JPH0435231A - Echo canceller - Google Patents

Abstract

PURPOSE:To surely cancel an echo component by providing a double talk state consecutive time measuring section so as to discriminate it as a pseudo double talk when the consecutive time in the double talk state is a setting time or over thereby reducing a detected threshold value of a double talk detection section. CONSTITUTION:When a double talk state consecutive time measuring section 5 uses a double talk detection section 4 to detect a double talk, the consecutive time of the double talk state is measured and when the state is consecutive for a setting time or over, a detected threshold value of the double talk detection section 4 is decreased. Then when the initial detected threshold value of the double talk detection section 4 is set to 6dB in compliance with The International Telegraph and Telephone Consultative Committee (CCITT), in the case of a pseudo double talk state, the detected threshold value is decreased by a prescribed level in response to the consecutive time. Thus, only the substantial double talk state is detected, an echo canceller is normally operated to cancel the echo component.

Description

[Detailed Description of the Invention] [Summary] Regarding an echo canceller that cancels echo components that circulate through a hybrid circuit, the present invention aims to distinguish between a pseudo double talk state and an original double talk state and operate the echo canceller normally. The purpose is to add the pseudo echo signal from the pseudo echo generator to a subtraction circuit to cancel the echo component that goes around through the hybrid circuit, and to add the pseudo echo signal from the pseudo echo generator to the subtraction circuit so that the echo component that goes around through the hybrid circuit is canceled, and the cancellation residual signal of the echo component becomes zero. In an echo canceller that updates a tap coefficient by a tap coefficient updating section and stops the updating operation of the tap coefficient updating section when double talk is detected by the double talk detection section, double talk is detected by the double talk detection section. The present invention further includes a double talk state duration time measurement section that reduces the detection threshold of the double talk detection section when the double talk state continues for a set time or more.

[Industrial application field]

The present invention relates to an echo canceller that cancels echo components that circulate through a hybrid circuit.

If the hybrid circuit, such as a hybrid transformer that performs 2-wire to 4-wire conversion, has an ideal configuration, no echo component will be generated that goes around from the receiving terminal of the 4-wire line to the transmitting terminal of the 4-wire line. In this case, an echo component loops through the hybrid circuit, and in a relatively long-distance line or a high-efficiency coding line, the amount of delay becomes relatively large, resulting in deterioration of speech quality due to the echo component. Therefore, an echo canceller is provided to cancel this echo component.

This echo canceller estimates the characteristics of an echo path including a hybrid circuit, generates a pseudo echo component based on a signal from a receiving terminal of a 4-wire line, and cancels the echo component with the pseudo echo component.

[Conventional technology]

For long-distance lines, it is common to use a 4-wire line that separates transmitting and receiving lines, and a hybrid circuit is used to connect the 2-wire line to which the telephone is connected and the 4-wire line. It will be established.

For example, as shown in FIG. 6, a telephone 61.62 is connected to a hybrid circuit 63.64 via a two-wire line,
The four-line line side of the hybrid circuits 63 and 64 will be separated into transmission and reception.

A part of the signal that should be sent to the 2-line line is returned via the dotted line echo path that includes the hybrid circuits 63 and 64, and this becomes an echo component that is transmitted to the normal domestic line. In this case, the line delay time is relatively small, about 50 m5, so deterioration in speech quality due to this echo component does not pose much of a problem. However, in international lines or high-efficiency encoded lines, 300 to 50
Since the line delay time is relatively large, about 0m5 or more, you will have to listen to the echo of your own voice while talking.
Call quality will deteriorate.

In order to cancel such echo components, echo cancellers 65 and 66 are provided.
5.66 includes a pseudo echo generator 67.68 that estimates the characteristics of the echo path and outputs a pseudo echo signal, and a subtraction circuit 69.66 that subtracts the pseudo echo signal from the actual echo component.
It consists of 70. Pseudo echo generator 67.6
8 is configured with a traversal filter, for example, and estimates the impulse response of the echo path. In order to respond to changes in this impulse response, a configuration is adopted in which the filter coefficients are adaptively updated using the received signal. has been done. Known algorithms for this case include the LMS method, maximum slope method, learning identification method, etc. Among these, the learning identification method, which has a simple hardware configuration and relatively good characteristics, is often used. .

FIG. 7 is an explanatory diagram of a high-efficiency encoding line, in which a 4-wire line is connected via high-efficiency code decoding units 91 and 92, and a 64Kb
ps audio signal is highly efficiently encoded into 32Kbps or the like and transmitted.Echo cancellers 89 and 90 are provided in the multimedia multiplexing devices 77 and 78 having the high efficiency code decoders 91 and 92. 89.
90, the hybrid circuit 79 of the exchange 75.76
This cancels the echo component that passes through the 80.

