KR20030091362A - Line echo canceller - Google Patents

Abstract

PURPOSE: An apparatus for cancelling a line echo is provided to cancel an echo in only an interval in which echo components exist. CONSTITUTION: A double-talk detector(210) detects whether double-talk is generated from transmission and reception paths. If the double-talk is not generated, a delay estimator(220) receives a transmission signal and an echo signal and estimates a delay component. A delay margin(230) includes a margin value in the delay component estimated by the delay estimator(220). An echo estimator(240) estimates an echo path by the delay component included in a margin value estimated by the delay margin(230) and an adaptive filter. A digital subtracter(250) subtracts only an echo component except for the delay component from the echo path estimated by the echo estimator(240).

Description

Line echo canceller

The present invention relates to a line echo canceller, and more particularly, to an echo canceller configured to perform echo cancellation only for a section in which an actual echo component exists by estimating when the track direction has a long delay component.

In general, reverberation in a communication system is caused by a hybrid impedance mismatch for 4-2 wire conversion. The voice is transmitted to the receiving terminal through the receiving side hybrid, and the echo signal returned by the impedance mismatch of the hybrid should be removed. This is the echo canceller.

In the case of voice communication using the PSTN network, a resonance inevitably occurs in which the caller hears his / her voice after a certain delay time by a hybrid circuit in the switching center. These echoes are offensive to the caller and degrade the quality of the call. Echo cancellation is generally used for adaptive echo cancellation using an adaptive filter. Adaptive echo cancellation is a method of modeling echo paths using adaptive filters and then estimating and subtracting echo components. With adaptive echo cancellation, the amount of computation and memory usage increases in proportion to the length of the echo path modeled by the adaptive filter.

An existing echo cancellation method for a digital network includes a method using an adaptive filter having a length corresponding to the entire echo length (hereafter full tap).

1 is a block diagram showing a conventional echo canceller (full-tap).

Referring to FIG. 1, a double-talk detector 110 confirms whether a voice is simultaneously input from a transmission / reception line, and a return path fed back from a reception path when not in a double talk state. The echo estimator 120 for modeling and estimating, and a digital signal for subtracting the echo signal corresponding to the echo path estimated by the echo estimator 120 from the speech signal input on the transmission path. A subtractor 130 and a non-linear processor 140 which receives an output signal of the digital subtractor 130 and removes a signal of a noise level.

Referring to Figure 1 attached to the conventional echo canceller configured as described above is as follows.

First, the double talk detection unit 110 connects to the transmission path and the reception path, and determines whether the voice is in the double talk state in which both voices are simultaneously input. At this time, only the pure voice signal Sin transmitted is input to the transmission path, and the voice signal or the echo signal Rin is received as the echo path.

The double talk detection unit 110 compares the echo signal on the reception path with the speech signal on the transmission path, and if the magnitude of the echo signal is relatively large, determines that the received speech signal exists, and temporarily suspends the operation of the adaptive filter. .

The double toggle detector 110 drives the echo estimator 120 and the nonlinear processor 130 when the double talk state is not in the double talk state, and the echo estimator 120 estimates the echo path by modeling the echo path from the reception path. do. Here, the echo estimator 120 is an adaptive filter.

The echo estimator 120 transmits the estimated echo path to the digital subtractor 130, and the digital subtractor 130 estimates the echo signal using the estimated echo path from the input transmission voice signal and transmits the estimated echo path. It is subtracted from the audio signal.

When the digital subtractor 130 transmits the transmission voice signal from which the echo signal is subtracted to the nonlinear processor 140, the nonlinear processor 140 removes the noise signal level from the transmission voice signal and outputs the noise signal level.

In the operation of the conventional half-type elimination device, the transmission signal and the reception signal are input, and the simultaneous call situation is determined. Subsequently, in a non-simultaneous call condition, the echo path is estimated using the adaptive filter, which is the echo estimator 120, and the nonlinear processor 140 estimates and subtracts the echo component of the estimated path from the digital subtractor 130. A transmission voice signal is output through the.

The length of the echo is generally about 8 ~ 16ms in the PSTN network. However, as the digital network environment such as VoIP develops, the net delay component is added to the front of the echo path and is increased to 32 ~ 128ms.

However, most echo signals are within 16m-23ms, so in digital network environments such as the Voice of Internet Protocol (VoIP), echo cancellers are required to remove echoes of 128 ms length with delay. Full-Tap) requires up to eight times the amount of computation actually required.

As described above, the conventional echo cancellation apparatus needs to estimate the entire echo path length for echo cancellation, and therefore, there is a problem in that an excessive amount of computation and memory are required to estimate the echo length. In addition, since the length of the adaptive filter becomes long, a decrease in convergence speed inevitably occurs.

Another echo canceller is a method of estimating a position having a maximum value among delayed echo components. This is a problem that it is impossible to estimate the exact net delay echo path because the time delay is estimated to be longer than the actual delay.

SUMMARY OF THE INVENTION An object of the present invention is to provide a track echo canceling apparatus capable of estimating when a track echo has a long delay component and performing echo cancellation only on a section in which an actual echo component exists.

