JPH06315173A - Transmission controller - Google Patents

Abstract

PURPOSE:To compulsorily lower the noise level at the time of transmitting a control signal by measuring reception power after a control signal is eliminated, not imparting attenuation to a reception signal when power is large and imparting attenuation when power is small. CONSTITUTION:A power measuring means measures the power of an input signal. Set values P1 and P2 in two comparators to be the components of a gain setting means and signal power are compared, and which one of three ranges of >=P2, <=P1 and between P1 and P2 signal power is discriminated. When signal power is >=P2, gain G0 (gain 1) is outputted, when signal power is <=P1, gain G1 (smaller than gain 1) is outputted and when signal power is between P1 and P2, gain proportional to signal power is outputted. A variable amplifier means amplifies an input signal by the gain determined by the gain setting means. A memory such as ROM, etc., is used as the setting means.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a transmission control device, and more particularly, it uses a line such as a two-wire type dedicated line to transmit and receive control signals such as a line capture signal or a dial signal within a voice signal band, The present invention relates to a transmission control device (in-band ringer) that corrects attenuation characteristics and the like.

[0002]

2. Description of the Related Art An in-band ringer is a device for transmitting a control signal such as a voice signal or a line capture signal within a voice signal band (0.3 to 3.4 kHz) by using a line such as a dedicated line. In such a device, it is common to use a four-line dedicated line because it is easy to obtain good call quality and has a simple structure. However, the 4-wire system has a problem that the maintenance cost of the leased line is high, and therefore there is a demand to realize the same function and performance with the 2-wire system dedicated line. 20
It is a block diagram of the whole system at the time of cascade-connecting a two-wire type | formula exclusive line via the relay PBX which performs long distance transmission.

In the in-band ringer, a control signal is superimposed and transmitted during dialing or during a call. However, it is necessary to remove the control signal in the in-band ringer so that the speaker cannot hear it. In order to remove the control signal, the signal frequency sent by the transmitter and the center frequency of the receiver filter on the receiver must match exactly.Therefore, use an oscillator or filter composed of analog parts such as resistors and capacitors. If it is used, there is a problem that complicated adjustment is required and it is difficult to manufacture. As a method of solving this, there is a method of realizing no adjustment by configuring an oscillator or a removal filter by a digital signal processing technique using an accurate reference frequency.

[0004]

In order to perform digital signal processing in order to accurately remove a control signal used between in-band ringers, all signals including a voice signal must be converted into a quantized signal. Quantized signals always contain quantization noise, and if there is a speech signal and its power is large, the quantization noise power is relatively small, so there is no problem, but if there is no speech signal, the quantization noise does not occur. Has a problem that it becomes relatively large and gives discomfort to the speaker. In particular, in an analog-digital converter that performs non-linear amplitude quantization such as μ-law, since the quantization interval of a large amplitude part is wide, when a control signal with a large amplitude is converted, a relatively large level of white noise occurs, Since the white noise remains after the control signal is removed, if the audio signal power at that time is small, the white noise will be heard as an offensive sound because the masking effect does not work. . It is an object of the present invention to reduce noise, which is noticeable when there is no sound, in a transmission control device to the extent that it does not cause discomfort.

[0005]

According to the present invention, the received power after removing the control signal is measured, the received signal is not attenuated when the power is high, and the received signal is attenuated when the power is low. It is characterized in that the noise level during control signal transmission is forcibly reduced.

[0006]

In the present invention, when the voice signal is silent, there is little discomfort even if the amplitude is extremely reduced. Therefore, the received signal power is measured, and when the power is small, the received signal is attenuated according to the power. By applying, the white noise caused by the quantization of the reception control signal can be attenuated without giving a feeling of strangeness to the speaker.

[0007]

21 is a block diagram showing the hardware structure of an in-band ringer according to the present invention. The in-band ringer 1 is connected between the 2-wire line and the PBX2,
The PBX2 is, for example, a 6-wire SR dedicated line interface (TTC standard JJ-21.10) as shown in FIG.
Connected by. Transmission line 4WS, reception line 4WR
Transmits audio signals in each direction, and the uplink control line S
The S and downlink control lines SR transmit control signals for calling, dialing, responding, etc. in the respective directions. In addition to this, as an interface with the PBX or the like, there is also a method of superposing a transmission / reception voice signal and a control signal such as a line capture signal, a call signal, and a dial signal on a two-line transmission line.

DSP (Digital Signal Processor) 1
Reference numeral 0 is a CPU having a high-speed arithmetic function. Figure 23 is D
It is a block diagram which shows an example of a structure of SP. Since various DSPs are commercially available and well known, detailed description thereof will be omitted. However, by incorporating an appropriate program into the DSP, various digital signal processing such as filtering, signal generation, and detection becomes possible. It is also possible to control an external circuit. H13 is a 2-wire to 4-wire conversion circuit, BN
14 is a balance circuit network including a plurality of circuits for balancing H13 (these circuits will be described later),
The AGC 15 is a variable gain amplifier (which will also be described later) for compensating for the amount of signal attenuation on the dedicated line.

The ADCb16 outputs the H13 on the 4-wire side (AG
The Cb) is an analog-digital converter for converting a digital signal, and the DACb12 is a digital-analog converter for converting a digital signal into an analog signal. ADCa11 is an analog-digital converter for converting an input audio signal into a digital signal, and DACa17 is a digital-analog converter for converting a digital signal into an analog signal. The photocoupler 18 receives the SS signal and outputs it to the control bus 20 of the DSP 10. The relay 19 sends a control signal to the SR line under the control of the DSP 10.

FIG. 1 is a functional block diagram showing the function of the two-wire in-band ringer of the present invention. The function of the in-band ringer of the present invention is selected by setting means (for example, a switch) not shown at the time of installation, and the function is set to the master station device or the slave station device. In the present invention, the in-band ringer set in the slave station first transmits a test signal for measuring the characteristics of the leased line, and the in-band ringer set in the master station is
In response to this, setting of amplifier gain and the like is started. The setting for the master station or the slave station distinguishes which of the test signals is transmitted first and which of the master station frequency and the slave station frequency is used as the control signal. Since the master station and the slave stations have the same configuration, each component will be described with reference to FIG. 1, and then the operation will be described with reference to FIG. 2 and subsequent figures.

In FIG. 1, ECb is a 2-line to 4-line converter H.
, An echo canceller for estimating the transfer function of the wraparound component, ADb is a subtractor for suppressing the wraparound by subtracting the output of ECb and the output on the 4th line of H, WHNa is a white noise generator, and SWb is the selection of the balance network BN. Alternatively, it is a changeover switch for inputting white noise to the 4-wire side input of H when the adaptive filter tap coefficient of the echo canceller ECb is set.

MOD is a modulator for modulating dial pulses, OSC is a variable frequency oscillator for outputting a control signal having one frequency or two frequencies, and BEFa is an occupied band of the control signal output from the MOD or OSC. A band elimination filter that removes the component from the voice signal input from the PBX side, SWa is a switch for passing the entire band of the voice signal in the call mode, DET is one frequency or two
A signal receiver that receives a control signal consisting of a frequency, measures its power, and determines the presence / absence of a signal, DEM is a demodulator that demodulates a dial pulse modulated wave, and SWc passes the entire band of a voice signal in a call mode. Switch for BE
Fb is a band elimination filter that removes a component corresponding to the band of the control signal from the signal received from the dedicated line.

The EQ has a plurality of equalizer curves, one of which is used to correct the high frequency attenuation of the received voice signal, and the NR corresponds to the power of the received voice signal in order to reduce the quantization noise. Noise reducer that changes gain, W
HNb is a white noise generator, SWd is a switch for inputting white noise to the receiving line 4WR when the tap coefficient of ECa is set, and ECa is the transfer function of the wraparound component generated in the 2-wire 4-wire conversion circuit of the PBX subscriber circuit. The echo canceller ADa for estimating is a subtracter for subtracting the output of ECa from the signal of the transmission line 4WS and suppressing the wraparound. CONT
Controls the functional blocks thus over the control bus. IN
T is an interface section with SS line and SR line, and S
Transfers control signal received from S line to CONT,
The control signal from the ONT is transmitted to the SR line.

Here, a signal processing method will be described for the above-mentioned blocks having complicated functions. The variable gain amplifier AGC whose circuit diagram is shown in FIG. 22 is configured by an operational amplifier, and the amplification gain is one of a plurality of feedback circuits (R1 to Rn).
This is realized by selecting one of them, and the selection of the feedback circuit is performed by giving a binary code to the analog switch SW.

The echo cancellers ECa and ECb are
When white noise is being input to the 2-line to 4-line converter, the subtractor A
This is a preset echo canceller that adjusts the tap coefficient so that the output of Da or ADb is minimized and holds the value after the adjustment of the tap coefficient. For example, a learning identification method is used for the identification algorithm. ECa and EC
During the training of b, the white noise is applied at the same level as the voice signal so that the training is surely performed.
In the case of ECb, there is no problem because it is performed during the initial setting operation, but in the case of ECa, since training is performed after the PBX speech path is formed, the speaker hears it as a large unpleasant noise. In order to reduce the discomfort, the called side performs training immediately after the line acquisition is detected or the calling side immediately after the acquisition response signal is stopped so that there is a masking effect due to a voice switching sound. In the white noise generators WHNa and WHNb, for example, a random signal generated in advance by a computer is stored in the ROM and is read out to generate the random signal. The modulator MOD is realized by phase modulation that outputs a positive-phase sine wave during the break period of the dial pulse and outputs a negative-phase sine wave during the make period, for example. In the case of phase modulation, signal power is always present and can be detected by the single frequency detecting means, which is convenient. Variable frequency oscillator O
In SC, for example, when the oscillation frequency is fi and the sampling frequency is fs, the nth sine wave sample value is represented by the real part or imaginary part of the following equation.

Exp (j2πnfi / fs) = exp (j2π (n-1) fi / fs)
・ Exp (j2πfi / fs). Therefore, by sequentially calculating the right side of this equation, that is, the coefficient cos (2πfi / fs) and si corresponding to the oscillation frequency are
By performing a complex operation using n (2 πfi / fs), a sine wave having a desired frequency can be generated. Band stop filter BEFa
Alternatively, BEFb is composed of an IIR filter, and the denominator coefficient and the numerator coefficient corresponding to a plurality of blocking characteristics are RO
These coefficients are stored in M, and these coefficients are selected and calculated according to the needs of the communication procedure. In removing the two-frequency signal, the coefficient is selected twice during the processing of one sample, and the filter operation is performed.

The signal detection in the signal receiver DET is shown in FIG.
As shown in FIG. 6, first, the electric power Pa in all the bands of 0.3 to 3.4 kHz is measured, and then the signal band is selected by the band pass filter that passes the sine wave of the desired frequency. The electric power Ps of the signal is measured, and in the detection determination, it is determined that the desired signal is present when the ratio Ps / Pa is 1 (actually, noise or an error causes a value rp smaller than 1). When the ratio Ps / Pa is small, the received signal contains a large amount of components other than the desired signal, and there is a high possibility of false detection. In DET, power measurement is realized by calculating the average within a time window, the band pass filter is composed of an IIR filter, and the denominator coefficient and the numerator coefficient corresponding to a plurality of pass characteristics are stored in ROM. Every time, these coefficients are selectively calculated according to the detected frequency, and in the case of 2 frequencies, it is realized by calculating twice like the band stop filter. The demodulator DEM is composed of a bandpass filter and a detector that determines the phase change of the captured signal, and demodulates the dial pulse. The equalizer EQ and the noise reducer NR will be described in detail later.

Next, the procedure for entering the initial setting operation of the in-band ringer according to the present invention and the standby state after completion of the initial setting operation will be described with reference to FIGS. 2 (a), 2 (b), 6 (a) and 6
An explanation will be given using (b). When installing the in-band ringer, the contractor connects the transmission line with the PBX or the like and the two-wire dedicated line, sets the master station device or the slave station device, and turns on the power.

When the power is turned on, the main station device is shown in FIG.
As shown in (a), the standby signal S11 having two frequencies is set to the OS.
Output from C. At this time, SWa falls to the BEFa side,
BEFa is programmed to block the two frequencies that make up S11. (Unless otherwise specified, when the master station device or the slave station device outputs a control signal from OSC or MOD, SWa falls to the BEFa side, and BEFa is programmed to block the same frequency component as the output signal, and SWc is also set to the BEFb side when receiving the control signal). The master station device outputs a reset signal S12 of one frequency after a lapse of a certain time (for example, 1 second) to instruct the slave station device to start the initial setting operation, and the slave station device for measuring the attenuation characteristic that also serves as a response. Waits for the test signal S23 to be sent. Test signal S2
When 3 is received, the operation shifts to the initial setting operation. The master station device receives the reset signal S2 from the slave station device in the standby state.
When 2 is received, the reset signal S12 is output, and the initialization operation is activated in the same sequence as when the power is turned on.

As shown in FIG. 6B, the slave station device outputs the standby signal S21 consisting of two frequencies for a certain period of time immediately after the power is turned on, and then outputs the reset signal S22 to start the initial setting operation to the master station device. It gives an instruction and waits for the main station device to send a reset signal S12 which also serves as a response. Upon receiving the reset signal S12, the slave device immediately outputs the attenuation characteristic measuring test signal S23 and shifts to the initial setting operation.

FIG. 2A shows the sequence until the master station device is powered on first and the master station device starts the setting operation, and FIG. 2B shows the slave station device is powered on first. The sequence until the setting operation is started in the master station device is shown.
As is apparent from FIGS. 2 (a) and 2 (b), according to the procedure of the present invention, the control procedure is synchronized regardless of the power-on sequence of the master station slave stations. The initial setting of is started, and the slave station outputs the signal required for it. That is, by synchronizing, one of the two devices outputs a test signal, and based on that, the other learns and corrects the characteristics of the leased line, thereby enabling initial setting without human intervention.

The reason why the standby signal is transmitted prior to the reset signal is that even if the initial setting is forcibly started by a reset switch or the like during the operation of the in-band ringer, both of them are put in the standby state. This is because the shift signal is transferred and the reset signal is reliably detected. In other words, the standby signal consisting of two frequencies can be easily distinguished from the voice signal, and the possibility of false detection due to noise is extremely low, so that the in-band ringer can reliably receive the signal and forcibly wait. Can return to the state. In the standby state, the voice signal is not output to the dedicated line, and even a single frequency signal such as a reset signal can be reliably detected. Although it is possible to configure the reset signal with two frequencies, it is disadvantageous to use both the standby signal and the reset signal as a two-frequency signal because of the circumstances described below, and only the standby signal has a two-frequency configuration.
In other words, 2 kHz or less of the 3.4 kHz band is not assigned to a signal that forcibly changes the operation because there is an MF signal for dial transmission. Frequency required, 3k
At around Hz, there is a reason that it is unsuitable for use as a control signal due to large attenuation by a dedicated line. Therefore, it is detected that both the reset signal and the standby signal are pushed into a narrow band as signals of dual frequency configuration. Will be at a disadvantage.

Next, the initial setting operation according to the present invention will be described. As shown in FIG. 7A, the master station device sets the AGC and EQ when receiving the test signal S23. The test signal S23 is a signal composed of two frequencies of a frequency fc (for example, 1.5 kHz) near the center of the transmission band and a frequency fe (for example, 3 kHz) near the end of the transmission band, and is transmitted from the slave station device with the same amplitude for both frequencies. To be done. The signal of frequency fc is a signal that also serves to measure the amount of attenuation of the dedicated line and to notify the start of the initial setting operation of the slave station device, and the signal of frequency fe is a signal for measuring the amount of attenuation in the high frequency band. The fe may be multi-frequency in order to improve the measurement accuracy. The main station device detects the test signal S23, that is, the frequency fc by the signal receiver DET, refers to the amplifier gain value table stored in the ROM based on the fc power at that time, and refers to the variable gain amplifier AGC to the corresponding gain value. To set.

Next, the main station apparatus measures the power of the frequency fe (since the processing is realized by time division processing, fc and fe are
Power measurements are performed simultaneously, that is, within one sample. ), And the equalizer characteristic is selected based on the ratio between this and fc power. The method of selecting the equalizer characteristic will be described in detail later. The main station device outputs the end signal S14 after the selection of the equalizer characteristic is completed. If the fc power is exhausted during S14 transmission, the process proceeds to the selection of the balance circuit network BN. The stop of the test signal S23 (fc) should detect the silence itself, but it is difficult to detect it directly because of the presence of noise. Therefore, the silence detection is replaced by the fact that the test signal S23 cannot be detected (silence detection thereafter is the same). The balance circuit network BN is selected with white noise applied from the WHNa to the 2-wire to 4-wire conversion unit H, which will be described later in detail.

After selecting the balance circuit network, the main station device performs the training of the tap coefficient of the echo canceller ECb as described above, and after the training is completed, the end signal S15.
Is sent. When the test request signal S25 is received from the slave device during the transmission of S15, the output of the test signal S13 is started. The test signal S13 has the same frequency structure as S23. During the period when S13 is sent, the slave station device
It is the time for setting the GC and EQ and is set in advance. If the end signal S24 cannot be detected even after the period has passed, it is considered that the dedicated line is broken or the slave station device is broken, and an alarm is displayed. When S24 is received, the signal transmission is stopped so that the slave station device does not interfere with the training by outputting white noise by itself. For silent transmission, if the end signal is not received after a certain period of time, it is determined that the wire is broken or there is a failure, and an alarm is displayed.

The end signal of the slave station is also used as the standby signal S21 since all the initial setting operations have been completed. The standby signal S21 is output from the slave device following the white noise, but the standby signal component is also included in the white noise, so that the master station may make an erroneous detection unless some measures are taken. In order to prevent erroneous detection, in the DET as described above, the ratio between the total band power and the desired signal power is referred to, and when both are approximately the same, it is determined that the true end signal has been received. In this way, the sending side does not have to remove the end signal component from the white noise in advance and send it. If a certain component must be removed from the white noise to perform BN selection or tap coefficient setting of the echo canceller, no information can be obtained about the removed band, and thus correct selection or setting cannot be performed. .

The initial setting operation of the slave device is started from the transmission of the test signal S23 as shown in FIG. 7 (b). S2
If the end signal S14 is not sent from the main station device even after a certain time elapses during the sending of 3, the alarm line is displayed because the disconnection of the dedicated line or the failure of the main station device is considered. If S14 is received, signal transmission is stopped, timer monitoring is performed, and if a certain time elapses, it is regarded as disconnection or failure and an alarm is displayed. If the end signal S15 is detected by the DET within a fixed time, the test request signal S
25 is output to wait for reception of the signal (fc) near the center of the transmission band included in the test signal S13. If S13 is detected, the gain of the AGC is set, the equalizer characteristic is selected, and the end signal S24 is output. When the test signal S13 is no longer detected after sending S24, white noise is applied to the 2-wire to 4-wire converter to train the balance circuit BN and the echo canceller ECb. After these operations are completed, the standby signal S21 that also serves as the termination signal is output and the standby state is entered.

In the alarm display described above, the main station device or the slave station device lights the alarm lamp and remains in the alarm state in which the standby signal is sent to the other device, and the reset is performed after the cause of the alarm is manually removed. Then, the initial setting operation starts again. FIG. 3 is a sequence diagram showing the initial setting operation described above with the master station device and the slave station device facing each other.

Next, a method of selecting the equalizer characteristic will be described with reference to FIGS. 12 and 13. The dedicated line has various amplitude frequency characteristics as shown in FIG. That is, when the leased line is a short distance, the amplitude frequency characteristic is almost flat as shown by C4, when it is a medium distance, the attenuation in the high range is larger than that in the low range as shown by C3 or C2, and when the long line is long. C
As shown in Fig. 1, the attenuation difference between the low range and the high range becomes extremely large. If this state is left as it is and the correction by the equalizer is not performed, the clarity of the voice is deteriorated and the call is disturbed. In FIG. 12, fc is a frequency near the center of the transmission band described above, and fe is a frequency at the end of the transmission band. In the figure, the amplitude frequency characteristic is expressed by standardizing the amplitude value at fc,
Further, r1, r2, and r3 are threshold values for equalizer characteristic determination, and are represented by using the attenuation amount (ratio) at fe.

Next, a method of selecting the equalizer characteristic will be described with reference to FIG. When the test signal is received, first the amplitude level Lc at fc is measured, and then the amplitude level Le at fe is measured.
Is measured, the ratio r = Le / Lc between the two is calculated, and the calculated ratio r
And r1, r2, r3 are compared in magnitude to select the corresponding correction characteristic, that is, the inverse characteristic of the amplitude frequency characteristic shown in FIG. In the equalizer setting method of the present invention, for example, r
For a dedicated line measured as a value in the range of r1 or more and less than r2, the optimum characteristic is determined as the inverse characteristic of C2 regardless of the line type and the transmission distance. Therefore, if the number of correction characteristics to be prepared is small, the risk of overcompensation or undercompensation increases. 12 and 13, the number of characteristics is 4 for convenience of explanation.
Although there are books, at least about eight books are actually necessary. In addition, the measurement of the amplitude level is made in DET,
The output of the bandpass filter is used. The selected equalizer characteristic is stored in the ROM as the tap coefficient of the FIR filter, and the selected tap coefficient is transferred to the coefficient RAM configuring the FIR filter and held until the initialization operation is performed again thereafter. The method of the present invention is completed in an extremely short time as compared with the method of manually inputting a measurement signal from the other side, measuring the signal level by the measurement side and manually selecting the optimum correction filter (bandpass filter). If the output is averaged and the measurement time is set to 100 ms, the measurement can be performed at two frequencies simultaneously, so the selection operation is completed in about 100 ms).

Next, the selection of BN will be described with reference to FIGS. 14 and 15. FIG. 14 is a circuit diagram showing the 2-line to 4-line converter H and the balance circuit network BN of FIG. In the figure, Z11 and Z21 in the frame indicated by BN
Constitutes a pair of voltage dividing circuits, and includes selection switches S11 and S2.
Selected by closing 1. Similarly, Z12 to Z14 and Z22 to Z24 also constitute voltage dividing circuits, and are selected by the selection switches S12 to S14 and S22 to S24. S11 to S14 and S21 to S24 are realized by analog switch ICs. ZL represents the impedance of the dedicated line, and Z0 is paired with ZL to form a voltage dividing circuit. Two
The line-to-line conversion unit is a bridge circuit, and the transmission signal is applied between points c and d and output to ZL.
If the potential of the connection point a is equal to the potential of the voltage dividing circuit connection point b in the BN, no voltage is generated between both points, and the transmission signal does not sneak into the reception signal.

Balance network Z11, Z21, Z12
Z22, Z13 / Z23, and Z14 / Z24 are calculated so that the divided potentials are as equal as possible in consideration of some cases of ZL (four cases in FIG. 14) that vary depending on the line type and the transmission distance. It is designed using. Next, a selection method for selecting an optimum one from these balance circuit networks will be described with reference to FIG. First, a signal (white noise) having a flat spectrum in the band is applied between points c and d in FIG. Next, the provisional minimum power Pmin is set to an extremely large value, and the balance network number Nmin for outputting the minimum power is set.
Is initialized to 0. Next, the balance network number n is set to 1, S11 and S21 in FIG. 14 are closed, the received power generated between points a and b in FIG. 14 is measured as the total power Pa in DET, and Pmin and Pa are compared.

Since Pa is always smaller when n = 1, Pmin = Pa and Nmin = 1. Next, set n = 2 and close S12 and S22 (open S11 and S21),
The received signal power Pa is measured, and Pmin and Pa are compared. If Pa is smaller, Pmin = Pa and Nmin = 2 are updated. If Pa is larger, Pmin and Nmin are not updated, and the next balance is calculated. In order to select the network, n is incremented and n = 3. Similarly, after selecting a balance network and measuring Pa, Pmin and Nmin are updated, and when the measurement is completed for all balance networks, the number stored in Nmin is taken as the optimum balance network number. , Fix the state of the selection switch. After that, this state is maintained until the initial setting operation is restarted.
For the sake of convenience of explanation, the number of balance circuit networks is four, but in reality, since the number of line types of dedicated lines is large, at least about eight pairs are necessary to reduce the amount of wraparound. BN of the present invention
In comparison with the method of manually selecting the wrap-around amount for a large number of frequency points and selecting the optimum one from, for example, eight or more sets, the selection of is clearly efficient and the setting time is extremely shortened.

Next, the operations in the standby state, the calling state, the called state and the talking state will be described with reference to FIGS.
8 to 11 are flowcharts showing the operation of the master station device, the slave station device also has exactly the same operation, and only signal symbols (for example, S11 and S21, S19 and S29, etc.) are different.

In the standby state, the standby signal S as shown in FIG.
11 is transmitted, and when the reset signal S22 from the slave station is received, the operation shifts to the initial setting operation, and when the call origination of the own station is detected, the transmission of the standby signal S11 is stopped and the operation transits to the call origination state.
If the capture signal S29 is received from the slave station, the standby signal S11
When the standby signal S21 is received from the slave station, it remains in the standby state, and when none of the signals can be received for a certain period of time or more, it is a disconnection of the leased line and the operator is disconnected. An alarm is displayed to inform you.

In the calling state, as shown in FIG.
19 is sent to notify the partner device of line capture. After sending the capture signal S19, the system waits for the capture response signal S28 to be sent from the partner device. If the capture response signal cannot be detected within a certain period of time, an alarm is displayed because the leased line is broken. When the capture response signal S28 is received, pulse dial modulation is performed. Pulse dial modulation is the acquisition signal S1
The frequency of 9 is used as the carrier wave. In the case of the PB dial signal, the pulse is not transmitted from the PBX, so the capture signal S19 is not modulated. It waits for the acquisition response signal S28 to be stopped at the same time as the pulse dial modulation (including the case where it is not modulated). In the embodiment of the present invention, no control signal is output in the call state, and all the bands are provided for the call. That is, the called device receives the capture signal and outputs PB while calling the PBX in parallel with the capture response signal output.
When there is a response from X, the capture response signal is stopped.

When the stop of the capture response signal from the partner device is detected, a silence is output, the voice signal is released in all bands, and a call state is entered. However, since there is a wink start response method described later, a timer is provided. After confirming that the capture response signal has been stopped for a certain period of time or more, the communication state is entered. Training of the echo canceller ECa is performed immediately after the stop of the acquisition response signal is detected. Since ECa identifies the transfer function of the detour path including the PBX speech path as already described,
It is necessary to perform training immediately after the calling side PBX forms a communication path, that is, immediately after the calling side in-band ringer notifies the PBX of the called response. However, since there is a wink start response system described later, a timer is provided and the tap coefficient is fixed only when the stop of the capture response signal is detected for more than the duration of the wink pulse and the state is changed to the call state.

In the called state, the capture response signal S18 is first output as shown in FIG. At the same time, the pulse dial is demodulated. In the case of the PB dial, since the capture signal S29 is not modulated, the dial pulse is not output to the PBX. If the capture signal S29 from the partner device is not received for a certain period of time before the local station stops the capture response signal S18, an alarm is displayed because it is a disconnection of the dedicated line. When it detects the line capture (response) of its own station, it outputs a silent sound, opens the entire band to the voice signal, and at the same time, trains the echo canceller ECa. In the case of a call, since the PBX communication path is formed immediately after the response is detected, the ECa is trained at this point to fix the tap coefficient and shift to the communication state. However, since there is a response system by wink start, which will be described later, a timer is provided and the tap coefficient is fixed only when the response is detected for the longest duration of the wink pulse and the state is changed to the talking state.

In the call state, as shown in FIG. 11, silence is output as the control signal, and the standby signal S21 is detected and the recovery is detected (line release detection). Standby signal S21
If is detected, it returns to the standby state after detecting the recovery of its own station. If recovery is detected before the other device, the standby signal S
It outputs 11 and waits for a response by the standby signal 21 from the other party. In the embodiment according to the present invention, since the control signal for detecting the disconnection is not sent during the call, the disconnection cannot be detected directly. However, if a disconnection occurs during a call, the speaker cannot continue the call and should hang up the handset. It is assumed that the speaker on the main station side puts the handset first. If the standby signal S11 is output and the wire is not broken, the other party also puts the handset within a few seconds and the standby signal S21 of the partner device is received. Utilizing this fact, the standby signal S21 is transmitted from the own station, and after a lapse of a certain time, the standby signal S21
If is not received, the disconnection alarm is displayed. When the standby signal S21 can be received within the fixed time, the process returns to the standby state.

As described above, when the control signal is transmitted in each state, it is detected that the scheduled response is returned within the scheduled time and the disconnection is detected, so that the disconnection can be detected at any time. In FIGS. 8 to 11, in the disconnection alarm state, a standby signal is output, the alarm lamp is lit up and remains in that state, and when the forced reset is performed, the initial setting operation is started.

FIG. 4 is a sequence diagram for explaining a case where the main station device makes a call in the calling state, the called state, and the call state described above, and notifies the end of the call first. The dial tone or ring back tone in the figure is output by the called PBX. FIG. 5 is a diagram in which the sequence until the ringback tone is output in FIG. 4 is replaced with the sequence executed by the wink start method using the SR leased line interface. In the wink start method, the calling side PBX uses the accumulated dials for the called side PBX.
The wink pulse emitted by the BX is received and then transmitted. In the embodiment of the present invention, the called device modulates and outputs the wink pulse in response to the stop of the capture response signal, and the calling device demodulates and transmits it to the PBX. In this case, since the capture response signal stops for a certain period of time, it will coincide with the called response, but the calling device holds the calling state for the time corresponding to the duration of the wink pulse, and captures for a predetermined time or longer. If the response signal continues to stop, the call state is entered.

Next, the training of the ECa in the called state will be described with reference to FIG. In the case of wink start, the state of the PBX during the output of the wink pulse is not in the state of forming a speech path, so if training is performed during this period, an incorrect loop transfer function will be identified. That is, as shown in FIG. 17, when the wink start method is not adopted (a), the communication path is formed at the time when the response is made, whereas when the wink start method is adopted (b), when viewed from the in-band ringer. Since the line is captured twice, the communication path is not formed in the first response, and the communication path is formed immediately after the second response. PBX for those who use in-band ringer
It is desirable not to have different inband ringer models or different settings even if the handshake method between them is different. Therefore, the training of the echo canceller must be performed normally regardless of whether the wink start method is used or not.

In order to solve this, it is necessary to always re-train immediately after the line is captured as shown in FIG. 17 (c). Therefore, in the called state of the embodiment of the present invention, as shown in FIG. 10, when the response is detected, the timer is started, and while the timer does not expire, the ECa is trained regardless of whether the PBX communication path is formed or not. When is expired, the tap coefficient is fixed. In addition,
In the ECa training, the wraparound attenuation amount is measured, and the tap coefficient is not updated when the wraparound attenuation amount exceeds a predetermined amount. In addition, since training is performed using white noise, the training time is short (50
within ms). Therefore, there is no direct relationship between the time of the timer set when the response is detected and the ECa training, and the training is completed in a time sufficiently shorter than the wink pulse width. Explaining the relationship between the training of the calling ECa and the wink start method, PB
At the time when the speech path of X is reliably formed, the calling side in-band ringer detects the stop of the acquisition response signal and sends the response signal to the PB.
Since it is the time of outputting to X, if the training result of ECa immediately before the transition to the talking state is adopted as shown in FIG. 9, normal training can be performed regardless of whether the wink start method or not.

Next, the noise reducer NR will be described. In the calling state, the capture signal and the capture response signal are mutually transmitted using a part of the transmission band from the time the speaker captures the line until the other party responds. Then, these control signals are removed by the band stop filter BEFb on the receiving side, and the rest is heard by the speaker as a call signal. In particular, the calling party speaker puts his / her ear on the ear during transmission of these control signals, so that it is uncomfortable to hear noise caused by the control signals that cannot be removed by BEFb. More specifically, since the control signal is transmitted at a level equivalent to the speech level, if a non-linear AD converter such as a PCM codec is used, the quantization interval becomes coarse in a large amplitude portion and the quantization noise increases. That is, when the control signal is AD-converted, quantization noise (white noise) having a wide band spectrum is added to the spectrum of the original signal. The original signal spectrum is removed by BEFb but the broadband noise is not removed. Since the level of this broadband noise is lower than that of the voice signal, it is masked when the voice signal is being transmitted at the same time as the control signal, so that there is no problem. However, the presence of voice signals is rare in the calling state,
The calling speaker perceives noise.

In the embodiment of the present invention, the noise reducer NR is inserted after BEFb in order to reduce the noise generated by the AD conversion to the extent that it is not perceived. 18 and 1
The operation of the noise reducer NR will be described with reference to FIG. When a high power signal such as a voice signal is superimposed on the quantization noise, it is masked and noise is not a concern, so no processing is required. On the other hand, when the quantization noise occupies most of the power, it is perceived and is attenuated. This is expressed in FIG. 18, where the amplification gain is G1 (attenuation) when the signal power is P1 or less, and G0 when the signal power is P2 or more.
(Gain 1), and if P1 or more and P2 or less, the gain is determined in proportion to the signal power. FIG. 19 is a flow chart for explaining the operation. First, the signal power P is measured. The power is measured, for example, by calculating the moving average over a 5 ms window. Next, the measured power P is compared with the threshold values P1 and P2 to determine the gain according to the power range. Since it is not necessary to operate the noise reducer in a call state, for example, the measured power P is forcibly set to a value of P2 or more and the operation is stopped. When the noise reducer according to the embodiment of the present invention is not inserted, the quantization step is uniform regardless of the signal amplitude in order to reduce the quantization noise, and the resolution equivalent to the non-linear quantization is provided. Therefore, a high resolution analog-to-digital converter is required.

FIG. 24 is a functional block diagram showing the internal structure of the noise reducer for realizing the function of the flowchart of FIG. The power measuring means measures (calculates) the power of the input signal. Two comparators, which are components of the gain setting means, compare the values of P1 and P2 in FIG. 18 with the signal power, and the signal power is P2 or more, between P2 and P1, P
It identifies which of the following three ranges it belongs to. In the gain calculation means, when the signal power is P2 or more, the gain G0
(Gain 1) is output, and gain G1 (G1
<G0, for example 0.3) is output, and when it is between P1 and P2, the gain G is proportional to the signal power (G1 <G <G0), for example, G = G1 + {(P-P1) / (P2 -P1)} ・
(G0-G1) is output. The variable amplification means amplifies the input signal with the gain determined by the gain setting means. It is also possible to use a memory such as a ROM as the setting means.

[0047]

As is apparent from the above description, the present invention has the following effects. (A) quantizing noise perceived when there is no audio signal,
Since it is detected that the quantization noise is the dominant constituent power of the signal power and the attenuation is performed only in that case, the noise can be reduced without giving an abnormality to the voice signal. (B) By inserting a noise reducer, it is possible to prevent the speaker from perceiving unpleasant noise without using an expensive analog-digital converter with high resolution.

[Brief description of drawings]

FIG. 1 is a functional block diagram showing a configuration of an exemplary embodiment of the present invention.

FIG. 2 is an operation sequence diagram immediately after reset.

FIG. 3 is a sequence diagram of an initial setting operation.

FIG. 4 is a sequence diagram including a call state.

FIG. 5 is a sequence diagram in a wink start system operation.

FIG. 6 is a flowchart illustrating a reset process.

FIG. 7 is a flowchart illustrating an initial setting operation.

FIG. 8 is a flowchart illustrating a standby state operation.

FIG. 9 is a flowchart illustrating a calling state operation of the master station device.

FIG. 10 is a flowchart illustrating a called state operation of the master station device.

FIG. 11 is a flowchart illustrating a call state operation of the main station device.

FIG. 12 is an explanatory diagram of equalizer characteristics.

FIG. 13 is a flowchart illustrating an equalizer characteristic selection operation.

FIG. 14 is a circuit diagram of a 2-wire / 4-wire conversion unit and a balance circuit network.

FIG. 15 is a flowchart illustrating a balance circuit network selecting operation.

FIG. 16 is a flowchart illustrating detection of a received signal.

FIG. 17 is a time chart showing a wink start method.

FIG. 18 is a characteristic explanatory diagram of the noise reducer.

FIG. 19 is a flowchart illustrating a noise reduction operation.

FIG. 20 is a configuration diagram of a network in which two-line dedicated lines are cascade-connected via a relay PBX.

FIG. 21 is a hardware block diagram of the device of the present invention.

FIG. 22 is a circuit diagram of an AGC circuit.

FIG. 23 is a block diagram showing an example of the internal configuration of a DSP.

FIG. 24 is a functional block diagram showing an example of an internal configuration of a noise reducer.

[Explanation of symbols]

1 ... in-band ringer, 2 ... PBX, 10 ... DSP, 1
1, 16 ... A / D converter, 12, 17 ... D / A converter,
13 ... 2-wire to 4-wire conversion circuit, 14 ... Balance circuit network, 15
… AGC circuit, 18… Photo coupler, 19… Relay

Claims (2)

[Claims] 1. A transmission control device for controlling a call origination / reception by a digital signal processing using a two-wire line, a band elimination filter unit for eliminating a control signal from a received signal, and the band elimination filter unit. Measuring means for measuring the power of the output signal; gain setting means for setting a gain value according to the output value of the measuring means; and variable gain amplifying means whose gain is controlled by the output gain value of the gain setting means. The transmission control device, wherein the gain setting means reduces the gain below a normal value when the output value of the measuring means is below a predetermined level. 2. The gain setting means comprises a first comparing means for comparing the output of the measuring means with a first predetermined value, and a second predetermined value smaller than the output of the measuring means and the first predetermined value. Based on the outputs of the second comparing means for comparing the value with the two comparators, the first setting value is output when the output of the measuring means is larger than the first setting value, and the output of the measuring means is second output. When the output of the measuring means is between the first set value and the second set value, the second set value is set between the first set value and the second set value. 2. The transmission control device according to claim 1, further comprising setting means for outputting a value proportional to the output value of the measuring means.