These exchanges 75.76 and telephones 71.72 are public network 7
3.74, the telephone 7172 may be directly accommodated in the exchange 75.76.

There are also exchanges 75, 76 and multimedia multiplexers 77.
78, and a level adjustment section 81 consisting of an attenuator, an amplifier, etc.
-88 are provided, and level adjustment is performed at the time of system start-up, etc. In this case, a single sine wave is often input and adjusted.

[Problem to be solved by the invention] In the echo canceller 65, 66, 89, 90,
A pseudo echo signal is formed based on the received signal from the 4-wire line, and the tap coefficients of the pseudo echo generators 67 and 68 are updated so that the echo component cancellation residual signal becomes zero. If the transmitted signal from the 2-line line to the 4-line line is also included, it will be mistakenly recognized as having an increased echo component cancellation residual signal. Therefore, a double talk state in which a signal from a far-end speaker from a 4-line line to a 2-line line and a signal from a near-end speaker from a 2-line line to a 4-line line exist simultaneously is detected; (Double) In the - state, updating of the tap coefficients is stopped and the immediately previous tap coefficient is held to cancel the echo component.

In that case, the echo path loss is specified as 6 dB or more in the CCITT (International Telegraph and Telephone Consultative Committee) recommendations.
If it is at least less than 6 dB, it can be determined that it is a double talk state.

Further, the level adjustment units 81 to 88 can be adjusted by, for example, 810H.
For example, as shown in FIG. 8, if the frequency characteristics of the line are large attenuation near 810 Hz, the level adjustment section 81 is adjusted to compensate for this attenuation. In steps 88 to 88, adjustments are made to amplify.

As a result, the echo path loss is 6 dB around 810 Hz.
As mentioned above, if it is less than 6 dB at other frequencies, it is determined to be a double talk state even though it is not a double talk state, and the update operation of the tap coefficient of the pseudo echo generator 67, 68 is stopped. , it becomes impossible to cancel the echo component adaptively.

Furthermore, if the update operation of the evening/amplifier coefficients is stopped from the time a call is set up, there is a drawback that the echo component is not canceled at all.

The object of the present invention is to distinguish between the above-mentioned pseudo double-talk state and the original double-talk state, and to operate the echo canceller normally.

[Means for Solving the Problems] The echo canceller of the present invention sequentially changes the double talk detection threshold so as to detect the original double talk state, and will be described with reference to FIG.

A circuit 2 subtracts the pseudo echo signal from the pseudo echo generator 1.
In addition, the tap coefficient of the pseudo-echo generation unit 1 is updated by the tap coefficient update unit 3 so that the echo component that goes around via the hybrid circuit 8 is canceled and the cancellation residual signal of the echo component becomes zero, thereby generating double talk. In the echo canceller that stops the update operation of the tap coefficient update unit 3 when the detection unit 4 detects double talk, it is activated when the double talk detection unit 4 detects double talk, and the double talk state is longer than a set time. A double talk state duration measuring section 5 is provided to reduce the detection threshold of the double talk detection section 4 if the double talk state continues.

and a call connection detection unit 6 that detects that a call is connected and puts the double talk state duration measurement unit 5 into an operable state;
A call disconnection detection section 7 is provided which detects that the call is disconnected and returns the detection threshold of the double talk detection section 4 to its initial state.

[Effect]

When the difference between the signal power from the 4-line line to the 2-line line via the high frit circuit 8 and the signal power from the hybrid circuit 8 to the 4-line line is equal to or greater than a detection threshold, the double talk detection unit 4 detects It is determined that the state is a one-way conversation, and if it is less than the detection threshold, it is determined that the state is a double talk state.

When the double talk detection section 4 detects double talk, the double talk state duration measuring section 5 measures the duration of the double talk state, and if the double talk state continues for more than a set time, the double talk state duration measuring section 5 sets the detection threshold of the double talk detection section 4. It is intended to reduce That is, in a normal telephone conversation, both parties rarely speak at the same time, and when one party speaks, the other party listens. Therefore, if the double talk state continues for longer than the set time, it is a pseudo double talk state and not an original double talk state.

Therefore, if the initial detection threshold of the double talk detector 4 is set to, for example, 6 dB as recommended by the CCITT, in the case of a pseudo double talk state, the detection threshold is reduced by a certain amount depending on the duration. As a result, only the original double talk state can be detected, so that the echo canceller can be operated normally to cancel the echo component.

Further, the call connection detection unit 6 detects the connection of the call by a ring bank tone or the like, and from that point on, the double talk state duration measuring unit 5 is enabled to operate, and at that time, the double talk detection unit 4 detects double talk. In this case, if the double talk state continues for more than a set time, it is a pseudo double talk state, so the detection threshold of the double talk detector 4 is reduced and controlled so that only the original double talk state can be detected. .

In addition, the call disconnection detection unit 7 detects that the call has been disconnected, for example.
This is detected based on the input/output signal power difference of the subtraction circuit 2, and the detection threshold of the double talk detector 4 is returned to the initial value.

Thereby, control can be performed so that false double talk detection does not occur even when a connection is made using a line with other line conditions.

〔Example〕

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

FIG. 2 is a block diagram of an embodiment of the present invention, in which 11 is a pseudo echo generation section, 12 is a subtraction circuit, 13 is a tap coefficient update section, 14 is a double talk detection section, and 15 is a double talk state duration measurement section. , 16 is a call connection detection section, 17 is a call disconnection detection section, 18 is a hybrid circuit such as a hybrid transformer, 19 is a receiving terminal of a 4-wire line, 20 is a transmitting terminal of a 4-wire line, 21.22 is a signal power calculation section , 23 is a comparison judgment section, 24 is a detection threshold setting section, 25.26 is a ring bank tone detection section, 27 is an OR circuit, Rin is a reception human signal, Rout is a reception output signal, Sfn is a transmission input signal,
5out is a transmission output signal.

Pseudo echo generator 11. The subtraction circuit 12 and the tap coefficient updating section 13 have the same configuration and operation as in the conventional example, and the double talk detection section 14 uses, for example, signal power calculation sections 21 and 22 to detect the received input signal Rin and the transmitted signal. The signal power with the input signal Sin is calculated, and the comparison/determination section 23 compares the difference with the detection threshold α set in the detection threshold setting section 24. If the difference is greater than the detection threshold α, double talk is performed. If the difference is less than the detection threshold α, it is determined that the state is not a double talk state.

The initial value of the detection threshold α in this case is 6 dB as described above.
It can be done.

The determination signal from the comparison and determination section 23 of the double talk detection section 14 is applied to the tap coefficient update section 13 and the double talk state duration measuring section 15, and the determination signal of the double talk state causes the tap coefficient update section 13 to The updating operation is stopped, and the immediately previous tap coefficient is held and applied to the pseudo echo generator 11. Also, the double talk state duration measuring unit 15 is activated and starts measuring the duration of the double talk state, and if it continues for more than the set time, it is determined to be a pseudo double talk state and the detection threshold setting unit 24 is set. The detection threshold value α is reduced by a certain value, and the detection threshold value α is adjusted so that the duration of the double talk state becomes equal to or less than the set time. This makes it possible to set a detection threshold α suitable for the system.

In addition, when the call connection detecting section 16 detects a call connection, the double talk state duration measuring section 15 is activated so that the actual double talk state can be detected as opposed to the pseudo double talk state from the time the call is connected. Make it operational. For example, when the call connection detection section 16 detects a ring bank tone in either of the ringback tone detection sections 25 and 26, and the call connection detection section 16 is making a call to the other party,
When it is determined that the call is connected, an enable signal is applied to the double talk state duration measuring section 15 via the OR circuit 27. In this case, if the calling party goes on-hook because the called party does not respond, there is no problem because the call disconnection detection section 17 detects the disconnection state.

FIG. 3 is a block diagram of the double talk state duration measuring section, showing a case where the call connection state is determined based on the signal power of the received input signal Rin and it becomes operational, 31 is a signal power calculation section, 32.34 is a comparator, and 33 is a duration measuring device. The operation will be explained below with reference to the flowchart in FIG.

The signal power calculating section 31 of the double talk state duration measuring section 15 calculates the signal power PRin of the received input signal Rin, and the comparator 32 compares it with the threshold value TR1n.
Determine whether Ri n>TRin.

If the received signal power PRin is not larger than the threshold value TR1n, it is determined that the call is not in the connected state, and if the received signal power PRin is larger than the threshold value TR1n, it is determined that the call is in the connected state, and the double talk detection unit 14 determines that the call is not in the connected state. It is determined whether or not the double talk detection signal has been added (2), and if double talk is not detected, the measurement time Td by the duration measuring device 33 is set to 0 (2).

When the double talk detection signal is applied from the double talk detection section 14, the duration measuring device 33 increases the measurement time Td by 1 at predetermined time intervals, and the comparator 34 sets the duration time threshold T.
t, determine whether Td≧Tt, and measure the measurement time Td.
is larger than the duration threshold Tt, it is determined to be a pseudo double talk state, and the adjustment amount K is subtracted from the detection threshold α in the double talk detector 14 to set the next detection threshold α. 33 and set the measurement time Td to 0 (■).

By repeating the above operations, the detection threshold α in the double talk detection section 14 can be set to an optimal value.