Another purpose is to estimate the position of the maximum value of the echo component, and to compensate for the difference in the margin value with the net delay component from the position, the line echo canceller which includes the margin value in the echo component to remove the echo component The purpose is to provide.

1 is a block diagram showing a conventional line echo cancellation device.

Figure 2 is a block diagram showing a line echo cancellation apparatus according to an embodiment of the present invention.

3 is a graph illustrating an example of positions of a net delay component and a local maximum in an echo path for explaining the present invention.

<Explanation of symbols for the main parts of the drawings>

210.Double torque detector 220 ... Delay estimator

230 Delay margin 240 Echo estimator

250 ... digital subtraction 260 ... nonlinear processor

In order to achieve the above object, the track echo canceller according to the present invention,

A double talk detection unit for detecting whether there is a double talk from a transmission and a reception path;

A delay estimator for receiving a transmission signal and an echo signal and estimating a delay component when the signal is not in a double talk situation;

A delay margin unit including a margin value in the delay component estimated by the delay estimation unit;

An echo estimator for estimating an echo path using an adaptive filter having a delay component including a margin value estimated by the delay margin unit;

And a digital subtraction unit for subtracting only the echo component except for the net delay component from the echo path estimated by the echo estimator.

Hereinafter, with reference to the accompanying drawings as follows.

2 is a block diagram showing a device for removing the echo echo of the track according to an embodiment of the present invention.

Referring to FIG. 2, a double talk detector 210 for detecting a double talk state by receiving a voice signal from a transmission and a reception line, a delay estimator 220 estimating a delay component from an echo path, and the delay estimator A delay path unit 230 including a margin value in the delay component estimated by 220 and a delay component including a margin from the delay margin unit 230 are used to generate an echo path using an adaptive filter. An echo estimator 240 that estimates, a digital subtractor 240 that subtracts the echo signal estimated by the echo estimator, and a nonlinear processor 250 that removes noise components of the speech signal are included.

Referring to the accompanying drawings, a track echo canceller according to an embodiment of the present invention as described above is as follows.

Referring to FIG. 2, the double talk detection unit 210 receives a voice signal transmitted / received by being connected to a transmission and reception line and determines whether or not a simultaneous call is present.

At this time, when the call is not in a simultaneous call state, the double talk detector 210 transmits a control signal indicating that the call is not in the simultaneous call state to the delay estimator 220, the echo estimator 240, and the nonlinear processor 260.

The delay estimator 220 estimates a delay component in the echo path (ie, the reception path). In this case, as shown in FIG. 3, the estimated delay component estimates the position having the maximum value T_Peak in the echo path as the delay component value. When echo cancellation is performed using this delay estimate, echo cancellation performance is degraded because it does not include a part of the transfer function.

To this end, the delay estimator 220 transfers the estimated delay component to the delay margin unit 230. The delay margin unit 230 includes a margin value M in the estimated delay component. That is, when the delay component is estimated from the position of the maximum value in the echo path, delay compensation is performed by including the margin value M in the delay component, thereby delaying the actual echo component having a margin value except the pure delay value T_Delay. It will have an estimated value. This increases or decreases from the maximum value of the echo to find the margin value M corresponding to the difference between the net delay component value T_Delay and the maximum value T_Peak of the echo. Therefore, it is possible to improve the echo cancellation performance.

The echo estimator 240 estimates the echo path in which the actual echo component exists by using an adaptive filter with the delay estimation value including the margin value. The digital subtractor 250 estimates and subtracts the actual echo component from the actual echo path, and finds a margin value that minimizes an error in the subtracted echo component. As an example, a margin value of 64 samples (= 8 ms) is set based on the echo environment provided by G.168-2000, which is the Recommendation for ITU-T Digital Network Echo Cancellation. In addition, the margin value thus set up has an effect of compensating for the estimation error occurring in the delay estimation process.

The nonlinear processor 260 removes the noise component of the speech signal and outputs the noise component.

As described above, the line echo canceller according to the present invention performs an echo canceller that compensates for a net delay component, thereby achieving a large gain in computation amount and memory usage compared to a conventional echo canceller.

In addition, when the network echo canceller is configured, an efficient implementation is possible, and the effect is expected to be great because more channels can be accommodated with the same resource in the multi-channel implementation.

Claims (4)

In the line echo canceller in a digital network environment, A double talk detection unit for detecting whether there is a double talk from a transmission and a reception path; A delay estimator for receiving a transmission signal and an echo signal and estimating a delay component when the signal is not in a double talk situation; A delay margin unit including a margin value in the delay component estimated by the delay estimator; An echo estimator for estimating an echo path using an adaptive filter having a delay component including a margin value estimated by the delay margin unit; And a digital subtraction unit for subtracting only the echo component except for the net delay component from the echo path estimated by the echo estimator. The method of claim 1, And the delay estimator estimates a delay component value from a position of a maximum value from an echo path. The method of claim 1, And the delay margin unit estimates an actual delay component including a margin value corresponding between the position of the maximum value of the echo path and the pure delay value. The method of claim 1, And the digital subtraction unit subtracts the actual echo component from the echo path estimated by the echo estimator to find a margin of minimum error from the echo component.