JP2570011B2 - Bidirectional amplifier - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a bidirectional amplifier for amplifying signals attenuated in a two-wire transmission line for telephone, facsimile, data, and the like.

[0002]

2. Description of the Related Art When performing bidirectional amplification for the purpose of amplifying an attenuated signal in bidirectional communication using a two-wire line for transmitting telephone, facsimile, data, etc., a two-wire to four-wire conversion is performed by a hybrid circuit. After that, a method of performing amplification in each direction by a unidirectional amplifier is adopted. However, general subscriber lines differ in line type and line length for each line, so that impedance is not constant, and perfect impedance matching cannot be achieved with a hybrid circuit having a fixed constant. If impedance matching cannot be achieved, that is, if the hybrid balance is lost, the signal at the 4-wire input terminal of the hybrid circuit goes around to the 4-wire output terminal. If amplification is performed without removing the sneak signal, in an extreme case, oscillation will occur in the four-wire loop formed by the hybrid circuit, and even if oscillation does not occur, a large echo will occur. For this reason, in the case where the two-wire / four-wire conversion is performed only by the hybrid circuit having a fixed constant, if the line conditions fluctuate, almost no amplification can be performed.

[0003] As a technique for solving this problem, Echo
There is a bidirectional amplifier using a suppressor system or an echo canceller system.

A two-way amplifier using the echo suppressor system utilizes the property that ordinary voice communication is conversational, that is, non-simultaneous two-way communication. For example, when the signal level of speaker A is higher than the signal level of speaker B, it is considered that there is no signal of speaker B, and only the signal of speaker A is amplified and the signal of speaker B is suppressed and attenuated. Let it. By controlling the call direction in this way, it is possible to amplify the signal while suppressing the gain of the 4-wire loop. The echo suppressor method has an advantage that the control is simple and the configuration can be made at low cost. On the other hand, since the control cannot be performed until the signal level of the speaker A or the speaker B is detected to determine which signal is to be suppressed, the beginning of the speech is necessarily present, and essentially the speech is cut off. Since simultaneous calls are not allowed, there is a disadvantage that natural calls tend to be hindered.

A two-way amplifier using the echo canceller system has a pseudo wraparound signal generation unit inside and has a two-wire
A signal in which a signal on the transmission side in a hybrid circuit that performs line conversion wraps around a line on the reception side is canceled by a pseudo wraparound signal output from a pseudo wraparound signal generation unit, thereby attenuating a signal circulating in a four-wire loop. To enable amplification. The pseudo wraparound signal generator consists of a variable tap coefficient transversal filter that can vary the amount of attenuation and delay time, and observes the wraparound signal generated by the hybrid circuit and modifies the tap coefficients to minimize it. I do. Therefore, even if the conditions of the telephone line change, it becomes possible to reduce the sneak signal adaptively. In the echo canceller system, since it is not necessary to control the communication direction, there is no talk head disconnection occurring in the echo suppressor system, and simultaneous two-way communication can be performed without any trouble.

A bidirectional amplifier using the echo canceller system is roughly classified into a preset type and a constantly adaptive type in terms of its control system. In the echo canceller type preset type, it is necessary to determine the tap coefficient of the transversal filter in order to set the amount of attenuation and delay time of the pseudo wraparound signal. It is transmitted from the 4-wire input terminal (transmitting side), monitors the amount of wraparound of the training signal appearing at the 4-wire input terminal (receiving side), and determines the optimum value of the tap coefficient of the transversal filter (optimization). ). This optimizing operation is completed within a fixed time period in which no communication signal exists, and thereafter, communication is performed while holding the value of the optimized tap coefficient.

In the always- adaptive type, the signal of the 4-wire input terminal of the hybrid circuit which appears during communication without using a special training signal is used, and the amount of the signal which wraps around to the 4-wire output terminal of the hybrid circuit is monitored. This optimizes the tap coefficients of the transversal filter.

An advantage of the preset type is that training is performed during a period in which no communication signal exists, so that a training signal can be arbitrarily selected. In the operation of optimizing the tap coefficients, how quickly it can converge to the optimum value (determining the optimum value) is a problem, which involves the input of the pseudo wraparound signal generation unit (the signal at the 4-wire input terminal of the hybrid circuit). ) Requires a signal with little autocorrelation. If a signal having a small autocorrelation is selected as a training signal, it is possible to quickly converge to an optimum value. Disadvantages of the preset type are that the optimizing operation must be completed within a certain period, and that if the line conditions change during communication, it cannot be dealt with. However, since the condition of the telephone line once connected rarely changes during communication, it cannot be a fatal disadvantage.

The advantages of the always-adaptive type are that the above-mentioned optimization can be performed at any time during communication, and that it can cope with a change in line conditions during communication.
However, since the optimization operation is performed using signals such as voice and modem signals, there is a disadvantage that a signal having relatively strong autocorrelation must be assumed. Assuming such a signal, the convergence to the optimum value becomes rather gentle, and the tap coefficient may not be optimized in the early stage after the start of communication. Therefore, it is necessary to control the amplification factor of the amplifier to be set small at the initial stage after the start of communication and then to increase gradually thereafter, which may impair natural communication or normal communication.

In the above comparison between the advantages and disadvantages of the preset type and the always-adaptive type of the echo canceller system, if it is assumed that the condition of the telephone line once connected rarely changes during communication, The preset type is excellent. In the case of the preset type using the echo canceller method, if the initial training fails, the tap coefficients are not optimized thereafter, so that the communication cannot be continued and the connection of the communication line has to be abandoned. Therefore, it is very important at what point in the communication to train. There are three conditions required for successful training. The first is that the connection of the line to be trained is completed (the line is terminated at the local exchange side), the second is that the communication is in a silent state, and the third is that the silent state is in effect. A certain period can be predicted.

The first condition is, of course, that training is possible only after the communication line is connected, so that it is natural.

The second condition is a condition related to the success / failure of training, which is important because training is possible only in a silent state. The pseudo wraparound signal generation unit detects that the signal of the 4-wire input terminal of the hybrid circuit is 4
It monitors how much leaks to the line side output terminal and generates a pseudo wraparound signal. If a signal other than the wraparound signal, for example, an audio signal, is superimposed on the 4-wire output terminal from the 2-wire terminal, the monitored content will be different from the true wraparound signal. As a result, the tap coefficients do not converge to the optimal value, and the training fails. Since it is difficult to know whether or not an unexpected signal is superimposed on the two-wire side terminal during transmission of the training signal, the training must be performed in a silent state.

The third condition is a condition necessary for controlling the training, and the continuation of the silent state must be guaranteed for a required period after the training is started. There are many silent periods that can satisfy the training period in communication, but most of these depend on the communication content, and it is difficult to schedule the training period in advance.

[0014]

In the echo canceller type preset type, the silent period in which the above three conditions are highly likely to be satisfied is a period from immediately after the line connection is completed to when a communication signal is actually output. Conceivable. This period is the time until a person picks up the receiver and starts a conversation or the time when the communication machine responds with a timer, and is present in most communication. That is,
If the training period is set immediately after the line connection is completed, the training will succeed in most communications. However, although there is a small probability, a communication signal may be output immediately after the line connection is completed, and even if it is not output immediately, a communication signal may be mixed before the end of training. There was an unsolved problem of failure.

[0015]

SUMMARY OF THE INVENTION In a preset type of the echo canceller system, a communication control circuit for detecting the absence of a communication signal and a hybrid circuit for performing two-wire / four-wire conversion for executing an optimizing operation. A white noise generator applied from the 4-wire input terminal, a changeover switch for switching between white noise and a communication signal at the 4-wire input end, and one of the white noise and the communication signal via the changeover switch An adaptive filter for applying a signal obtained from the 4-wire input terminal, attenuated based on the adaptation permission signal to the 4-wire output terminal of the hybrid circuit as a pseudo-wraparound signal, and a communication signal from the communication control circuit When a non-existence signal indicating that no signal is present is obtained, the changeover switch is switched to the white noise generator side and the adaptive filter is optimized. When it is completed, it is instructed to hold the attenuation and delay, which are the optimum values obtained by the optimization operation, and when the application of the non-existence signal stops during the optimization operation, the changeover switch is turned on. A success / failure determiner is provided for switching to the communication signal side, stopping the current optimization operation, and instructing the adaptive filter to keep the optimum value obtained by the previous optimization operation as it is. In determining success or failure, the received signal and the pseudo
Past data using the residual signal that is the difference from the echo signal
Residual signal power or peak value with smaller weight
And the residual signal power is below a predetermined level.
Is successful when it is, and fails when it is above a certain level
And

[0016]

The training is performed by the white noise which is the training signal immediately after the connection of the communication line is completed, and after a lapse of a predetermined time, the level of the training signal (white noise) at the output terminal on the 4-wire side of the hybrid circuit sufficiently exceeds a predetermined level. Judgment is made as to whether it is below, and if it is below, training is terminated at that point, and if it is not below, training operation is continuously performed, so that training time can be secured immediately after connection of the communication line is completed. In the case of communication, the training is completed before a communication signal is output. Even if the training time cannot be secured immediately after the connection of the communication line is completed and the communication signal is superimposed, the optimization operation is performed when the communication signal becomes silent. The possibility of operation failure has been significantly reduced, and communication reliability has been improved.

The bidirectional amplifier according to the present invention does not require any special control from the outside, and can be replaced with a conventional bidirectional amplifier incorporated in a conventional telephone device.

[0018]

An embodiment of the present invention will be described with reference to FIG.

FIG. 1 shows a local line L0 and a local line L1 (hereinafter referred to as a local line L0 and a local line L1).
Communication lines are called office lines) connected by a bidirectional amplifier. In the figure, the circuits on the left and right sides of the central dashed line have exactly the same circuit configuration.

In FIG. 1, an office line L0 is connected to a hybrid circuit 11a and a call control circuit 30a via an office line interface 20a. An A / D converter 15a converts an analog signal from the hybrid circuit 11a into a reception signal y ₀ (n) which is a digital signal. 1
6a is a D / A converter, which converts the transmission signal u ₀ (n) into an analog signal and applies it to the hybrid circuit 11a,
Two-line / four-line conversion is performed, and the result is transmitted to the office line L0 via the office line interface 20a. If the impedance matching with the local line L0 is not completely achieved, the reception signal y ₀ (n)
Contains a wraparound component d ₀ (n) of the transmission signal u ₀ (n) from the hybrid circuit 11a. Adaptive filter 12a to generate a pseudo echo signal d _0p (n) is a pseudo echo component inputs the transmission signal u ₀ (n), wraparound component d ₀ contained in the received signal y ₀ (n) (n) And a delay filter operation for outputting the estimated value as a pseudo echo signal d _0p (n), and an adaptive permission signal INH ₀ , a transmission signal u ₀ (n), and a residual signal e ₀ (n) as inputs. An adaptive operation (hereinafter, referred to as training) for optimizing a filter coefficient is performed so that the pseudo echo signal d _0p (n) approaches the wraparound component d ₀ (n). Subtracter 18a subtracts the pseudo echo signal d _0p (n) from the received signal y ₀ (n), and outputs the residual signal e ₀ (n). Therefore, the residual signal e
₀ (n), the reception signal y ₀ from the local line L0.
The component of (n) is preserved, and the wraparound component d ₀ (n) is suppressed. The multiplier 19a is an amplifier, and the residual signal e ₀
(N) is multiplied by a gain constant α ₀ to obtain an amplified reception signal x
Output as ₁ (n). White noise generator 14
a is an uncorrelated signal with variance σ _N ² (eg, Gaussian noise)
As white noise N ₀ (n). white·
The noise generator 14a is realized, for example, by previously storing an uncorrelated signal with a variance σ _N ^{2 in} a ROM (Read Only Memory) and sequentially reading the signal. Changeover switch 17a is adapted permission signal switched by INH _0, adaptive permission signal INH ₀ is transmission signal u white noise N ₀ a (n) when showing the adaptive authorization
Output as ₀ (n). When the prohibition of adaptation is indicated, the changeover switch 17a is switched to output the amplified reception signal x ₀ (n), which is the output of the circuit on the local line L1, to the transmission signal u ₀ (n). The success / failure determiner 13a receives the call state signal S ₀ and the residual signal e ₀ (n), and outputs a signal for controlling whether to permit or inhibit the adaptive operation to the adaptive permission signal IN.
Output as H ₀ . The call control circuit 30a is connected to the station line L0.
Is determined by receiving a signal from the line interface 20a as to whether the telephone is in a call state or a call end state.
Output to the success / failure determiner 13a as ₀ .

FIG. 2 shows an example of the configuration of the adaptive filter 12. In FIG. 2, the adaptive filter 12 includes an adaptive filter group 101 for performing a filter operation, a tap coefficient correction circuit 110 for performing an adaptive operation, and a tap coefficient unit 111. Adaptive filter group 101 includes adaptive filter elements 102-1 to 102- (N-1) for obtaining unit delay time T,
Multipliers 103-1 to 103-N, adders 104-1 to 1
04- (N-1), and the pseudo echo signal d
_0p (n) is represented by d _0p (n) = Σh _k (n) u ₀ (nk) (1) Where Σ represents the sum from k = 0 to N−1, k also represents time kT, and h _k (n)
Is a coefficient to be multiplied by a signal delayed by kT time at the time nT, and represents a tap coefficient from each of the tap coefficient units 111-1 to 111-N applied to the adaptive filter group 101, and N is the number of taps of the adaptive filter group 101. Represents

The tap coefficient correction circuit 110 is constituted by an arithmetic circuit, and the adaptation permission signal INH ₀ is “adaptive prohibited”.
, The tap coefficient is not corrected to stop the adaptive operation, and h _k (n + 1) = h _k (n) (2) where k = 0 to N−1. When the adaptation permission signal INH ₀ indicates “adaption permission”, the adaptive operation is performed, and the tap coefficient correction circuit 110 outputs h _k (n + 1) = h _k (n) + μAe ₀ (n) (3) I do. Where A = u ₀ (n−k) (Nσ _N ² ) ⁻¹ , k = 0 to N−1, 0 <μ <2, μ is a correction coefficient, and σ _N ² is white noise N _Represents the variance (power) of ₀ (n). The second term on the right side of the equation (3) represents the correction amount of the tap coefficient, normalizes the transmission signal u ₀ (nk) by N · σ _N ² , and makes the transmission signal u ₀ (nk) proportional to the residual signal e ₀ (n).

FIG. 3 shows the structure of the success / failure judging unit 13, which includes a timer 131, a power detector 132, and a comparator 13.
3. It is composed of a judgment circuit 134.

The timer 131 outputs a judgment timing signal TM used in the judgment circuit 134. When the call state signal S ₀ (n) indicates the “end state”, the timer 131 is cleared and the determination timing signal TM is not output. When the speech state signal S ₀ (n) indicates “communication state”, the adaptive filter 12a when the transmission signal u ₀ (n), which is the input signal of the adaptive filter 12a, is white noise for a predetermined time τ. Is set to a value larger than the convergence time of τ. In this embodiment, τ = 75 ms.)
The determination timing signal TM is output every time. For example, the residual signal power p (n) can be obtained as a moving average value of the square amplitude e ₀ (n) ² of the residual signal e ₀ (n) using the following equation (4).

P (n) = λp (n−1) + (1−λ) e ₀ (n) ² (4) where λ is a forgetting coefficient that makes the weight smaller for past data, and 0 <λ Choose <1. The comparator 133 calculates the residual signal power p (n) and a predetermined residual signal power threshold PSH.
Are compared, and the comparison result is output as a comparison value CP (n). The determination circuit 134 receives the call state signal S ₀ , the determination timing signal TM, and the comparison value CP (n) as inputs, and outputs an adaptation permission signal INH ₀ .

FIG. 4 is a timing chart for explaining the operation of the decision circuit 134. When the call state signal S ₀ of the (a) is "end call state", the judgment timing signal TM of (b), regardless of the comparison value CP (n) of (e), the adaptive permission signal INH of (d) ₀ is "adaptation prohibited". When the communication state signal S _{0 in} (a) changes from the “end state” to the “communication state”, the adaptive permission signal I in (d)
NH ₀ is “adaptation permission”. Thereafter, if the comparison value CP (n) of (e) at the determination timing of the determination timing signal TM of (b) indicates that p (n)> PSH, the adaptation permission signal INH of (d) ₀ is left as “adaptive permission”, and at the determination timing executed for each cycle τ of the determination timing signal TM in (b),
If the comparison value CP (n) in (e) indicates that p (n) ≦ PSH, the adaptation permission signal INH ₀ in (d) is set to “adaptation prohibited”. After becoming once "adaptive prohibition" is, in (a) of the call status signal S ₀ is again "call end state" until the "call state", the judgment timing signal TM and (b) (e) Regardless of the state of the comparison value CP (n), the adaptation permission signal INH _{0 is set} to “adaptation prohibited”.

The operation of the bidirectional amplifier configured as described above will be described using the circuit on the local line L0 as an example.

The office line L0 in FIG. 1 outputs the call control circuit 30a when the call end state is the communication state signal S ₀ "call end state". At this time, the success / failure determiner 13a
Outputs "adaptation prohibition" to the adaptive permission signal INH _0. In addition, the values of the tap coefficients h _k (n) of the adaptive filter 12a are all initialized to 0 (or an adaptive value in the previous call state may be used).

Next, outputs the office line L0 is the call control circuit 30a becomes the call state is "call state" to the communication state signal S _0. The success / failure determiner 13a sets the adaptation permission signal INH _{0 when the} talk state signal S ₀ becomes “talk state”.
Is set to "adaptive permission", and the changeover switch 17a is controlled.
White noise N ₀ (n) as the transmission signal u ₀ (n)
And the adaptive filter 12a starts the adaptive operation.

During the adaptation operation, if there is no voice or modem signal on the local line L0, there is no noise during the adaptation operation.
The ₀ (n) only wraparound component d ₀ (n) of the hybrid circuit 11a is output. It faithfully reflects the wraparound characteristics of the hybrid circuit 11a. Therefore, in this case, the pseudo echo signal d _0p (n) approaches the wraparound component d ₀ (n) with time, and the residual signal e
₀ (n) approaches 0, and the residual signal power p (n) also approaches 0. On the other hand, if a signal such as a voice or a modem is present on the local line L0 during the adaptive operation, it acts as noise in the adaptive operation, so that the received signal y ₀ (n) faithfully reflects the looping characteristic of the hybrid circuit 11a. The transfer function (adaptive filter) obtained by the adaptive operation is far from the transfer function that generates the wraparound component d ₀ (n) in the hybrid circuit 11a.
Therefore, the wraparound component d ₀ (n) cannot be canceled using the pseudo echo signal d _0p (n), and furthermore, the residual signal e ₀
Since the signal of voice or modem is stored in (n), the residual signal power p (n) becomes a large value.

Therefore, the success / failure determiner 13a sets the residual signal power p at every fixed period τ from the start of “adaptive permission”.
Observe (n) and monitor the success or failure of the adaptation. That is, the residual signal power p (n) is compared with a predetermined residual signal power threshold PSH, and the residual signal power p (n) is compared.
If (n) is larger than the residual signal power threshold PSH, it is determined that some noise has mixed (failed). If it is determined to have failed, the adaptive permission signal INH ₀ is kept “adaptive permitted” in order to cause the adaptive filter 12 a to perform the adaptive operation again, and after a lapse of a predetermined time τ, the residual signal power p (n)
To determine whether the adaptive operation has succeeded or failed. These operations are repeated until it is determined that the operation is successful. When it is determined to be successful, the adaptive permission signal INH _{0 is set} to “adaptive prohibited”, the adaptive operation of the adaptive filter 12a is stopped (the tap coefficient is held), the switch is controlled, and the transmission signal u ₀ (n) is output. The amplified reception signal x ₀ (n) is output. Thereafter, the transmission signal u ₀ superimposed on the reception signal y ₀ (n)
The wraparound component d ₀ (n) of (n) is converted into a pseudo echo signal d _0p which is the output of the adaptive filter 12a that has stopped the adaptive operation.
Using (n), the difference is canceled by the subtractor 18a.

The above-described adaptive operation is performed in parallel in the circuit on the local line L1. When the above adaptive operation is completed in both the local line L0 and the local line L1, the received signal y ₀ (n) on the local line L0 is amplified by α ₀ times in the multiplier 19a, and the transmitted signal u on the local line L1 is amplified. ₁ (n), and the received signal y ₁ (n) on the local line L1 is amplified by α ₁ times in the multiplier 19b and output as the transmitted signal u ₀ (n) on the local line L0. . The gain constants α ₀ and α ₁ are set to values smaller than the amount of wraparound attenuation obtained by the hybrid circuit 11a and the adaptive filter 12a to prevent oscillation.

The success / failure determiner 13a performs the comparison using the power of the residual signal e ₀ (n).
The comparison may be performed using the voltage or the peak value of the residual signal e ₀ (n). This is because there is a relationship that the voltage is large when the power is large and the amplitude value is small when the power is small. When a signal having strong autocorrelation such as voice is present on the local line L0 during the adaptive operation, the residual signal e
_{Since 0} (n) also has a strong autocorrelation, the residual signal e
_The comparison may be performed using the autocorrelation value of the residual signal e ₀ (n) instead of the power of ₀ (n).

Next, an embodiment in which the bidirectional amplifier according to this embodiment is applied to a main unit of a telephone exchange will be described with reference to FIG.

FIG. 5 shows an example composed of a main unit and a plurality of telephones connected to the main unit. The main unit includes the station lines L0,
L1 is connected. The main units are office trunks 50-1 and 50-2 as office interfaces, extension trunks 60-1 and 60-2 as extension interfaces, a bidirectional amplifier trunk 10 according to the present invention, and a switching device for connecting the trunks. 80, a termination state code transmission unit 70, a highway bus HW, a microcomputer 90 for controlling the operation of the main unit, and a control bus 99.

The highway bus HW is connected to the office trunk 5
0-1, 50-2, extension trunks 60-1, 60-2,
Bidirectional amplification trunk 10, call end state code sending unit office line trunk 50-1 and 50-2 of a plurality of upward highway HW _U and switching device 80 towards the switching apparatus 80 from 70, the extension trunk 60-1 , Bidirectional amplification trunk 10
Consisting of a plurality of downward highway HW _D towards. Central office trunk 50-1 and 50-2, the voice from the central office line L0, L1, signals and the like to convert A / D, and sends the upward highway HW _U, the sound from the downward highway HW _D, a signal such as D
/ A conversion and send it to the office lines L1 and L0.

The extension trunks 60-1 and 60-2, the voice from the telephone set, a signal or the like to convert A / D, and sends the upward highway HW _U, the sound from the downward highway HW _D, a signal such as D / A Convert and send to phone. The end state code transmitting section 70 outputs a signal (for example, silence) indicating the end state to the upstream highway HW _U. The switching device 80 follows the instruction of the microcomputer 90 and
By W _U and switched connection the speech signal of the downward highway HW _D, the connection between the central office line L0, L1 and the phone connection between the telephones to each other, the connection of the station line L0, L1 and bidirectional amplification trunk 10, The telephone and the two-way amplification trunk 10 are connected to form a communication path. In addition, for trunks in the call end state (bidirectional amplification trunk 10, central office trunk 50, extension trunk 60), the upstream highway HW _U of the call end state code sending unit 70 is connected to the downstream highway HW of the trunk in the call end state. Exchange connection to _D. Microcomputer 90
Controls the operation of the main unit via the control bus 99.

FIG. 6 shows the configuration of the office trunk 50, which includes an office interface circuit 51, a hybrid circuit 52, an A / D conversion circuit 53, a D / A conversion circuit 56, a highway transmission circuit 54, and a highway reception circuit. It is composed of a circuit 55.

The office line interface circuit 51 is connected to the office line L0.
It has a DC closing circuit, a dial sending circuit, an incoming call detecting circuit, and a reverse detecting circuit (not shown). Each of these detection circuits notifies the microcomputer 90 of the detection result via a control bus 99, and a DC closing circuit and a dial sending circuit (not shown) are controlled by the microcomputer 90 via the control bus 99. . The hybrid circuit 52 includes two
Performs line-to-line conversion.

The A / D conversion circuit 53 converts an analog signal from the local line L0 into a digital signal (for example, a PC which is digitized by a coding rule called μ-1aw or A-1aw).
M signal). The highway transmission circuit 54 has an A /
According transmission provisions stipulated output of D converter circuit 53 in the main unit, and outputs the upward highway HW _U. The highway transmission circuit 54 outputs a silent signal during a period in which the line interface circuit 51 is not performing DC connection.

The highway receiving circuit 55 in accordance with the transmission provisions stipulated in the main unit, and inputs the digital signal from the downward highway HW _D. D / A conversion circuit 56 converts the digital signal input from the downward highway HW _D into an analog signal. The hybrid circuit 52, the A / D conversion circuit 53, and the D / A conversion circuit 56 included in the trunk trunk 50 in FIG. 6 are the hybrid circuit 11a, the A / D converter 15a, and the D / A converter 16a in FIG. Is equivalent to

FIG. 7 shows the configuration of the extension trunk 60. The extension interface circuit 61, the hybrid circuit 62, the A / D conversion circuit 63, the D / A conversion circuit 66, the highway transmission circuit 64, and the highway reception circuit 65 It consists of.

The extension interface circuit 61 is controlled by the microcomputer 90 via a control bus 99 to transmit and receive voice signals and control information to and from a telephone connected via a transmission line. The hybrid circuit 62 performs two-wire / four-wire conversion. The A / D converter 63 converts an analog signal from the telephone into a digital signal. Highway transmission circuit 6
4 outputs the output of the A / D converter 63 to the upstream highway HW _U in accordance with the transmission rules defined in the main device. Also, the highway transmission circuit 64 outputs a silent signal during a period when the extension interface circuit 61 is not transmitting and receiving a voice signal with the telephone. The highway receiving circuit 65 is connected to the downstream highway H
Inputting a digital signal from the W _D. D / A conversion circuit 66
Converts the digital signal input from the downward highway HW _D into an analog signal. The hybrid circuit 62, the A / D converter 63, and the D / A converter 66 of the present circuit
The hybrid circuit 11a, A / D converter 15a,
It corresponds to the D / A converter 16a.

FIG. 8A shows the structure of the end-of-call state code transmitting section 70, which is composed of an end-of-call state code generation circuit 71 and a highway transmission circuit 72. Ending state code sending section 7
Numeral 0 is a circuit for outputting a signal indicating a call end state, and outputs a constant code defined in the main device. As an example of this code, there is a silence code, and in the case of the μ-1aw code rule, 16
It is "00" in hexadecimal and "2A" or "AA" in hexadecimal in the case of A-1aw coding rule. Highway Bus H
When the signal polarity of W is negative logic, it becomes "FF" for the code μ-1aw coding rule and "D5" or "55" for the A-1aw coding rule. These codes represent silence, and are generally used in the main unit as end-of-call state codes for preventing click noise from being generated during a transition to a talk state.

FIG. 8 (b) shows an end state code generation circuit 71.
Shows the state of the end state code generation circuit 71 when the end state code 79 in which the error occurs is "00". Highway transmitting circuit 72 outputs the upward highway HW _U in accordance with the provisions stipulated the End state code 79 in the main device.

FIG. 9 shows the configuration of the bidirectional amplifying trunk 10. Bidirectional amplification trunk 10 is used to connect while performing amplifying the downward highway HW _D 1 and upward highway HW _U 2 and the downward highway HW _D 2 and upward highway HW _U 1. In the figure, the left and right sides of the dashed line at the center have exactly the same circuit configuration, so only the left side circuit will be described. The same components as those shown in FIG. 1 are denoted by the same reference numerals.

In FIG. 9, the highway receiving circuit 21a
(Abbreviated as R ₀ is in the figure) is the main device in accordance with a transmission provisions stipulated in the downward highway HW _D 1 from the PCM received signal c
₀ Enter (n). The PCM-to-linear converter 22a (abbreviated as P / L in the figure) converts the PCM received signal c ₀ (n) into a linear signal and outputs it as a reception signal y ₀ (n). The linear-PCM converter 23a (abbreviated as L / P in the figure) encodes a transmission signal u ₀ (n), which is a linear signal, and outputs it as a PCM transmission signal b ₀ (n). The highway transmission circuit 24a (abbreviated as T _{0 in the} figure) is a PCM
The transmission signal b ₀ (n) is output to the upstream highway HW _U 1 in accordance with the transmission rules defined in the main unit.

The call state detector 40a receives the PCM reception signal c ₀ (n) and outputs a call state signal S ₀ as a signal indicating the call state. Adaptive filter 12a, subtractor 18a, success / failure determiner 13a, changeover switch 17a,
The white noise generator 14a and the multiplier 19a are the same as those shown in FIG.
The configuration is the same as that described in the above. Therefore, FIG.
Output y ₀ of the A / D converter 15a (n) is transmitted via the highway bus HW ₀ 1 in, or input of the D / A converter 16a in FIG. 1 u ₀ (n) is the highway bus HW _U 1 , Except that the signal is transmitted through the circuit of FIG. However, the function of the call control circuit of FIG. 1 is different from that of the embodiment of FIG.
This is realized by a combination of the switching device 80, the microcomputer 90, and the end-of-call state code transmitting section 70 of FIG. When a call termination state is set by these combinations, the microcomputer 90 switches the call termination state code transmission unit 70 and the bidirectional amplification trunk 10 so as to switch connection.
To instruct. Therefore, the bidirectional amplification trunk 10 receives the PCM reception signal c ₀ received from the downstream highway HW _D 1.
By observing (n) in the call state detector 40a, the end state of the call can be known. Similarly, when making a call, the microcomputer 90 instructs the switching device 80 to switch and connect the two-way amplification trunk 10 to the office trunk 50 or the extension trunk 60. Therefore, the bidirectional amplification trunk 10 can know the call state by the fact that the call state detector 40a cannot detect the end state code of the PCM reception signal c ₀ (n) received from the downstream highway HW _D1 .

FIG. 10A shows the structure of the call state detector 40a in FIG. 9, which is composed of an end-of-call state code generation circuit 41, a collation circuit 42, and a judgment circuit 43. The end state code generation circuit 41 is a circuit that outputs a signal MUC indicating the end state, and is a fixed code defined in the main device (the end state code generation circuit of the end state code transmission unit 70 in FIG. 8). 71, which is the same as the end state code 79 output by the terminal 71. The collation circuit 42 collates the PCM reception signal c ₀ (n) with the end-of-call state code MUC at each time n, as shown in the flow of the operation in FIG. 10B (S1). (S1Y) Check signal PC (n) = "1"
2) If they do not match, the collation signal PC (n) is set to "0" (S
3) Output. The determination circuit 43 has a built-in counter UC, counts the collation signal PC (n), and calculates PC (n) =
And it outputs a "call end state" to call state signal S ₀ if the number of occurrences has become many "1".

In the call end state, the PCM reception signal c ₀
Since the end state code is received at (n), the occurrence probability of the collation signal PC (n) = "1" is 1 (each sample matches) unless there is interference due to noise or the like. In the talking state, the A / D converter 53 (FIG. 6) of the office trunk 50 or the extension trunk 60 is added to the PCM reception signal c ₀ (n).
63 (FIG. 7) are connected. If there is a voice or signal from the line L0, the PCM reception signal c ₀
Since (n) takes various values, the collation signal PC (n) =
The probability of occurrence of "1" will be much lower than 1. If there is no voice or signal from the local line L0, the A / D converters 53 and 63 should output a code representing silence. However, in practice, the A / D converters 53 and 63 perform quantization. Because of the presence of noise, the LSB of the PCM reception signal c ₀ (n)
Since the (least significant) bit changes, the collation signal PC (n)
= The occurrence probability of "1" is considerably smaller than 1.

As described above, the probability of occurrence of the collation signal PC (n) = "1" is considerably smaller than 1 in a call state regardless of the presence or absence of a voice or signal from the central office L0. Utilizing the above-mentioned empirically known statistical properties, when the occurrence probability of the collation signal PC (n) = "1" is close to 1, it is determined to be "end state" and the collation signal PC (n) = "1". When the occurrence probability is close to 0, the state is set to “communication state”.

FIG. 11 shows an operation flow of the judgment circuit 43 shown in FIG. 10 (a).
The counting operation of C and the determination of the call state are performed. Hereinafter, the “end state” is represented by S ₀ = “0” and the “call state”
Is represented by S ₀ = “1”. The counter UC (not shown) included in the determination circuit 43 uses a reset counting method as a counting method, and when S ₀ = “1” (S11N),
When the call signal is detected and the verification signal PC (n) = 0, (S12
Y), the counter UC is reset to the initial value 0 (S1)
3), continue to output the call state S ₀ = “1” (S1)
4). When the collation signal PC (n) = "1" (S12N),
Since no call signal is detected, the value of the counter UC is increased (S15). If the value of the counter UC has not reached the predetermined value M (for example, 47) (S16
N), the value of the counter UC is held as it is (S17),
The determination circuit 43 keeps the communication state signal S ₀ = “1” (S18). If no call signal is detected in the next cycle, the process goes to step S11.
11N is selected, and the operations of S12 to S18 are repeated. When the value M (for example, 47) of the counter UC is detected (S16Y), the value of the counter UC is reset (S16Y).
19), the call status signal S _{0 is set} to “0” (S 20),
Displays the end state.

When the call state signal S ₀ is “0” and the call end state is displayed (S 11 Y), when no call signal is detected, PC (n) = “1” (S 21 Y), and the counter UC of the counter UC is set. value is held at a UC = 0 (S22), call status signal S ₀ be kept of S ₀ = "0", and displays the end state (S23).

Next, from this state, the call state (PC (n))
= “0”) (S21N), the output of the A / D converters 53 and 63 of the local trunk 50 or the internal trunk 60 is connected to the PCM reception signal c ₀ (n), and c ₀
A code (for example, a voice code) that is not the end state code is output to (n). In the present embodiment, the end state S ₀ =
The upper limit value N of the counter UC at “0” is set to 1. This is because the noise mixed into the highway HW or the like can be ignored in practice, and when the end state code is not detected about twice consecutively, that is, when the speech code is detected twice consecutively, it is regarded as a call state. Because you can do it. Therefore, a code other than the end state code is output from the matching circuit 4
2. If two consecutive detections are made in P2 (P
C (n) = “0”), the determination circuit 43 determines that the state is “communication state” (S24Y), resets UC = 0 (S25),
Output the S ₀ = 1 (S26) to. The counter value UC is N
When not reached (S24N), the value of the counter 1 advances (S27), reaches a set count value N, that remain held in the "0" to call state signal S ₀ (S28).
So a call signal is detected (S21N), because they reached the count value UC = N (S24Y) counter is reset (S25), call status signal S ₀ becomes "1" (S26), call status Is displayed.

Since the operation is performed as described above, the call end state (S ₀
When = speech signal when in "0") is detected, immediately S ₀ = "1" is changed to display a call status, becomes once call state, the detection call signal is for example 47 times in succession Then, S ₀ = “0” and the end state is displayed. The error rate of the highway HW is 10 ⁻¹¹ , S ₀
= The occurrence probability of the verification signal PC (n) at "1" is 1 /
When 2, N = 1 and M = 47, the average time for which the determination circuit 43 keeps outputting the correct call state signal S ₀ is several hundred years or more, so that the correct call state signal S ₀ is always output. You can think of it as continuing.

In the description of the speech state detector 40 in FIG. 10, the termination state code is determined to be one type in the main unit, and the explanation has been made on the assumption that the speech state detector 40 is aware of it. It is not always necessary to specify the end state code. That is, the call state detector 40 can operate without knowing what kind of code is used as the end state code in the main apparatus. FIG. 12 shows the configuration of such a call state detector 40B. Show.

In FIG. 12, the matching circuit 42 and the judgment circuit 43 are the same as those described with reference to FIG. The collation circuit 42 outputs the PCM of the current cycle in the input code string.
The received signal c ₀ (n) is compared with the PCM received signal c ₀ (n−1) one cycle past from the delay circuit 44, and if they match, a matching signal PC (n) = “1” is output, and no match is found. In this case, the matching signal PC (n) = "0" is output. As described above, since one kind of continuous code is never output from the A / D converters 53 and 63 of the office trunk 50 or the extension trunk 60, the collating circuit 42 for performing such collation is used. The reason why “1” is continuously output to the output PC (n) is that the bidirectional amplifying trunk 10 is connected to the termination state code transmission unit 70, that is, when the trunk is in the complete termination state. Limited.

The call state detection circuit 40B receives the coded reception signal c ₀ (n) as an input signal, and receives the coded reception signal c ₀ (n) as an output of the PCM-linear converter 22a (FIG. 9). The same purpose as described above can be achieved even if ₀ (n) is used. However, in this case, the reception signal y ₀ (n) has a larger number of bits than the PCM reception signal c ₀ (n) (for example, the μ-law PCM signal is 8 bits and is converted to a linear signal). 1 in case
4 bits), and the size of the matching circuit 42 becomes large.
In the description of the main device of FIG.
Although the time required for the control of 0 was ignored, it was actually 10 times.
The switching device 80, the office trunks 50-1 and 50-2, and the extension trunks 60-1 and 60-2 of FIG.
Are performed. Therefore, the actual line trunk 50
-1 and 50-2 are connected to the office lines L0 and L1, and the microcomputer 90 instructs the switching device 80 to switch the bidirectional amplification trunk 10 and the office line trunks 50-1 and 50-2. And there is a time difference of several hundred ms. If the call state detectors 40 and 40B of the present invention (FIGS.
If 2) is not used and the bidirectional amplifier trunk 10 is also informed of the call state via the control bus 99 by an instruction from the microcomputer 90, several hundred ms will be added to the time difference. Assuming that the instructions of the microcomputer 90 are given in a random order, the bidirectional amplifying trunk 10 starts the adaptive operation according to the instructions of the microcomputer 90 before the switching device 80 performs the switching connection,
Microcomputer 90 to bidirectional amplification trunk 10
May be delayed. If the adaptation operation starts before the exchange connection is completed, the adaptation operation is performed in a state where the hybrid circuit is not connected, so that an erroneous operation occurs, and the microcomputer 9
If the instruction from 0 is delayed, the adaptive operation will not be performed at an ideal time, which will hinder communication. However, according to the present invention, since the call state detectors 40 and 40B are used, the adaptation operation can be completed at a time immediately after the connection to the office line where there is little possibility of interfering with the communication. If the method of knowing the call state by providing the call state detectors 40 and 40B according to the present invention is used, the call state can be known directly from the PCM reception signal input to the two-way amplification trunk 10, so that the microcomputer 90 However, the influence of the time required for control can be almost ignored.

[0059]

As is apparent from the above description, according to the present invention, the training time is secured immediately after the line connection is completed by providing the success / failure determiner in the bidirectional amplifier of the preset type echo canceller system. In addition, even in a communication in which a communication signal is superimposed immediately after the completion of the line connection, a correct adaptive operation can be performed by using a silent period of the communication existing thereafter. In addition, a signal having a small autocorrelation such as white noise can be used as a training signal, and the convergence time is extremely short.
(For example, 75 ms) , the training is successful if a slight silence period exists during communication. The initial stage of communication is a full-duplex communication transmitted by both the caller and the recipient, regardless of whether it is a human conversation or a communication between machines. In most cases, there will always be silence. Therefore, the adaptive operation can be completed in the initial stage of the communication, and the amplification can be performed in both directions from that point, so that the user does not feel uncomfortable or the communication is not interrupted. In addition, since the printed circuit board on which the bidirectional amplifier of the present invention is mounted can be replaced with a conventional bidirectional amplifier in a main unit of an existing exchanger, the present invention can be applied to a conventional main unit. is there.

Further, the training can be performed without being restricted by the control time of the microcomputer. Therefore, the effect of the present invention is extremely large.

[Brief description of the drawings]

FIG. 1 is a circuit diagram showing an embodiment of the present invention.

FIG. 2 is a circuit configuration diagram of an embodiment of an adaptive filter which is a component included in the circuit shown in FIG. 1;

FIG. 3 is a circuit configuration diagram of an embodiment of a success / failure determiner which is a component included in the circuit shown in FIG. 1;

FIG. 4 is a timing chart showing an operation of a determination circuit which is a component of the success / failure determiner shown in FIG. 3;

FIG. 5 is a circuit configuration diagram when the bidirectional amplifier of the present invention is incorporated in a main unit of a telephone exchange.

6 is a circuit configuration diagram of a central office trunk which is a component of the main device shown in FIG.

FIG. 7 is a circuit configuration diagram of an extension trunk which is a component of the main device shown in FIG. 5;

8 is a circuit diagram of an end state sending unit which is a component of the main apparatus shown in FIG. 5 and a code diagram showing an end state code generated from the end state transmitting unit.

9 is a circuit configuration diagram of an embodiment of a bidirectional amplification trunk according to the present invention, which is a component of the main device shown in FIG.

10 is a circuit configuration diagram of a call state detector which is a component of the bidirectional amplification trunk shown in FIG. 9 and a flowchart showing an operation flow of a matching circuit included therein.

11 is a flowchart showing an operation flow of a determination circuit which is a component of the call state detector shown in FIG.

FIG. 12 is a circuit configuration diagram of another embodiment of the call state detector.

[Explanation of symbols]

Reference Signs List 10 Bidirectional amplification trunk 11a, 11b Hybrid circuit 12a, 12b Adaptive filter 13a, 13b Success / failure determiner 14a, 14b White noise generator 15a, 15b A / D converter 16a, 16b D / A converter 17a, 17b Switching Switch 18a, 18b Subtractor 19a, 19b Multiplier 20a, 20b Local line interface 21a, 21b Highway receiving circuit 22a, 22b PCM-linear converter 23a, 23b Linear-PCM converter 24a, 24b Highway transmitting circuit 30a, 30b Communication control Circuits 40a, 40b Communication state detector 41 Termination state code generation circuit 42 Matching circuit 43 Judgment circuit 50 Office trunk 51 Office line interface circuit 52 Hybrid circuit 53 A / D converter 54 Highway transmission circuit 55 C Way receiving circuit 56 D / A converter 60 extension trunk 61 extension interface circuit 62 hybrid circuit 63 A / D converter 64 highway transmitting circuit 65 highway receiving circuit 66 D / A converter 70 termination state code transmission section 71 termination state Code generation circuit 72 Highway transmission circuit 79 Termination state code 80 Switching device 90 Microcomputer 99 Control bus 101 Adaptive filter group 102 Adaptive filter element 103 Multiplier 104 Adder 110 Tap coefficient correction circuit 111 Tap coefficient unit 131 Timer 132 Power detector 133 Comparator 134 Judgment circuit α ₀ , α ₁ Gain constant b ₀ (n), b ₁ (n) PCM transmission signal c ₀ (n), c ₁ (n) PCM reception signal d ₀ (n), d ₁ (N) Sneak component d _0p (n), d _1p (n) Pseudo echo signal e ₀ (N), e ₁ (n) residual signal h ₀ (n) to h _N-1 (n) tap count p (n) residual signal power τ cycle u ₀ (n), u ₁ (n) transmission signal _{x 0 (n), x 1} (n) amplified received signal _{y 0 (n), y 1} (n) received signal CP (n) comparison value HW highway HW _D down highway HW _U upward highway INH _0, INH ₁ adaptive permission signal L0, L1 office line _{n 0 (n), n 1} (n) white noise PC (n) matching signal PSH residual signal power threshold S _0, S ₁ call state signal TM judgment timing signal

Claims (3)

(57) [Claims] A white noise generating means for generating white noise during a silent period of a communication signal in order to measure in advance a quantity of a part of a transmission signal as a wraparound signal in a reception signal; Attenuating and delaying the white noise during the silent period to obtain a pseudo echo signal (d _op (n)), and superimposing the pseudo echo signal on the reception signal so that the reception signal has a minimum value; An adaptive filter means (12) for performing an operation for optimizing; and determining that the operation for optimizing is completed when the operation for optimizing is completed within the silent period, and holding the state of the adaptive filter means; If the operation for optimizing is not completed within the time period, it is determined to be a failure, and the operation for optimizing is performed in the next silent period by the adaptive filter means (1). 2) Repeat and execute, the success and failure
In the determination of loss, the received signal and the pseudo echo signal
(D _op (n)) and the residual signal (e ₀ (n))
Residual signal power using smaller weights for past data
(P (n)), and the residual signal power (p (n)) is
Success when below a predetermined level (PSH), predetermined
To determine failure if the level is higher than (PSH)
Bidirectional amplifier including a success / failure determination means (13). 2. A part of a transmission signal wraps around a reception signal.
Communication to measure in advance the amount that appears as a signal
White noise during silent periods of a signal
White noise generating means (14) for attenuating and delaying the white noise during the silent period.
To obtain a pseudo echo signal (d _op (n))
The received signal is the minimum value
Adaptive filter operating to optimize as shown
The means (12) and the operation for optimizing are completed within the silent period.
Sometimes it is determined to be successful and the state of the adaptive filter means is changed.
The operation for holding and optimizing within the silence period is
If it is not completed, it will be judged as failure and will occur next
In the silent period, the adaptive operation is
Successful deletion of the order allowed to run repeatedly to other means (12)
Loss determination means (13), encodes the transmission signal and the reception signal, and when the code value of the current cycle and the code value of one cycle before match with each other frequently is determined, it is determined that the call has ended, and the match is determined. If the frequency is low, it is determined that the call is in the call state, and the call state detection means (40, 4
0B) and a bi-directional amplifier, including. 3. A transmission signal partly wraps around a reception signal.
Communication to measure in advance the amount that appears as a signal
White noise during silent periods of a signal
White noise generating means (14) for attenuating and delaying the white noise during the silent period.
To obtain a pseudo echo signal (d _op (n))
The received signal is the minimum value
Adaptive filter operating to optimize as shown
The means (12) and the operation for optimizing are completed within the silent period.
Sometimes it is determined to be successful and the state of the adaptive filter means is changed.
The operation for holding and optimizing within the silence period is
If it is not completed, it will be judged as failure and will occur next
In the silent period, the adaptive operation is
And repeats the execution,
In the determination of loss, the received signal and the pseudo echo signal
(D _op (n)) and the residual signal (e ₀ (n))
Using one of the voltage value, peak value and autocorrelation
The smaller the weight of the previous data, the lower the value of the predetermined level (P
Success when SH) or less, predetermined level (PSH)
Success / failure determination method for determining failure when
A bidirectional amplifier comprising a stage (13) .