Although the call connection is detected by the comparison output of the comparator 32, the duration measuring device 33 can also be made operational by the call connection detection signal from the call connection detection section 16 in FIG. It is possible, and the duration measuring device 33 can also be enabled by a logic that combines them.

FIG. 5 is a block diagram of the call disconnection detection section, in which the signal power PSin of the transmission input signal Sin and the transmission output signal 5out is
, PSout are calculated in the signal power calculation units 51 and 52, and the thresholds TSin and TSin are calculated in the comparator 5354, respectively.
Compare with TSout. When the call is disconnected, if the transmitted input signal Sin does not include an echo component, it becomes zero, and the pseudo echo signal from the pseudo echo generator 11 of the echo canceller that estimates the characteristics of the echo path disappears. It will take some time. Therefore, a pseudo echo signal is included in the transmitted output signal 5out.

Therefore, when the transmission input signal power PSin is less than or equal to the threshold value TSin, the transmission output signal power PSout is equal to or less than the threshold value TS o u
Comparators 53 and 54 determine whether or not the condition is greater than or equal to
"'1") is output and added to the double talk detection section 14, and the detection threshold setting section 24 of the double talk detection section 14
The detection threshold α is set to an initial value of, for example, 6 dB, in preparation for double talk detection in the case of the next call connection state.

In the above-mentioned embodiment, when the 4-wire line side is transmitted as a digital signal by high-efficiency encoding, etc., the functions of each part of the echo canceller can be realized by the calculation functions of a digital signal processor. In addition, when the hybrid circuit 18 is configured with a hybrid transformer, a DA that converts the received input signal into an analog signal is installed between the hybrid circuit 18 and the echo canceller.
A converter and converting the transmitted input signal into a digital signal
It will be sufficient if a D converter is provided. Further, the call connection detection section 16 and the call disconnection detection section 17 can be replaced by other detection mechanisms.

(Effects of the Invention) As described above, the present invention includes the double talk state duration measuring section 5, and determines that the double talk state is a pseudo double talk state when the duration of the double talk state is equal to or longer than a set time. This is to reduce the detection threshold of the double talk detector 4, so that the detection threshold can be adjusted to match the line characteristics even if many false double talk detections occur as a result of level adjustment in the case of a line with poor frequency characteristics. Since this can be automatically adjusted, only the original double talk can be detected, and the echo component can be reliably canceled.

In addition, a call connection detection section 6 and a call disconnection detection section 7 are provided, and the double talk state duration measuring section 5 is enabled to operate upon call connection. Since the talk detection threshold can be automatically adjusted, it is possible to operate the echo canceller normally.

Furthermore, when the call disconnection detector 7 detects a call disconnection, by returning the detection threshold of the double talk detector 4 to its initial state, double talk detection can be initialized the next time a call is connected via a different line. You can start from the state.

Therefore, it is possible to operate the echo canceller normally to cancel the echo component and prevent deterioration in speech quality.

[Brief explanation of drawings]

Fig. 1 is a diagram explaining the principle of the present invention, Fig. 2 is a block diagram of an embodiment of the invention, Fig. 3 is a block diagram of a double talk state duration measuring section, and Fig. 4 is a double talk state duration measuring section. 5 is a block diagram of the call disconnection detection section, FIG. 6 is an explanatory diagram of the echo canceller, FIG. 7 is an explanatory diagram of the high efficiency coding line, and FIG. 8 is a diagram of the frequency characteristic curve of the line. . 1 is a pseudo-echo generation section, 2 is a subtraction circuit, 3 is a tap coefficient update section, 4 is a double talk detection section, 5 is a double talk state duration measurement section, 6 is a call connection detection section, and 7 is a call disconnection detection section. Section 8 is a hybrid circuit.

Claims (2)

[Claims] (1) The pseudo-echo signal from the pseudo-echo generator (1) is added to the subtraction circuit (2) to cancel the echo component that goes around via the hybrid circuit, so that the canceled residual signal of the echo component becomes zero. The pseudo echo generating section (1
) is updated by a tap coefficient updating unit (3), and when double talk is detected by a double talk detection unit (4), an echo canceller stops the updating operation of the tap coefficient updating unit (3). , is activated when double talk is detected by the double talk detection unit (4), and when the double talk state continues for a set time or more, the double talk detection unit (4) is activated.
1.) An echo canceller comprising a double talk state duration measuring section (5) that reduces the detection threshold of (5). (2) a call connection detection unit (6) that detects that a call is connected and makes the double talk state duration measurement unit (5) operational; and a call connection detection unit (6) that detects that the call is disconnected; A call disconnection detection unit (7) that returns the detection threshold of the double talk detection unit (4) to its initial state.
2. The echo canceller according to claim 1, further comprising: