JPH09233002A - Voice switch - Google Patents

Abstract

PROBLEM TO BE SOLVED: To prevent reception/transmission speech blocking to prevent production of howling and to suppress a sense of reverberation by tentatively inserting a prescribed amount of attenuation to a transmission side speech circuit when a received voice signal is interrupted and a silence state is generated. SOLUTION: A switch control circuit 10 controls an inserted attenuation of attenuation circuits 20, 21 based on a transmission voice signal received from a microphone 2 and a received voice signal sent from a remote end side terminal equipment. Delay circuits 30, 31 delay the transmission and reception voice signals by the same delay time as that of a signal from the switch control circuit 10. The switch control circuit 10 detects an echo of the received voice signal bypassed to the microphone 2 while a remote end talker speaks as a transmission voice signal. When the received voice signal is interrupted and a silence state is generated, the switch control circuit 10 decreases the inserted attenuation of the attenuation circuit 20 in matching with a change in smoothed fluctuation of the transmission voice signal. Thus, the received voice signal at a remote end side is kept to a prescribed level below a voice detection threshold level.

Description

Detailed Description of the Invention

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice switch for voice calls. In a hands-free call system that consists of a telephone and video conferencing equipment that have a hands-free function that allows you to talk with a speaker and a microphone without a handset, acoustic coupling between the speaker and the microphone, and a side-blocking circuit on the transmission path It is necessary to prevent howling that occurs when the gain of the one-cycle closed circuit formed by the sidetone coupling due to 1 exceeds 1. The voice switch is used as one of the means, and howling is prevented by reciprocally inserting the attenuation amount into the transmitter / receiver communication circuit based on the transmitted / received sound detected in the communication terminal. It has a configuration.

FIG. 14 is a diagram showing a closed circuit constituted by a hands-free call system. In this case, the echo of the sound output from the speaker 3 wraps around the microphone 2 on the near end side to cause acoustic coupling, and the echo of the sound output from the speaker 3 ′ wraps around the microphone 2 ′ on the far end side. A circular loop is formed by the acoustic coupling. Further, when the call terminal is connected to the two-wire transmission line, the sidetone coupling that occurs in the sidetone prevention circuit on the transmission line and the sound that occurs when the returned sidetone is output from the speaker 3 and wraps around the microphone 2. The combination also forms a circuit cycle.

Therefore, a voice switch is provided on each of the terminals on both sides to judge the current communication state by comparing the received voice and the transmitted voice detected in the terminal, and the voice signal from the speaker side is passed. Attenuation amount is not inserted in the call circuit on the other side, and a large amount of attenuation is inserted in the call circuit on the opposite side, so that the gain of the formed loop circuit is controlled to be always less than 1. There is. That is, the voice switches of the terminals on both sides are such that the attenuating circuit 20,
At the same time when the insertion attenuation amount of 21 'is set to 0, the attenuation circuit 21,
The insertion attenuation amount of the attenuation circuits 20 and 21 'is increased while the insertion attenuation amount of the attenuation circuits 21 and 20' is set to 0 during the utterance of the far-end speaker. As a result, the voice of the speaker is transferred to the other side without being attenuated, and the gain of the closed loop is less than 1, so that howling is suppressed.

When the howling is prevented by using the voice switch as described above, the amount of attenuation is always inserted in either one of the transmitting side / receiving side communication circuits, so that the communication is necessarily unidirectional. , This unidirectionality does not make the conversation impossible because the speakers alternate in the conversation.
However, if the switching of the communication direction by the voice switch is slow, problems such as disconnection of the talk occur, and therefore it is necessary to switch the voice switch as quickly as possible in order to provide a smooth conversation.

By the way, an echo canceller is well known as a howling preventing means other than a voice switch.
Echo canceller drastically solves one-way communication
It is compatible with two-way calling and howling prevention. However, since the echo canceller requires a certain amount of convergence time for changes in acoustic coupling, there is an undeniable possibility that howling may occur due to a decrease in the amount of echo cancellation (that is, an increase in the closed loop gain) during that time. Therefore, even in a hands-free telephone terminal incorporating an echo canceller, a configuration is generally adopted in which a howling is suppressed by using a voice switch together. In addition to such high availability,
The voice switch has very high practical value because it has a simple structure, small size, and low price.

[0005]

2. Description of the Related Art FIG. 15 shows a configuration example of a conventional voice switch. This voice switch is provided in the hands-free call terminal for making a call with the microphone 2 and the speaker 3, and the attenuation circuits 20 and 21 and the power calculation filter 210 are provided.
(S) and 210 (r), and the determination circuit 200.

The power calculation filter 210 (s) calculates the power P _S of the transmission voice input from the microphone 2, and the power calculation filter 210 (r) calculates the power P _R of the reception voice transmitted from the far end side. Calculate. Under the control of the determination circuit 200, the attenuating circuits 20 and 21 insert the amount of attenuation into the transmitting side talking circuit and the receiving side talking circuit, respectively. The determination circuit 200 compares the power P _{R of the} reception voice with the power P _{S of the} transmission voice, determines that the reception state is established if aP _R ≧ P _S , and sets the insertion attenuation amount of the reception side communication circuit to 0. Then, the attenuation circuits 20 and 21 are controlled so as to increase the insertion attenuation amount of the communication circuit on the transmitting side, and if aP _R <P _S , it is judged to be the transmitting state, and the insertion attenuation of the communication circuit on the receiving side is performed. The attenuation circuits 20 and 21 are controlled so that the amount is increased and the insertion attenuation amount of the transmitter side speech circuit is set to zero. Here, a is a constant determined from the magnitude of the acoustic coupling assumed in advance.

[0007] Here, the problem occurs because the echo of the received voice that has sneak into the microphone 2 from the speaker 3 cannot be distinguished from the transmitted voice that the near-end speaker has input to the microphone 2. For example, during the utterance of the far-end speaker, the power of the echo of the received voice that has sneak into the microphone 2 from the speaker 3 on the near end side and the received voice sent from the far-end side are compared. since input to the microphone 2 and the echo is slightly delayed speech, and when both are compared to vary violently level, are fully contemplated may become an aP _R <P _S (transmission state). In this case, since the voice switch inserts a large amount of attenuation into the attenuation circuit 21 in order to switch the call direction, a phenomenon occurs in which the received voice is cut off while the far-end speaker is speaking. This phenomenon is called reception blocking and continues until the power of the subsequent reception voice becomes large and it is determined that aP _R ≧ P _S (reception state).

When the transmitted voice sent from the near-end side to the far-end side returns to the near-end side as an echo due to acoustic coupling on the far-end side, the same process as in the case of receiving blocking is performed. The phenomenon that the transmitted voice is cut off may occur. This is called transmission blocking, but since it is the same phenomenon as reception blocking, the following description deals with reception blocking and omits transmission blocking.

In order to suppress the reception blocking with this configuration, the time constant of the power calculation filter 210 (r) of the reception voice is set to be as long as the reverberation time of the room (the time until the echo attenuates and falls below the ambient noise level). At the same time, by increasing the constant a and considering the power of the received voice to be larger than the actual power, it is possible to reduce the possibility that aP _R <P _S (transmitting state). However, if the time constant of the power calculation filter 210 (r) is lengthened, the receiving state (during the utterance of the far-end speaker) is forcibly extended even if the received voice becomes silent, and the constant a is increased. Then, a transmission voice with a large power is required to switch from the reception state to the transmission state. Therefore, if this means is taken, the switching operation of the voice switch will be delayed,
It becomes difficult to perform the smooth switching operation required for a normal conversation in which the speaker changes in time.

In order to solve this problem, the following voice switch control method has been proposed. That is, the power of the transmitted / received voice is not compared, but the presence / absence of the transmitted / received voice is determined, and the switching control is performed based on the following basic rule. When only the transmitted voice is detected, if the current state is the receiving state, the state is switched to the transmitting state. When only the received voice is detected, if the current state is the transmitting state, the state is switched to the receiving state. When both the transmitted and received voices are detected, the current state is given priority and the call state is not switched. When neither the transmitted voice nor the received voice is detected, the current state is prioritized and the call state is not switched.

By applying this basic rule, for example, in the receiving state, even if the power of the echo of the received voice sneaking into the microphone becomes larger than the power of the received voice sent from the far end side, the received voice continues. Is detected (=
The listening state is maintained (if the far-end speaker is speaking) and the listening blocking does not occur. However, with only the control according to this basic rule, the voice switch malfunctions in the following cases, and reception blocking occurs.

