JPS6259937B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an improvement in an echo canceller, and more particularly to an improvement in an echo canceller inserted into a 2-wire 4-wire conversion section in a long-distance telephone line.

Echo cancellers are used in long-distance telephone lines using communication satellites and submarine cables.
This device improves the quality of speech by eliminating reflections caused by impedance mismatching of the hybrid coil provided in the line conversion section, and the device mainly includes a correction amount calculator and a superimposed speech detector.

The superimposed call detector conventionally used in this type of device uses an analog circuit to detect the sending audio signal and the receiving audio signal in order to determine the status of the call by level comparison. This method involves separately detecting the voice signal and the received voice signal, comparing the smoothing levels, and providing the necessary hangover time. Determining this call state was extremely important in the process of simulating echo path characteristics. On the other hand, for an echo suppressor that is partially similar to this echo suppressor, the value of the time constant and hangover time of the smoothing circuit used in the superimposed call detector is This was an extremely important issue because it concerned the cutting of the ending of the story. Therefore, the echo canceller used the same superimposed speech detector as the echo compressor.

Furthermore, the superimposed speech detector of an echo suppressor or echo canceller connected to a normal PCM telephone line simulates the detection and smoothing means of a conventional analog circuit for a superimposed speech detector. For example, for outgoing audio signals, the non-linear code used in normal PCM telephone lines, which corresponds to the outgoing audio signals, is first converted into line codes, and then the input data is converted into line codes.
Means for sequentially adding hundreds of samples, means for subtracting old data each time new input data is added, means for subtracting old data each time new input data is added, and means for subtracting the discharge time constant each time new input data is added in order to approximate the effect of the discharge time constant of a smoothing circuit. This was done by multiplying by a close constant, and by combining several of these methods. The same process was performed for the received voice signal, and by comparing the data obtained from both, the levels were compared and the call status was determined.

The above-mentioned means were taken to imitate analog circuits as much as possible since echo suppressors using analog circuits have been widely used and their performance has been sufficiently evaluated. However, simulating the conventional superimposed call detector using an analog circuit with a digital circuit as described above has the disadvantage of being complicated and large in scale.

SUMMARY OF THE INVENTION It is therefore an object of the present invention to provide an echo canceller that is small and economical.

In order to achieve the above object, the echo canceller of the present invention takes advantage of the fact that the superimposed speech detector in the device does not need to be particularly fast, and operates at relatively long time intervals. This is for simplification.

According to the present invention, there is provided an echo path model that receives a correction signal and synthesizes a controlled echo from a received voice signal, an addition/subtraction calculator that outputs a difference between an echo included in a transmitted voice signal and the synthesized synthesized echo; A superimposed call detection circuit compares the levels of the transmitting audio signal and the receiving audio signal to determine the state of the call and issues a determination output; In the echo canceller, the echo canceler includes a correction amount calculator that generates the correction signal to control the synthesized echo outputted by the echo path model so that the difference between the outputs becomes small according to the state, and the superimposed speech detection circuit Each of the input signal systems for the speaking audio signal and the receiving audio signal includes an input signal level holding means for selectively updating the level of the input signal using a control signal, and an input signal level holding means for updating the input signal level of the input signal level and the signal to be held by the input signal level holding means. The output levels are compared, and if the comparison result shows that the former is greater than the latter, the control signal is issued and counting is restarted from zero, and if the comparison result shows that the former is not greater than the latter, the number of input signals is and a comparison counter means which issues the control signal when a predetermined number of times τ has elapsed and restarts counting from zero from that point, and between the two input signal systems, the two input signal levels are There is obtained an echo canceller characterized in that it has comparison means that compares the output levels of the signals held by the holding means and generates the discrimination output.

Next, the present invention will be explained in detail with reference to the drawings.

The figure shows the configuration of one embodiment of the present invention in blocks. In the figure, 1 is a transmitting audio input terminal, 2 is a transmitting audio output terminal, 3 and 4 are receiving audio input terminals and output terminals, 5 is a superimposed speech detector, 6 is a correction amount calculator, 7 is an echo path model, 8 is a subtractor. The superimposed speech detector 5 is a characteristic circuit in the echo canceler according to the present invention, and its detailed configuration is shown within the frame of the broken line. The operation of the echo canceller as a whole will be explained using a detector as one unit, and in the meantime, what properties this superimposed speech detector 5 should have will be explained.

The echo path model 7 is a circuit that synthesizes the echo of the echo path using a received voice signal, and has a built-in transversal filter. The output of this echo path model 7 is guided to a subtractor 8, and the subtractor 8 subtracts the synthesized echo synthesized by the echo path model 7 from the echo guided to the transmission audio input terminal 1, and outputs the difference. The output of the subtractor 8 and the received voice signal are led to the correction amount calculator 6. A correction amount calculator 6 uses these two signals to calculate a correction amount in order to correct the tap gain of the transversal filter of the echo path model 7 so that the difference appearing in the output of the subtractor 8 becomes small. This calculation result is given to the echo path model 7, and the echo path model 7 simulates the characteristics of the echo path. When treated as a unit circuit as described above, the superimposed speech detector 5 determines the speech state by comparing the levels of the transmitting and receiving audio signals, and stops the calculation of the correction quantity calculator 6 in the transmitting state. , to prevent disturbances in the tap gain of the transversal filter of the echo path model 7.

On the other hand, an echo suppressor has a different operating principle from an echo canceller, and as is well known, a resistor attenuator is placed between the input terminal and the output terminal of the transmitting audio signal. When an attenuator is inserted and a transmitting state is determined, the resistive attenuator is controlled to be removed to reduce the echo by the attenuation amount of the resistive attenuator.
Therefore, in the case of simultaneous calls with an echo suppressor,
Depending on whether it is a transmitting state or a receiving state, insertion or removal of the resistive attenuator may cause the beginning and end of the voice to be cut off.

From the above explanation, since the operating principles of an echo suppressor and an echo canceller are completely different, the superimposed speech detector used in the echo canceller can reduce the disturbance of the transversal filter of the echo path model 7 by quickly determining the speech state. The main focus is on doing. Note that for the double talk detector used in the echo suppressor, care is taken in setting the hangover time so that voice disconnections during calls do not occur too often. In this way, the echo canceller does not need to pay much attention to setting the hangover time associated with the syllables of the audio signal. Therefore, the configuration of the superimposed speech detector used in the echo canceller does not need to be simulated by the analog circuit described above.

Next, the superimposed call detector 5, which is the main part of the present invention, will be explained in detail. In the figure, 11 is a data input terminal for transmitting voice signals to this detector;
Reference numeral 2 indicates a data input terminal for a received voice signal to the detector; 13, 14 and 15 are comparators that compare the magnitude of the two input data and output the comparison results; 16 and 17 are counters that count hangover time; 18 and 19 are registers, and 20 is an output terminal of this detector. 1 in the diagram
The connections of the circuits 1, 13, 16 and 18 are the same as those of 12, 14, 17 and 19. The data input terminal 11 of the transmitted voice signal is connected to the input terminal of the comparator 13 and the input terminal of the register 18, and the output terminal of the register 18 is connected to the other input terminal of the comparator 13 and the input terminal of the comparator 15. Also, comparator 1
The output terminal of the counter 3 is connected to the input terminal of the counter 16, and the output of the counter 16 is connected to the control input terminal of the register 18.

The register 18, which receives the transmitted voice signal data from the data input terminal 11, updates the stored contents to the input data when it receives a control signal from the control counter 16. The comparator 13 compares the input data and output data of this register 18 and sends the comparison output to the control counter 16. When the comparison output indicates that the input data is larger than the output data, the control counter 16 issues the control signal and temporarily sets the count to zero and restarts counting; when the input data is not larger than the output data, the control counter 16 outputs the control signal and resets the count to zero. continues to count the number of input signals without issuing a control signal, but
When the number of inputs exceeds a predetermined number τ, the control signal is generated even if the input data is still smaller than the output data, and counting is restarted from zero as before. Therefore, new input data is always stored in the register 18 when a time period that does not exceed τ times (inconsistent) has elapsed (inconsistent).
For example, if the input is 8 kHz and τ is 256 times, the maximum update time will be 32 ms.

