JPS6340377B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a preset echo cancellation method that uses a training signal to estimate an echo path.

Generally, telephone lines are composed of two-wire and four-wire circuits, and these lines are connected by a hybrid. In this case, it would be fine if the impedance matching was ideal, but in actual circuits, the impedance is not constant depending on the line connection status, and as the impedance matching breaks down, echo signals due to wraparound occur at this connection point. Occur. This echo signal lags considerably behind the original signal over long distance lines, thus interfering with normal communication. In addition, in loudspeaker telephones that allow calls without the handset, the acoustic coupling between the speaker and microphone of the loudspeaker telephone is greater than that of a regular telephone, so if the amount of attenuation of the echo signal generated in a hybrid is small, signal loops may occur. This can cause howling.

An echo canceller is one of the methods for attenuating the echo signal as described above. This method is
The echo path characteristics, which are the characteristics when looking from the 4-wire side to the 2-wire side in a hybrid, are estimated in the form of an impulse response, and based on this impulse response, a pseudo-reverberation signal of the echo signal generated at this node is generated. The echo signal is generated and subtracted from the actual echo signal to cancel the echo signal, and a loss corresponding to the signal level is added to the line to attenuate the echo signal.
It is well known that, compared to the suppressor method, it has the advantage of essentially not having problems with word breaks at the beginning or end of words, and with less risk of howling.

The present invention relates to a preset type of the eco-canceller system in which echo path estimation is performed using a training signal suitable for estimating echo path characteristics when a line is connected.

A conventional preset-type echo canceling device estimates the echo path by sending a training signal to the line for a fixed period of time at different times for the calling side and the called side after the line is connected. This is to prevent the training signal of one telephone from having a negative effect on the estimated operation of the other telephone when the training signals of the calling and called parties overlap. Although the training time required to estimate is not constant depending on the line condition, in this case the training time is constant, so the problem is that echo path estimation does not finish within the training time. points may occur. In view of these problems, the present invention is designed to train both the calling and receiving telephones until the estimation of echo path characteristics is completed, and to prevent these training signals from overlapping. If one of the incoming telephones is not a public address telephone,
Both systems are characterized by the fact that training can be completed and calls can be started in less training time than when using a public address telephone.

Hereinafter, the present invention will be explained in detail with reference to Examples.

FIG. 1 shows an outline of an embodiment of the present invention.
10 is a loudspeaker telephone with line connection detection function, call origination,
It has an incoming call determination function and outputs a line connection detection signal and an outgoing/incoming call determination signal, respectively. 1
1 is a training signal source, which is a circuit that generates a signal suitable for the echo path estimation operation of the 18 sidetone cancellation circuits. 12 is a training start signal generation circuit;
A training start signal is generated which is a sine wave signal with a constant frequency _f1 . 13 is a training end signal generation circuit which generates a training end signal which is a sine wave signal with a constant frequency _f2 . 14 is a selection circuit which selects one signal from the transmission signal from the loudspeaker telephone 10, the training signal from 11, the training start signal from 12, the training end signal from 13, and no signal, and selects one signal from the 19 hybrid This is a circuit that sends data to the line via the circuit. 18 is the transmission signal from 10 loudspeaker telephones,
This is a circuit that cancels the sidetone generated by the unbalanced hybrid circuit of No. 19 that reaches the receiving side. An example of the configuration of this circuit is shown in Fig. 2.

In FIG. 2, numeral 21 is an A/D converter for converting a transmission signal into a digital signal, and the input signal digitized by this circuit is held in the X register as a value past the current time by the number of taps. 23 is an H register that holds the echo path characteristics in the form of an impulse response; when the echo canceller finishes estimating the echo path characteristics, the convolution integral result of 24 becomes equal to the sidetone. The estimation operation is to calculate the difference between the result of the convolution during estimation and the sidetone using 26 subtraction circuits, and then multiply this value by the value of each tap in the X register and a constant value α using 27 multiplication circuits. Then, the process proceeds by repeatedly performing a series of operations in which the value is added to the value of the H register of the corresponding tap in the H register update circuit 28.

In addition, the input side sound signal of this side sound cancellation circuit is as follows:
Since the estimation operation is disturbed when a signal other than the sidetone of the transmitted signal is added, an update prohibition signal is input to the H register update circuit to prohibit the estimation operation except during training. 16 is a training end signal detection circuit, an example of its configuration is shown in FIG.

In the figure, 31 is a border filter that blocks only the frequency f ₂ of the training end signal;
35 is a threshold circuit that outputs a logic signal "1" only when the output is below a certain value, 32 is a bandpass signal that passes only f ₂ , and 34 is a circuit that takes the absolute value of its output. Circuit that takes the absolute value of the output, 3
6 is a threshold circuit that outputs a logic signal "1" which is a training end signal detection signal only when the output level is above a certain value, and 37 is a threshold circuit that outputs a logic signal "1" which is a training end signal detection signal, and 37 is a threshold circuit that outputs a logic signal "1" when the outputs of 35 and 36 become "1" at the same time. This is an AND gate circuit that outputs a logic signal "1".
With the configuration shown in FIG. 3, it is possible to detect that only the training end signal is included in the received signal.

Reference numeral 17 denotes a line estimation completion detection circuit, an example of its configuration is shown in FIG. In the figure, 41 is an absolute value circuit that takes the absolute value of the input signal, 42 is an integration circuit that integrates the output of 41 for a certain period of time, and 43 is a logic circuit that outputs a line estimation end signal only when the output level of 42 is below a certain value. Outputs signal "1". Then, the 15 control circuits apply this output to the 44 AND gates using the signals during the training period, and the output becomes "1", indicating that the sidetone has fallen below the predetermined level ( This is a circuit for detecting the completion of line estimation.

20 is a training start signal detection circuit, an example of its configuration is shown in FIG. In the figure, 51 is a band pass filter that passes only the training start signal of frequency _f1 . 52 is a circuit that takes the absolute value of the output, and 53 is a threshold circuit that outputs a logic signal "1" which is a training start signal detection signal only when the output is above a certain value.

15 inputs a line connection detection signal, an outgoing/incoming call determination signal, a training end signal detection signal, a line estimation end detection signal, and a training start signal detection signal, and based on these signals, the selection circuits 14 and 18 This circuit sends a control signal to the sidetone cancellation circuit of the system and controls training after the line is connected.

The training sequence in the present invention is shown in FIG. 6 and will be described below.

First, when the telephone set in FIG.
A control signal is sent to the selection circuit, and a training start signal is sent to the line for a certain period of time (about several tens of milliseconds). At this time, a signal is sent to the sidetone canceling circuit to inhibit the estimation operation.

Next, as shown in b, the control circuit sends a control signal to the 14 selection circuits to transmit a training signal to the line, and simultaneously starts the estimation operation of the sidetone canceling circuit. Thereafter, the control circuit continues to send training signals to the line until a line estimation completion detection signal indicating that the sidetone has been attenuated to a predetermined amount is received. When the line estimation end detection signal is input, the control circuit ends the estimation operation of the sidetone canceling circuit and causes the 14 selection circuits to selectively send the training end signal to the line for a certain period of time (about several tens of milliseconds) as shown in c. After that, a training start signal detection signal is generated for a certain period of time (approximately 100 msec) indicating that a training start signal has been received from the partner station. If the training start detection signal is not received during this period, the control circuit determines that the other party is not a loudspeaker telephone, as shown in d, and immediately causes the selection circuit 14 to send out the transmission signal of the loudspeaker telephone, thereby starting the conversation. enable. On the other hand, when the training start signal is detected as shown in e, the control circuit receives the training end signal from the partner station by 1.
The selection circuit 14 selects no signal and does not send anything on the line until the training end signal detection circuit 6 detects the training end signal.After that, when the training end signal is detected from the partner station, the selection circuit selects the signal from the loudspeaker telephone. To enable a call by selecting a transmission signal.

Next, when the telephone in FIG. 1 is on the called side, the control circuit 15 connects the line and learns that it is the incoming call by the line connection signal from the loudspeaker telephone 10 and the calling/incoming call determination signal. , fixed time (about 100msec)
Set the selection circuit to no signal and receive a training start signal from the partner station. If, during this period, no training start detection signal is received from the training start signal detection circuit 20, the control circuit determines that the other party is not a public address telephone, and immediately sends the training signal to the selection circuit 14 as shown in j. is selected, and the training signal is sent out on the line until the completion of line estimation is detected in step 17, and at the same time, the sidetone canceling circuit is caused to perform the estimation operation. Thereafter, when the control circuit receives a line estimation end signal as shown in h, the selection circuit selects the transmission signal from the loudspeaker telephone to enable a telephone conversation.

On the other hand, when the control circuit inputs the training start signal detection signal from the partner station, the control circuit waits for the training end signal detection signal which is the output of 16. After that, when the training end signal is detected, the control circuit immediately executes the training start signal detection signal as shown in α. A training start signal is sent from the selection circuit to the line for a certain period of time (about several hundred milliseconds), and then, as shown in g, the selection circuit selects a training signal and sends the training signal onto the line. At the same time, the estimation operation of the sidetone canceling circuit is also started. This operation continues until the line estimation completion detection signal is input. Next, when the line estimation end detection signal is input, the control circuit ends the estimation operation of the sidetone canceling circuit, and causes the selection circuit to send out a training end signal for a certain period of time (about 100 msec) as shown in h. Thereafter, as shown in i, the selection circuit selects the transmission signal from the loudspeaker telephone to enable a telephone conversation.

In FIG. 6, origination indicates the timing of the calling side, and terminating indicates the timing of the called side.

FIG. 7 is a circuit example of the above-mentioned control circuit.

In the same figure, 701, 702, 704, 70
8, 710, 711, and 714, the output Q becomes high level "1" for a certain period of time at the rising edge or falling edge of the input CK pulse, and then becomes 0. Monostable multivibrator, 703,7
05,706,707,709,712,71
3,700, Q becomes "1" at the falling edge of the S input signal, and Q becomes "0" at the falling edge of the R input signal.
This is an RS flip-flop. is the output of the inverse logic of the Q of each flip-flop. It is also assumed that the flip-flops 700 to 714 are initially cleared by the CL pulse and set to Q="0" and "1".

When the telephone in FIG. 1 is on the calling side, when a pulse A indicating that the line is connected is input, the output Q of 701 becomes "1" for a certain period of time, and the selection circuit 14 sets the signal F to "1" for a certain period of time. select the training start signal. Next, when signal F falls, a flip-flop 704 is set to generate a signal G that causes 14 to select a training signal and 18 to perform an echo path estimation operation. After that, when the echo path estimation is completed and the output B of the line estimation end detection circuit becomes "1", the rising signal causes the Q of 702 to become "1" for a certain period of time, and the training end signal is sent to the selection circuit 14 for a certain period of time. A signal E is generated to select the .
Next, when E falls, the output Q of 704 becomes 1 for a certain period of time.
During this time, when the training start signal C from the called side is input, the output Q of 706 becomes "1" due to the output of the AND gate of 718. When the training end signal D from the called side is input next, the output Q of 706 becomes "1". Due to the output of the AND gate 717, the output of the output Q716 of 705 becomes 1, and a signal H is generated that causes 14 to select the transmission signal.

On the other hand, if the detection output C of the training start signal from the other station does not become "1" while the output Q of 704 remains "1" for a certain period of time, the Q of 704 becomes "1".
At the same time as the falling edge of , the output of 707 becomes "1", and a signal H that causes 14 to select a transmission signal is generated.

On the other hand, if the telephone set in Figure 1 is the called party, when the signal J indicating that the line is connected is received, it will take 70 seconds for a certain period of time.
If the output Q of 8 becomes "1" and the training start signal from the other station does not enter within that section,
The output I of the training start signal detection circuit does not enter, and at the same time the output of 708 falls, 721, 71
1 sets L to "1" for a certain period of time, causes 14 to send a training signal, and then 700, 72
3 sets K to ``1'' and causes 14 to select a transmission signal to enable communication. Also, when the output of 708 is "1", the training start signal from the other party is received, and when the training start signal detection signal is input, the output Q of 709 becomes "1", and then the training from the called side is received. When the end signal detection signal T is input, the output of 710 and M become "1" for a certain period of time, causing 14 to send out a training start signal. Next, when the sending of the training start signal is finished, 712 and 714 set L to ``1'' until N, which is the output of the line estimation end detection circuit, becomes ``1'', causing 14 to send out the training signal, As soon as N becomes ``1'', set 1 for a certain period of time to make 14 send a training end signal, then set output Q of 713 to ``1'', set K to ``1'', and send a transmission signal to 14. send it out. 71
Reference numeral 5 denotes a signal P for determining whether the telephone is on the calling side or the called side, and is a multiplexer that selects each control signal of FGEH and each control signal of LMK and uses it as a control signal for the 14 selection circuits. It is assumed that the selection circuit No. 14 does not output anything when all of the outputs ABCD of this multiplexer are "0".

The above circuit can realize the control procedure described above.

It should be noted that there is essentially no problem with the circuit and control procedure described above even if the positions of the calling side and the called side are reversed.

As explained in detail above, the echo cancellation method of the present invention provides
Since it sends out an identification signal that can be easily detected on the receiving side, it is possible to wait for the end of the training signal on the other party's phone before sending out the training signal, thereby preventing overlapping training signals.
If one of the telephones is not a loudspeaker telephone, since there is no identification signal at the start of training for the other telephone, it can be easily determined whether it is a loudspeaker telephone, and the training time can be shortened.

[Brief explanation of drawings]

FIG. 1 is a diagram showing an embodiment of the present invention, FIG. 2 is a diagram showing an example of the configuration of a sidetone cancellation circuit, FIG. 3 is a diagram showing an example of the configuration of a training signal detection circuit, and FIG. FIG. 5 is a diagram showing an example of the configuration of a line estimation detection circuit, FIG. 5 is a diagram showing an example of the configuration of a training start signal detection circuit, FIG. 6 is a timing diagram of training, and FIG. 7 is an example of the configuration of a control circuit. It is a diagram. 10... Loudspeaker telephone, 11... Training signal source, 12... Training start signal source, 13...
...Training end signal source, 14...Selection circuit,
15... Control circuit, 16... Training end signal detection circuit, 17... Line estimation end detection circuit, 1
8... Sidetone cancellation circuit, 19... Hybrid circuit, 20... Training start signal detection circuit.

Claims (1)

[Claims] 1. A training signal, a transmission output signal from a local telephone, a training start signal indicating the start of the training signal, and a training end signal indicating the end of training are selectively transmitted through 2-4 lines. A selection circuit sends out to the hybrid circuit for conversion, a training signal sent out from this selection circuit, and the sidetone signal to estimate the sidetone characteristics and create a pseudo sidetone signal, and this signal is sent to the own station. a sidetone canceling circuit that eliminates a sidetone signal by subtracting it from a received input signal to the telephone; and a first sidetone canceling circuit that detects a training start signal and a training end signal from the other station from the received signal.
a second detection circuit for detecting completion of sidetone characteristic estimation by the sidetone cancellation circuit; and a second detection circuit for controlling transmission of the training signal and estimation of the sidetone characteristic by the sidetone cancellation circuit. and a control unit further configured to start the training when the line is connected to the other station or when the first detection circuit does not detect a training start signal from the other station within a predetermined time. a first means for sending a signal to the hybrid circuit via the selection circuit; and following this means, sending the training signal and detecting that the estimation of sidetone characteristics is completed by the second detection circuit; a second means for continuing to send the training signal to the hybrid circuit until the training signal is detected; a third means for subsequently sending the training end signal to the hybrid circuit; and fourth means for waiting for training of the own station until the first detection circuit detects the training end signal from the partner station when the training start signal from the station is detected within the predetermined time. A reverberation canceling device featuring: