CN1219391C - Echo eliminator and method for evaluating echo - Google Patents

Abstract

The invention discloses an echo canceller, which includes an adaptive filter, a far-end input signal buffer, a near-end input signal buffer and a signal adder, and adds a far-end signal down-sampling processing module and a near-end signal down-sampling processing module and the cross-correlation echo effective part search module; the cross-correlation echo effective part search module performs cross-correlation operation on the signal down-sampled by the far-end signal down-sampling processing module and the near-end signal down-sampling processing module to obtain the echo effective parameters, and adaptive filtering The echo estimation device performs echo estimation on the received signal according to the effective parameters of the echo to generate an echo estimation value; the signal adder subtracts the echo estimation value from the input signal mixed with the echo signal received from the outside of the echo canceller to perform echo cancellation. At the same time, a method for echo estimation using the echo canceller is disclosed. The application of the scheme of the invention not only does not affect the effect of echo cancellation, but also improves the efficiency and can greatly save the system cost.

Description

An echo canceller and echo estimation method

technical field

The invention relates to echo processing technology in a communication system, in particular to an echo canceller and an echo estimation method.

Background technique

In the fixed telephone communication system, the switchboard uses a four-wire system, while the user's telephone line uses a two-wire system. In this way, a converter (Hybrid) is needed to realize the transmission of signals from the switchboard to the user during the transmission process. transmission. Since the impedance of the four-wire system and the two-wire system cannot be accurately matched, the signal will be reflected during transmission, and when the reflected signal reaches the user side, the user will hear his own voice returning, forming a echo. Echo will greatly reduce the call quality, and it is a problem that must be solved in the communication system. As shown in FIG. 1 , the signal at terminal A passes through the converter 101 and returns to terminal A, so that user A hears his own echo.

At present, the echo cancellation function can be realized by using an echo canceller. Fig. 2 is the structural block diagram of prior art echo canceller, as shown in Fig. 2, prior art echo canceller comprises an adaptive filter 202, a far-end input signal buffer 201, a near-end input signal buffer 203 and a signal adder 204 . The far end refers to the four-wire end, and the near end refers to the two-wire end. Its core is the adaptive filter 202, by using the adaptive filter 202 to simulate the transfer characteristics of the converter, this adaptive filter 202 refers to the far-end input signal and the near-end input signal sent by the far-end input signal buffer 201 The near-end input signal sent by the cache 203 and the near-end output signal as the feedback signal of the adaptive filter are used to estimate the value of the echo. The function of the adaptive filter 202 is to simulate an environment similar to a real reflection path (echopath) , and how long it can make the echo estimated by itself close to the real echo, the difference between the estimated echo and the real echo is less than a specified value, and then the mixed signal received from the outside of the echo canceller is passed through the signal adder 204 Echo cancellation is performed by subtracting the estimated echo value from the echo signal. The smaller the difference between the estimated echo and the true echo, the better the echo cancellation will be.

Figure 3-1 and Figure 3-2 are schematic diagrams of the principle of echo generation. Among them, Figure 3-1 is a schematic diagram of sending an ideal pulse excitation signal, and Figure 3-2 is a schematic diagram of the response of the ideal pulse excitation signal sent in Figure 3-1. For point A in Figure 3-1, at point A An ideal impulse signal sent out will get the response in Figure 3-2, where tp is the net echo part, that is, the effective part of the echo, tr is the pure delay part, and td is the total echo delay. As shown in Figure 3-2, the net echo part accounts for a relatively small part of the total echo delay. It can be seen that if the net echo part can be effectively canceled, the echo cancellation can be realized. However, the current echo cancellers all cancel the signal of the whole time td segment. Obviously, the result of doing this is to bring a lot of waste of system resources.

It can be seen that the main disadvantage of the existing technology is that the characteristics of the echo are not considered when performing echo cancellation, and the signal of the entire time period of the echo is directly canceled, so too many resources are required, which makes it impossible to improve the utilization of system resources and reduce costs. .

Contents of the invention

The basic principle generated by the echo shows that the effective part of the echo only occupies a small part of the entire echo length. If this part can be found, the effect of echo cancellation can be realized with less resources, which can save a lot of system cost .

In view of this, the main purpose of the present invention is to provide an echo canceller, which can find the effective part of the echo, and only cancel this part of the echo, which not only does not affect the effect of echo cancellation, but also improves the efficiency and can save a lot system cost.

Another main purpose of the present invention is to provide an echo estimation method, which can efficiently and accurately obtain an echo estimation value.

According to an aspect of above-mentioned object, the present invention provides a kind of echo canceller, it comprises adaptive filter, far-end input signal buffer, near-end input signal buffer and signal adder; This far-end input signal buffer will receive The far-end input signal is sent as a reference signal to the adaptive filter; the near-end input signal buffer sends the received near-end input signal to the adaptive filter; the adaptive filter generates an echo estimate and sends it to the signal Adder; the signal adder subtracts the echo estimated value from the input signal mixed with the echo signal received from the outside of the echo canceller to perform echo cancellation, and generates a near-end output signal, and sends the near-end output signal as a feedback signal to an adaptive filter; the echo canceller further comprising:

The remote signal down-sampling processing module, which performs down-sampling processing on the remote input signal received from the remote input signal buffer, and sends the processed signal to the cross-correlation echo effective part search module;

A near-end signal down-sampling processing module, which performs down-sampling processing on the near-end input signal received from the near-end input signal buffer, and sends the processed signal to the cross-correlation echo effective part search module;

Cross-correlation echo effective part search module, which performs cross-correlation calculation on the above two processed signals received, finds the echo maximum value from the calculation result, and selects the farthest distance on the left and right sides of the echo maximum value according to the communication protocol The part of the echo is used as an effective part, and it is sent to the adaptive filter as an effective echo parameter;

The adaptive filter receives the signal sent by the far-end input signal buffer and the near-end input signal buffer as well as the near-end output signal, and receives the effective echo parameters, and finds the effective echo part in the received signal according to the effective echo parameters. Echo estimation is performed on the effective echo part to generate an echo estimation value.

Wherein, the far-end signal down-sampling processing module or the near-end signal down-sampling processing module may further include a far-end signal low-pass filter or a near-end signal low-pass filter, which are respectively used to cache the far-end input signal or the near-end input The far-end input signal or the near-end input signal received by the signal buffer is low-pass filtered, and then down-sampled.

The far-end signal low-pass filter and the near-end signal low-pass filter may respectively be low-pass filters with a cutoff frequency of 1000 Hz.

According to another aspect of the above-mentioned purpose, the present invention provides a method for echo estimation, which uses the above-mentioned echo canceller to perform echo estimation, and the method includes the following steps:

1) The far-end signal down-sampling processing module and the near-end signal down-sampling processing module respectively down-sample the far-end input signal and the near-end input signal;

2) Send the two down-sampled signals to the cross-correlation echo effective part search module to perform cross-correlation operations;

3) The cross-correlation echo effective part search module finds the maximum value of the echo from the cross-correlation results, and according to the communication protocol on the left and right sides of the maximum echo value, takes the part with the farthest distance as the effective part of the echo, and uses it as the effective parameter of the echo, and sends to the adaptive filter;

4) The adaptive filter finds the effective echo part in the received signal according to the effective echo parameters, performs echo estimation on the effective echo part of the received far-end input signal, near-end input signal and near-end output signal, and generates an echo estimate value. Wherein, step 1) of the method may further include: performing low-pass filtering respectively before down-sampling the far-end input signal and the near-end input signal. Step 3) The method for finding the maximum value of the echo may be: searching for the maximum value of the echo by comparing the amplitude values of each point of the cross-correlation result.

As can be seen from the above-mentioned scheme, the key of the present invention is: firstly down-sampling the far-end input signal and the near-end input signal respectively; then performing cross-correlation calculation on the above-mentioned signals; searching out the effective part of the echo from the cross-correlation result, and finding Echo effective parameter; the adaptive filter estimates the echo value based on this parameter. Therefore, this echo canceller and echo estimation method of the present invention can accurately locate the effective echo part, and only perform echo cancellation on the effective part of the echo, which not only does not affect the effect of echo cancellation, but also improves efficiency and saves a lot system cost. Since the present invention adopts the down-sampling method, the amount of computation and data storage can be reduced when performing delay search.

Description of drawings

Fig. 1 is a schematic diagram of a system that generates an echo in a prior art communication system;

Fig. 2 is the structural block diagram of prior art echo canceller;

Figure 3-1 is a schematic diagram of sending an ideal pulse excitation signal;

Figure 3-2 is a schematic diagram of the response of the ideal pulse excitation signal sent in Figure 3-1;

Fig. 4 is a structural block diagram of a preferred embodiment of the echo canceller of the present invention;

5 is a schematic diagram of the external application of the echo canceller of the present invention;

Fig. 6 is a schematic diagram of the principle of cross-correlation calculation in the present invention;

Fig. 7 is a schematic diagram of the echo estimation process in the present invention.

Detailed ways

In order to make the object, technical solution and advantages of the present invention clearer, the following examples are given and the present invention is further described in detail with reference to the accompanying drawings.

Fig. 4 is the structural block diagram of a preferred embodiment of the echo canceller of the present invention, as shown in Fig. 4, the echo canceller of the present invention comprises: an adaptive filter 402, a far-end input signal buffer 401, a near-end input A signal buffer 406 , a far-end signal down-sampling processing module 403 , a near-end signal down-sampling processing module 405 , a cross-correlation echo effective part search module 404 and a signal adder 407 .

Wherein, the far-end input signal buffer 401 sends the received far-end input signal to the adaptive filter 402 as a reference signal. The near-end input signal buffer 401 sends the received near-end input signal to the adaptive filter 402 . The remote signal down-sampling processing module 403 performs down-sampling processing on the remote input signal received from the remote input signal buffer 401, and the remote signal down-sampling processing module 403 may also include a remote signal low-pass filter. A low-pass filter with a cutoff frequency of 1000 Hz is adopted. This module first performs low-pass filtering on the remote input signal received from the remote input signal buffer 401, then performs down-sampling processing, and sends the processed signal to Cross-correlation echo effective part search module 404 . The near-end signal down-sampling processing module 405 performs down-sampling processing on the near-end input signal received from the near-end input signal buffer 406, and the near-end signal down-sampling processing module 405 may also include a near-end signal low-pass filter. A low-pass filter with a cutoff frequency of 1000 Hz is adopted. This module first performs low-pass filtering on the near-end input signal received from the near-end input signal buffer 401, then performs down-sampling processing, and sends the processed signal to Cross-correlation echo effective part search module 404 . The cross-correlation echo effective part search module 404 performs a cross-correlation operation on the above-mentioned two processed signals received, that is, performs a correlation operation on the above-mentioned two signals, and then searches for the maximum value and the effective part of the echo from the cross-correlation results. , and then generate effective echo parameters according to the maximum value and effective part of the echo, and send the parameters to the adaptive filter 402 . The adaptive filter 402 receives the signals sent by the far-end input signal buffer 401 and the near-end input signal buffer 406, and the near-end output signal as the feedback signal 402 of the adaptive filter, and simultaneously receives effective echo parameters, and according to the above signals and The effective echo parameters generate an echo estimate and send it to the signal adder 407 . The signal adder 407 subtracts the estimated echo value from the input signal mixed with the echo signal received from outside the echo canceller to perform echo cancellation.

Since the sampling rate of the signal in the communication system is relatively high, for example, the sampling rate in the public circuit-switched telephone network (PSTN) is 8k and so on. If the range that the system needs to search is 64ms, then the number of samples that need to be calculated and searched for cross-correlation when performing cross-correlation is 512 points, among which there are 8 samples per 1 ms. In this way, the amount of calculation is too large, which is not conducive to improving the utilization rate of system resources. Therefore, the present invention performs some processing before performing the correlation operation. The way to deal with it is downsampling.

The principle of downsampling is to extract the original sampled data uniformly again. For example, for the data with a sampling rate of 8k in the public circuit switched telephone network (PSTN), extract one out of every four data, that is, downsample by 4. At this time, for the data for cross-correlation calculation, it is equivalent to reducing the sampling rate to 2kHz, and the amount of calculation is greatly reduced. Although the spectrum energy of the speech signal is relatively wide, the part where the energy is relatively concentrated is below 1000Hz. Because it can keep the concentrated part of the original signal frequency energy, it will not affect the search result when performing an effective echo search. If a lower sampling rate is used, it is likely to affect the frequency band where the energy of the speech signal is relatively concentrated, which will affect the search for effective echoes.

In practical applications, in order to avoid the spectrum leakage and spectrum aliasing caused by the low sampling rate after downsampling, so as to affect the calculation results, the input signal is low-pass filtered when downsampling, due to the cross-correlation In terms of sample points, the sampling rate is actually only 2kHz, so a low-pass filter with a passband not higher than 1000Hz is used for low-pass filtering. At the same time, to ensure that the concentrated frequency band of the speech signal is preserved as much as possible, a filter with a passband higher than 800 Hz but not higher than 1000 Hz is used, for example, a 1000 Hz low-pass filter used in this embodiment.

Figure 5 is a schematic diagram of the external application of the echo canceller of the present invention, as shown in Figure 5, the dotted line in the figure is the echo signal generated when the A terminal sends a signal to the B terminal, and the signal is canceled by the echo canceller.

FIG. 6 is a schematic diagram of the principle of cross-correlation calculation in the present invention. Among them, a is the initial excitation signal, b is the response signal, and c is the result. As shown in Figure 6, ts is found by finding the maximum value from the result c, and then through correction, the correction method is: in the echo maximum value According to the communication protocol, the left and right sides take the farthest part as the effective echo part tr, so that the effective echo part of the system can be found, and after finding it, the echo cancellation can be performed on the effective echo part with very small resources.

Therefore, for the echo estimation method of the present invention, FIG. 7 is a schematic diagram of the echo estimation process in the embodiment shown in FIG. 4 . It includes the following steps:

First step 701 and 702, the remote input signal and the near-end input signal are respectively down-sampled. In this embodiment, the far-end input signal and the near-end input signal are respectively subjected to steps 7011 and 7021, and the cutoff frequency is 1000 Hz. Pass filtering, and then perform steps 7012 and 7022 to down-sample the signal by 4.

Then in step 703, the two down-sampled signals are subjected to a cross-correlation operation, and in this embodiment, the cross-correlation of 128 sample points is calculated.

Next, step 704 searches out the echo maximum value and the effective part from the cross-correlation result, and obtains the effective parameters of the echo. In this embodiment, the echo maximum value is found by comparing the amplitudes of each point of the cross-correlation result. According to the communication protocol, the left and right sides take the part with the farthest distance as the effective part of the echo, and send it to the adaptive filter as the effective parameter of the echo.

In the last step 705, the adaptive filter finds the effective echo part of the received signal according to the effective parameters of the echo, and at the same time according to the effective echo part, the far-end input signal, the near-end input signal and the feedback signal as the adaptive filter The echo value is estimated from the near-end output signal.

It can be seen that the echo canceller and the echo estimation method of the present invention only perform echo cancellation on the effective part of the echo without affecting the effect of echo cancellation, which improves efficiency and saves a lot of system cost. Since the present invention adopts the down-sampling method, the amount of computation and data storage can be reduced when performing delay search. At the same time, performing low-pass filtering before down-sampling ensures that the signal will not generate spectrum aliasing due to down-sampling so as to affect the accuracy of delay estimation.

Claims (8)

1. An echo canceller, which includes an adaptive filter, a far-end input signal buffer, a near-end input signal buffer and a signal adder; the far-end input signal buffer sends the received far-end input signal as a reference signal to adaptive filter; the near-end input signal buffer sends the received near-end input signal to the adaptive filter; the adaptive filter generates an echo estimate and sends it to the signal adder; the signal adder takes the The input signal mixed with the echo signal received outside the echo canceller subtracts the estimated value of the echo to cancel the echo, and generates a near-end output signal, and sends the near-end output signal to the adaptive filter as a feedback signal; it is characterized in that, The echo canceller further includes: The remote signal down-sampling processing module, which performs down-sampling processing on the remote input signal received from the remote input signal buffer, and sends the processed signal to the cross-correlation echo effective part search module; A near-end signal down-sampling processing module, which performs down-sampling processing on the near-end input signal received from the near-end input signal buffer, and sends the processed signal to the cross-correlation echo effective part search module; Cross-correlation echo effective part search module, which performs cross-correlation calculation on the above two processed signals received, finds the echo maximum value from the calculation result, and selects the farthest distance on the left and right sides of the echo maximum value according to the communication protocol The part of the echo is used as an effective part, and it is sent to the adaptive filter as an effective echo parameter; The adaptive filter receives the signal sent by the far-end input signal buffer and the near-end input signal buffer as well as the near-end output signal, and receives the effective echo parameters, and finds the effective echo part in the received signal according to the effective echo parameters. Echo estimation is performed on the effective echo part to generate an echo estimation value. 2. The echo canceller according to claim 1, characterized in that: said remote signal down-sampling processing module further comprises a remote signal low-pass filter for buffering the received remote input signal After the input signal is low-pass filtered, it is then down-sampled. 3. The echo canceller according to claim 2, characterized in that the low-pass filter for the remote signal is a low-pass filter with a cutoff frequency of 1000 Hz. 4. The echo canceller according to claim 1, characterized in that: said near-end signal down-sampling processing module further comprises a near-end signal low-pass filter for buffering received near-end input signals After the input signal is low-pass filtered, it is then down-sampled. 5. The echo canceller as claimed in claim 4, characterized in that the low-pass filter for the near-end signal is a low-pass filter with a cutoff frequency of 1000 Hz. 6. A method for echo estimation, characterized in that the echo canceller according to claim 1 is used for echo estimation, the method comprising the following steps: 1) The far-end signal down-sampling processing module and the near-end signal down-sampling processing module respectively down-sample the far-end input signal and the near-end input signal; 2) Send the two down-sampled signals to the cross-correlation echo effective part search module to perform cross-correlation operations; 3) The cross-correlation echo effective part search module finds the echo maximum value from the cross-correlation results, and according to the communication protocol on the left and right sides of the echo maximum value, takes the part with the farthest distance as the echo effective part, and sends it as the echo effective parameter to adaptive filter; 4) The adaptive filter finds the effective echo part in the received signal according to the effective echo parameters, performs echo estimation on the effective echo part of the received far-end input signal, near-end input signal and near-end output signal, and generates an echo estimate value. 7. The echo estimation method according to claim 6, characterized in that step 1) further comprises: performing low-pass filtering on the far-end input signal and the near-end input signal before downsampling. 8. The echo estimation method according to claim 6 or 7, characterized in that the method of finding the maximum value of the echo in step 3) is: by comparing the amplitude of each point of the cross-correlation result to search for the echo maximum value.

Images