The operation of the circuits 11, 13, 16, and 18 has been explained above, but the circuits 12, 14, and 1
The operation of the portion consisting of 7 and 19 is performed in the same manner. Therefore, new input data of the received voice signal is updated and stored in the register 19 at every time as described above. The comparator 15 compares the data stored in the two registers 18 and 19, and sends the comparison result of the transmitted audio signal and the received audio signal to the correction amount calculator 6 from the terminal 20.
Control the control signal. The subsequent operations are as described above, and disturbances in the echo path model can be effectively prevented with a simple configuration.

The present invention is particularly effective when the echo canceller is interfaced to a PCM telephone. In other words, the code format of the PCM telephone line is a nonlinear code, that is, A-law or μ-
lam, etc., and consists of a mantissa part and an exponent part, so the exponent part and part of the mantissa part can be used as input data for the superimposed call detector to have a data length of several bits. Therefore, it can be configured with a small number of bits without performing code conversion from non-linear code to linear code, and comparators 13, 14, 15 and registers 18, 19
This has the advantage that it can be constructed more compactly compared to conventional methods.

In the embodiment described above, the superimposed call detector 5 is formed of a well-known but relatively basic circuit, so it can be constructed using a general-purpose microprocessor.

As explained above, by guiding the transmitting and receiving voice signals to the superimposed speech detector, and guiding the output of the superimposed speech detector to the correction amount calculator to control the correction amount calculator, the echo path model is disturbed. can be effectively prevented. Especially in the case of digital interface echo cancellers, the usual PCM
This non-linear code can be used directly without the need for code conversion means from a non-linear code used in a telephone line to a linear code. Therefore, an echo canceller having a superimposed speech detector as described above is compact and economical.

[Brief explanation of the drawing]

The figure is a diagram showing the configuration of an embodiment of the present invention. Explanation of symbols: 1 is a transmitting audio input terminal, 2 is a transmitting audio output terminal, 3 is a receiving audio input terminal, 4 is a receiving audio output terminal, 5 is a superimposed speech detector, 6 is a correction amount calculator, 7 is a Echo path model, 8 is a subtractor, 13, 14, 15 are comparators, 16, 1
7 represents a counter, and 18 and 19 represent registers, respectively.

Claims (1)

[Scope of Claims] 1. An echo path model that receives a correction signal and synthesizes a controlled echo from a received voice signal; an addition/subtraction calculator that outputs a difference between the echo included in the transmitted voice signal and the synthesized synthesized echo; a superimposed call detection circuit that compares the levels of the outgoing audio signal and the incoming audio signal to determine the state of the call and outputs a determination output; In the echo canceller, the echo canceller includes a correction amount calculator that generates the correction signal to control the synthesized echo output by the echo path model so that the difference between the outputs becomes small according to the state of the echo path model, and the superimposed speech detection circuit For each input signal system of the transmitting audio signal and the receiving audio signal,
An input signal level holding means that selectively updates the level of the input signal by a control signal, and an input signal level holding means that compares the level of the input signal of the input signal level holding means and the output level of the signal to be held, and the comparison result is that the former is greater than the latter. If the comparison result indicates that the former is not greater than the latter, the number of input signals is counted, but after a predetermined number of times τ has elapsed, the control signal is emitted and the counting is restarted from zero. and a comparison counter means for restarting counting from zero from that point in time, and comparing the output levels of the signals held by the two input signal level holding means between the two input signal systems. An echo canceller characterized by having a comparison means for emitting a discrimination output. 2. The echo canceller according to item 1 above, characterized in that a nonlinear input signal is adopted for the input signal of the superimposed speech detector.