sent from the receiving side output terminal 2 to the near-end speaker as a reception signal RO. The signal Ro passes through the echo path and is sent to the transmission input terminal 3 at the transmission input Si.
Enter as. In the echo canceller 5, the transmitting side output signal
The series of receiving power R1 and the sending output S0 so that o becomes zero
Create a pseudo echo path from In this way, the echo signal is canceled and the line is then connected.

Thereafter, the echo path is identified using the far-end speaker's voice. If a near-end speaker exists, the echo path cannot be estimated successfully. Therefore, the presence of a near-end speaker signal is detected and estimation of the echo path is stopped.

The voice switch 32 compares the received input signal R1 and the transmitted output signal Eo with the call state determination means 33, determines whether the call is to be sent or received, and when the call is to be sent, the switches 34 and 35 are turned to the b side to reduce the loss on the receiving side. The means 36 is enabled and the transmitting side loss means 37 is disabled. The audio switch 32 is always connected to the reception input terminal 1.
It works to add a predetermined loss between the transmission input terminal 4 and the transmission input terminal 4, suppressing howling and echo when the echo path fluctuates.

Problems to be Solved by the Invention However, in the above configuration, it is difficult to adapt by using a training signal at the initial stage of adaptation of the echo canceller.
If the echo path changes and the loss of the voice switch is insufficient, the echo cancellation of the echo canceller may deteriorate, resulting in more echoes or howling. This had the disadvantage that simultaneous communication was impaired.

In view of the above problems, the present invention provides an echo cancellation device that does not require a training signal at the initial stage of adaptation of the echo canceller, can suppress echoes and howling caused by fluctuations in the echo path, and does not impair simultaneous two-way communication performance. This is what we provide.

Means for Solving the Problem To achieve this objective, the echo canceller of the present invention includes an echo canceller, detects the presence or absence of a near-end talker, and identifies the echo path of the echo canceller. a detector, an audio switch, a howling detector for detecting howling, and an echo path loss calculation means for calculating the loss of an echo path (hereinafter referred to as echo path loss) including an echo canceller, and from the calculated echo path loss. Insertion loss calculation means for calculating the loss to be inserted, loss storage means for gradually reducing the insertion loss of the audio switch when the echo canceller is adapting, maximum value selection means, loss reduction means, and howling It is composed of a margin adding means for adding margin.

Effect: With this configuration, the audio switch has enough loss to prevent howling in the early stages of adaptation of the echo canceller. The echo path loss calculation means calculates the loss of the echo path including the echo canceller, the insertion loss calculation means calculates the insertion loss to be inserted, and the loss reduction means stores the loss within a range not exceeding the convergence speed of the echo canceller. The reduced loss amount is reduced, the echo path loss calculated by the maximum value selection means is compared with the reduced loss amount, the larger one is stored as the loss amount in the loss storage means, and a howling margin is added by the margin addition means.

As the adaptation of the echo canceller progresses, the echo path loss increases, and by repeating the above process, the insertion loss can be reduced. Further, in the initial stage when the near-end speaker speaks and the echo path changes due to fC, the two-way simultaneous conversation detector stops the application of the echo canceller. In this way, the near-end speaker does not identify the false telepathic echo path, but in the early stages of fluctuations in the echo path, the number of echoes increases and howling may occur.

Therefore, the howling detector detects the howling and informs the loss storage means of the howling, and the loss storage means replaces it with the initial loss amount of the adaptation. The voice switch determines whether the call is sent or received and inserts a loss. The amount of loss inserted at this time is sent from the margin addition means. In this way, the voice switch eliminates the need for initial adaptation operations, minimizes deterioration in simultaneous two-way communication due to insertion loss, and provides a highly reliable echo cancellation device that does not cause howling. .

EXAMPLE An example of the present invention will be described below with reference to the drawings. FIG. 1 shows the configuration of an echo canceling device in one embodiment of the present invention. In FIG. 1, 1 is a receiving side input terminal, 2 is a receiving side output terminal, 3 is a transmitting side input terminal, and 4 is a transmitting side output terminal. The above configuration is the same as the configuration shown in FIG.

Reference numeral 5 denotes an echo canceller that cancels echoes returning from the receiving side output terminal 2 to the transmitting side input terminal 3.

Reference numeral 6 denotes a two-way simultaneous conversation detector which detects the presence or absence of a near-end speaker, and performs adaptation of the echo canceller 5, input selection of the loss storage means 110, and control of the loss reduction means 13.

This is a howling detector that detects sFi howling and controls the insertion loss of the audio switch 7. 9 is an echo path loss calculation means for calculating the amount of echo loss from the transmission output signal and reception input signal of the echo canceller, and 1° is the echo path loss and echo path loss means when the echo canceller does not cancel any echo signal. Insertion loss calculation means for calculating the calculated difference in echo path loss (difference in logarithmic expression); 11 is loss storage means for storing the amount of loss; 12 is the output of the loss reduction means 13 and the output of the echo path loss calculation means 9. 14 is a maximum value selection means for comparing and selecting the larger one; and 14 is a margin addition means for adding margin to the loss storage means 11. Reference numeral 16 denotes a switch which is switched by the adaptive control signal adp of the two-way simultaneous call detection means 6.

Figure 2 is a block diagram of the echo path loss calculation means in Figure 1.
0 is a transmission signal level detection means for detecting the transmission output signal level LeO of the echo canceller, 21 is a reception signal level detection means for detecting the reception signal level Lri of the echo canceller, and 22 is the ratio of the reception signal level Lri to the transmission signal level Lsi. This is a level ratio calculating means for calculating.

The operation of the echo canceling device configured as described above will be explained below.

First, in the initial stage of adaptation of the echo canceller, the audio switch has enough loss to not cause howling. Assume that you are now in a complete listening state.

The echo canceller has been adapted and the echo path loss calculation means 9
The echo path loss KRL calculated by is also increased. As shown in FIG.
The transmission output signal level Leo of the echo canceller is detected by the reception signal level detection means 21, the reception input signal level Le11 of the echo canceller is detected by the reception signal level detection means 21, and the ratio of the two is calculated by the level ratio calculation means 22 as the echo path loss ERL. The insertion loss calculation means 10 calculates the difference (difference in logarithmic expression) between the echo path loss ERLO and the calculated echo path loss KRLO when no echo canceller is used, that is, the insertion loss instantaneously or in a specific band. Since it is adapted for audio, there are variations in reverberation loss depending on the frequency band. Therefore, the maximum value selection means 12 selects the larger of the output of the loss reduction means 13 and the output of the insertion loss calculation means 10 so as to select the worst value of the variation (maximum value of insertion loss). Since the adaptation of the echo canceller is ON, the switch 16 is turned to the A side. The loss storage means 11 includes a maximum value storage means 12.
The output of (less than the previously stored value) is stored. The margin addition means 14 stores the loss KRL of the loss storage means 11.
Add all the howling margin to MAX and send it to the audio switch.

As adaptation of the echo canceller progresses, the echo path loss increases, and by repeating the above process, the insertion loss Loss can also be reduced.

Next, a case where a near-end speaker utters a voice will be explained. The two-way simultaneous conversation detector 6 stops the application of the echo canceller 6. In this way, erroneous echo path identification by near-end speakers is not performed. At this time, since the adaptation is OFF, the switch 16 is turned to the B side and the contents of the loss storage means 11 are not changed. Further, the loss reduction means 13 also does not operate.

Finally, the case where the echo path changes will be explained.

At the beginning of a change in the echo path, the echo canceller 6 stops adapting. At this time, the operation is the same as a two-way simultaneous call, and an echo is returned because the insertion loss of the voice switch 7 does not increase, but eventually the two-way simultaneous call detector 6 detects that it is a change in the echo path and adapts. resume. When adaptation is restarted, the maximum value selection means 12 selects a loss amount corresponding to the amount of echo cancellation of the echo canceller 6. Later,
Perform the same operation as the initial adaptation. However, if the fluctuations in the echo path are large enough to exceed the howling margin, howling occurs. Therefore, howling detector 8 detects howling,
This is notified to the loss storage means 11, and the loss storage means 11 replaces the loss with the loss amount at the initial stage of adaptation (state in which the echo cancellation of the echo canceller is not performed).

Based on the loss amount LO8SMAX stored in the loss storage means 11 in the above three states (early adaptation of the echo canceller, when the near-end speaker utters, and when the echo path fluctuates), the margin addition means 14 generates howling noise. A margin is added and given to the audio switch as insertion loss Loss. The voice switch 7 determines whether the call is sent or received and inserts the loss sent from the margin addition means 14. In this way, by adaptively controlling the loss of the voice switch, a training signal for the initial adaptation of the echo canceller is no longer required, minimizing the deterioration of simultaneous two-way communication caused by the conventional voice switch, and eliminating howling. This makes it possible to provide a highly reliable echo cancellation device that does not cause any noise.

As described above, according to this embodiment, in the initial stage of adaptation of the echo canceller, the audio switch has enough loss to prevent howling. The echo path loss calculation means calculates the loss of the echo path including the echo canceller, the insertion loss calculation means calculates the insertion loss to be inserted, and the loss reduction means stores the loss within a range not exceeding the convergence speed of the echo canceller. The reduced loss amount is reduced, the echo path loss calculated by the maximum value selection means is compared with the reduced loss amount, the larger one is stored as the loss amount in the loss storage means, and a howling margin is added by the margin addition means. As the adaptation of the echo canceller progresses, the echo path loss increases, and the above process can be repeated to reduce the insertion loss. In addition, at the initial stage when the near-end speaker speaks or the echo path changes, the two-way simultaneous conversation detector stops the adaptation of the echo canceller. In this way, the near-end speaker does not identify the echo path incorrectly, but at the beginning of the echo path variation,
There is a possibility that the number of echoes increases and howling occurs. Therefore, the howling detector detects the howling and informs the loss storage means of the howling, and the loss storage means replaces it with the initial loss amount of the adaptation. The voice switch determines whether the call is sent or received and inserts a loss. The amount of loss inserted at this time is sent from the margin addition means. In this way, the voice switch eliminates the need for initial adaptation operations, minimizes deterioration in simultaneous two-way communication due to insertion loss, and provides a highly reliable echo cancellation device that does not cause howling. .

As described above, in the present invention, in the initial stage of adaptation of the echo canceller, the audio switch has enough loss to prevent howling. The echo path loss calculation means calculates the loss of the echo path including the echo canceller, the insertion loss calculation means calculates the insertion loss to be inserted, and the loss reduction means stores the loss within a range not exceeding the convergence speed of the echo canceller. The reduced loss amount is reduced, the echo path loss calculated by the maximum value selection means is compared with the reduced loss amount, the larger one is stored as the loss amount in the loss storage means, and a howling margin is added by the margin addition means.

As the adaptation of the echo canceller progresses, the echo path loss increases, and the above process can be repeated to reduce the insertion loss. In addition, at the initial stage when the ν echo path produced by the near-end speaker fluctuates, the two-way simultaneous conversation detector stops the adaptation of the echo canceller.

In this way, the near-end speaker will not incorrectly identify the echo path, but in the early stages of fluctuations in the echo path, there will be many echoes and howling may occur. Therefore, the howling detector detects the howling and informs the loss storage means of the howling, and the loss storage means replaces it with the initial loss amount of the adaptation. The voice switch determines whether the call is sent or received and inserts a loss. The amount of loss inserted at this time is sent from the margin addition means. In this way, the voice switch eliminates the need for initial adaptation operations, minimizes the deterioration of two-way simultaneous communication by the voice switch, and provides a highly reliable echo canceller that does not cause howling. Practical effects are like dogs.

[BRIEF DESCRIPTION OF THE DRAWINGS] FIG. 1 is a block diagram showing the configuration of an echo cancellation device in an embodiment of the present invention, and FIG. 2 is a block diagram showing the configuration of an echo path loss calculation means in an embodiment of the present invention. , FIG. 3 is a block diagram showing the configuration of a conventional echo canceling device, and FIG. 4 is a block diagram showing the configuration of the audio switch in FIG. 3. 6...Echo canceller, 6...Two-way simultaneous call detection unit, 7...Audio switch, 8
・・・・・・／・Uring detector audio, 9・・・・・・
Reverberation loss calculation means, 10... Insertion loss calculation means, 11... Loss storage means, 12... Maximum value selection means, 13... Loss reduction means. , 14・
..... margin addition means, 16 .... switch,
2゜...Transmission signal level detection means, 21...
. . . Received signal level detection means, 22 . . . Level ratio calculation means. Name of agent: Patent attorney Toshio Nakao and one other name
Li
, Figure 3

Claims (2)

[Claims] (1) An echo canceller that adaptively identifies the echo path characteristics from the receiving side output signal and the transmitting side input/output signal and cancels the echo signal included in the transmitting side input signal, and near-end communication such as two-way simultaneous calls. a two-way simultaneous conversation detector for detecting the presence or absence of a person and controlling the identification of an echo path of the echo canceller; and an echo path represented by a ratio of a transmitting side input signal of the echo canceller to a transmitting side output signal of the echo canceller. echo path loss calculation means for calculating the loss; insertion loss calculation means for calculating the difference between the echo path loss and the echo path loss in a state where the echo canceller has not canceled the echo signal; an audio switch that adds loss to a signal and applies a loss to a transmitter output when receiving a call; a howling detection means that detects howling; and a loss storage means that stores an amount of loss obtained by subtracting howling margin from the loss applied to the audio switch. , a loss reduction means for reducing the amount of loss stored in the loss storage means when the echo canceller is performing adaptation; and the output of the insertion loss calculation means and the greater of the two inputs of the loss reduction means. a switch for connecting the input of the loss storage means to the maximum value selection means when adaptation is being performed and to the loss storage means when adaptation is stopped; An echo canceling device characterized by comprising margin adding means for adding a howling margin to the loss stored in the means and sending the insertion loss to the audio switch. (2) The echo cancellation effect calculation means includes a transmission signal level detection means for detecting the signal level of the transmission output of the echo canceller, a reception signal level detection means for detecting the signal level of the reception input of the echo canceller, and a transmission output signal. 2. The echo canceling device according to claim 1, further comprising level ratio calculating means for calculating a ratio between the level and the received input signal level to produce an echo canceling effect.