Even when the far-end speaker is speaking, a temporary short silence period may occur immediately before the explosive consonant or the consonant. During the period, the echo of the received voice is slightly delayed and input to the microphone 2, and when it is detected as the transmitted voice, the state is switched to the transmitted state according to the above rule.
However, if the reverberation time is longer than this silent period,
Even if the received voice is detected again, the transmission state is maintained by the above-mentioned current priority rule until the echo is attenuated and disappears. As a result, a phenomenon such as a delay in the return to the receiving state and a disconnection of the consonant part of the talk occurs.

Usually, even if the received voice from the far-end speaker is temporarily silenced due to a burst consonant or the like, the near-end speaker rarely speaks during that time, and even if there is, it is important to the extent of a confusion. It is considered to be infrequent. Based on this idea, the following exceptions are made to the above basic rules. That is, after the switching from the reception state to the transmission state, the reception priority state is set at least within the reverberation time, and the reception state is immediately returned when the reception voice is detected regardless of the detection of the transmission voice. In addition, the transmission blocking is similarly solved by providing the transmission priority state.

FIG. 16 shows an example of the configuration of a voice switch having a blocking prevention function depending on the receiving / transmitting priority state. This is provided with voice detection circuits 310 (s) and 310 (r) in place of the power calculation filters 210 (s) and 210 (r) in the conventional configuration shown in FIG. 15, and a transmission priority setting circuit 320 (s). And a receiving priority setting circuit 320 (s) is newly provided.

The voice detection circuit 310 (s) detects the presence / absence of transmitted voice input from the microphone 2, and the voice detection circuit 310 (r) detects the presence / absence of received voice transmitted from the far end side. The transmission priority setting circuit 320 (s) and the reception priority setting circuit 320 (r) control the duration of each priority state, and terminate the priority state when a predetermined time has elapsed after the transition to the priority state. Notify the determination circuit 300. The determination circuit 300 includes the voice detection circuits 310 (s) and 3 (s).
10 (r) and the notification from the transmission priority setting circuit 320 (s) and the reception priority setting circuit 320 (r), the switching of the communication state and the insertion attenuation amount of the attenuation circuits 20 and 21 are controlled.

The operation of this voice switch will be described with reference to the state transition diagram of FIG. (A) Priority on the current state of the receiving state When the voice switch is in the receiving state, the voice detecting circuit 310
If (s) and 310 (r) both detect the transmitted / received voice, or if neither of them is detected, the determination circuit 300 maintains the current state according to the above basic rule or. (B) Transition from the receiving state to the receiving priority state However, the receiving voice is interrupted and becomes silent during the receiving state,
When only the transmission voice is detected, the determination circuit 300 switches to the reception priority state, inserts the attenuation amount into the attenuation circuit 21 on the reception side, and removes the attenuation amount from the attenuation circuit 20 on the transmission side. (C) Returning from the receiving priority state to the receiving state Since the call direction in the receiving priority state is the same as that in the transmitting state,
The transmitted voice input from the microphone 2 is transferred to the far end side. However, during this period, the current priority rule is not applied, and even if the transmitted voice is being detected, if the received voice is detected, the state returns to the receiving state. As a result, the head is not cut off when the received voice is temporarily silenced and then resumed, and if the near-end speaker starts speaking during the temporary silence of the received voice, The voice is transferred to the far end.

Since the receiving priority state is a transient state for preventing the blocking of the reception caused by the echo of the reception voice, there is no reason to maintain the state over the reverberation time. Therefore, we set a limit on the duration of the listening priority state,
After that time has passed, it is assumed that the state transitions to another state according to the above-mentioned basic rule. For example, the reception priority setting circuit 320 (r) notifies the determination circuit 300 of the termination of the reception priority state when the reverberation time elapses after the interruption of the reception voice is detected by the voice detection circuit 310 (r).
After this, if only the received voice is detected, the determination circuit 300
Switches to the listening state. To return from the receiving priority state to the receiving state, the attenuation amount is inserted in the attenuation circuit 20 on the transmitting side,
This involves control for removing the amount of attenuation from the attenuation circuit 21 on the receiving side. (D) Transition from the receiving priority state to the transmitting state Further, when only the transmitting voice is detected after the reception priority setting circuit 320 (r) notifies the end of the receiving priority state, the determination circuit 300 transmits the transmission. Switch to the state. Since the call direction is the same in the receiving priority state and the transmitting state, the call direction switching control is not involved at this time. The case where the state changes from the receiving state has been described above. However, when the state changes from the transmitting state, similarly, the state becomes the transmitting priority state and then returns to the transmitting state or transits to the receiving state.

[0018]

However, even when the voice switch having this structure is used, a problem arises when a hands-free communication system is constructed by facing a communication terminal equipped with it. This problem will be described with reference to FIG. Immediately after the interruption of the received voice is detected on the near end side, the echo of the received voice wraps around the microphone 2 and is detected as the transmitted voice, so that the voice switch transits to the receiving priority state and the attenuation circuit 20 on the transmitting side. Insertion attenuation is 0
become. Therefore, the echo that sneaks into the microphone 2 is sent back to the far end side. There is no problem if the far-end speaker resumes speaking before the echo sent back is detected by the voice switch on the far-end side. In this case, since the voice switch on the far end side maintains the transmission state, the echo sent back from the near end side is blocked by the attenuation circuit 21 ', while the voice transmitted by the far end speaker is transmitted to the near end side. Transferred. As a result, the near-end voice switch detects the received voice and immediately returns to the receiving state.

The problem arises when the echo sent back as described above is detected before the far-end speaker speaks. In this case, the voice switch on the far end side immediately transits to the transmission priority state, and the insertion attenuation amount of the attenuation circuit 21 'becomes 0, so the echo returned from the near end side is output from the speaker 3'. Then, the voice switch on the far end side detects the echo sneaking into the microphone 2'as a transmission voice, and immediately returns to the transmission state and returns the insertion attenuation amount of the attenuation circuit 20 'to zero. As a result, the echo is sent back to the near end side, where it is detected as a received voice. As a result, the voice switch on the near end side returns from the reception priority state to the reception state, and the insertion attenuation amount of the attenuation circuit 21 is returned to 0.
The echo sent back is output from the speaker 3 and further circulates around the microphone 2. In this way, a phenomenon occurs in which the echo repeatedly circulates through the closed circuit even in a short time, which causes howling.

The present invention has been made in view of the above problems, and it is an object of the present invention to provide a voice switch capable of preventing reception / sending blocking, suppressing howling, and suppressing reverberation. To aim.

[0021]

The present invention is applied to that the transmitting side communication circuit and the receiving side communication circuit are provided with attenuation means for inserting an attenuation amount into the communication circuit, respectively. A voice switch used for a hands-free function that suppresses howling by inserting a large amount of attenuation in the circuit and by inserting a large amount of attenuation in the talking circuit on the transmitting side in the receiving state, enabling a call using a microphone and a speaker. Is.

In the present invention, in order to solve the above-mentioned problems, as a first form, when a state transition from a receiving state to a transmitting state occurs due to a receiving voice circulated as an echo in a microphone, the other party is temporarily There is provided a voice switch characterized in that the received voice transmitted to the terminal side is inserted into the transmitting side communication circuit with an attenuation amount such that the received voice is equal to or lower than the voice detection level at the partner terminal side.

With such a configuration, even if the received voice circulated as an echo to the microphone is sent back to the partner terminal side, there is no possibility that the partner terminal determines it as the speaker's voice and malfunctions.

In a second aspect of the present invention, when a state is changed from a receiving state to a transmitting state by a receiving voice which is used in a hands-free call system having a side-tone prevention circuit and which circulates as an echo in a microphone. Temporarily, the component of the received voice transmitted to the other terminal side, which has sneak into the own receiving side communication circuit through the protection side sound circuit, becomes below the voice detection level in the receiving side communication circuit. A voice switch is provided, characterized in that a large amount of attenuation is inserted into the transmitter side speech circuit.

With such a structure, there is no possibility that the echo returning around the anti-side sound circuit is judged to be the received voice sent from the partner terminal side and malfunctions.

In the voice switches of the first and second modes, the echo sneaking into the microphone is below the voice detection level in the talk side speech circuit during the period in which the attenuation amount is inserted into the talk side speech circuit. Can be up to.

With such a configuration, the period for inserting the attenuation amount into the transmitter side speech circuit can be suppressed to a necessary minimum, so that it is possible to avoid the disturbance of conversation.

The voice switches of the first and second modes can be constructed so that the amount of attenuation inserted in the transmitter side speech circuit is gradually reduced according to the reverberation characteristic. Such a configuration further includes a reverberation monitoring filter for observing the power change of the echo that wraps around the microphone, and the attenuation amount to be inserted into the transmitter side speech circuit in accordance with the power change of the echo output by the reverberation monitor filter. It is also possible to control.

With this configuration, the voice can be detected on the other terminal side if the speaker speaks even during the period in which the attenuation amount is inserted in the talk side communication circuit. Can be switched more smoothly to provide high call quality.

Further, the voice switch according to the second embodiment described above is connected to the receiving side speech circuit through the protection side sound circuit while controlling the attenuation amount according to the reverberation characteristic or the output of the reverberation monitoring filter. It is possible to adopt a configuration in which an attenuation amount that eliminates the reverberation of the received voice that wraps around as an echo is inserted into the receiving side communication circuit.

Further, the voice switch of the above-mentioned second embodiment can be constructed so as to immediately remove the attenuation amount inserted in the receiving side communication circuit when the receiving side communication circuit detects the reception voice.

In the voice switches of the first and second modes, when the voice is detected by the receiving side speech circuit for a predetermined period immediately after the state transition from the receiving state to the transmitting state,
A configuration can be adopted in which the state immediately shifts to the receiving state regardless of whether or not the transmitting side communication circuit detects the transmitting voice.

Further, in the voice switches of the first and second modes, even when voice is detected by the receiving side call circuit in the transmitting state, the transmitting voice signal is transmitted by the transmitting side call circuit. While the call is continuously detected, the transmitting state can be maintained.

In the voice switches of the first and second modes, the transmission voice detected by the transmission side communication circuit is interrupted and the transmission state is changed from the transmission state to the reception state. If the transmitting side speech circuit detects the transmitted voice again after the lapse of a predetermined time which can be regarded as a temporary silent section, it is possible to immediately return to the transmitting state.

With such a configuration, if the interruption of the transmitted voice is a temporary silent section that occurs immediately before the explosive consonant or the like, it is possible to immediately return to the transmitted state when the transmitted voice resumes. There is no possibility that a voice transmission block will occur due to interruption of the voice received from the partner terminal during the silent period.

Further, the voice switches of the first and second forms described above are provided when the receiving voice detected by the receiving side communication circuit is interrupted within a predetermined time which can be regarded as a temporary silent section of the transmitting voice. Can be configured to return to the transmitting state after the echo circling around the microphone falls below the voice detection level in the transmitting side communication circuit.

With such a configuration, when the received voice is interrupted within the temporary silent section of the transmitted voice, there is a possibility that the echo is judged as the transmitted voice of the speaker and the voice switch malfunctions. It can be lost.

[0038]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is an example of the configuration of a hands-free call system. However, here, only the configurations of the voice switches on the near end side and the far end side are shown, and other parts of the call terminal are omitted. In the figure, 1 is a voice switch as an embodiment of the present invention, 2 is a microphone, and 3 is a speaker. The voice switch 1 includes a switch control circuit 10, attenuation circuits 20 and 21, and delay circuits 30 and 31.

The attenuator circuit 20 inserts the attenuation amount into the transmitting side speech circuit in the terminal, and the attenuating circuit 21 inserts the attenuation amount into the receiving side communication circuit in the terminal. The switch control circuit 10 controls the insertion attenuation amount of the attenuation circuits 20 and 21 based on the transmitted voice input from the microphone 2 and the received voice transmitted from the far-end side terminal. The delay circuits 30 and 31 delay the transmission voice and the reception voice by the same amount as the delay time of the signal generated by the switch control circuit 10.

FIG. 2 shows the configuration of the voice switch 1. The switch control circuit 10 includes a determination circuit 100, a transmission side voice detection circuit 110 (s), and a reception side voice detection circuit 110.
(R), and the reverberation monitoring filter 120.
The voice detection circuit 110 (s) monitors the transmitted voice input from the microphone 2, and if the transmitted voice is detected, S =
1. If the transmitted voice is not detected, the determination circuit 100 determines S = 0.
Notify. Similarly, the voice detection circuit 110 (r) monitors the received voice sent from the far-end side terminal, and if the received voice is detected, R = 1 is set, and if the received voice is not detected, R = 0 is set to the determination circuit 100. Notice. The reverberation monitoring filter 120
This is a low-pass filter that calculates and outputs the power of the echo of the received voice that wraps around the microphone 2. Judgment circuit 1
00 controls the insertion attenuation amount of the attenuation circuits 20 and 21 based on the notifications from the voice detection circuits 110 (s) and 110 (r) and the output of the reverberation monitoring filter 120.

FIG. 3 shows a configuration example of the voice detection circuit 110. In the figure, 111 is a voice level detection filter, 112 is a noise level detection filter, 113 is a comparator, and 114 is a multiplier whose multiplication coefficient is b. Voice level detection filter 1
Reference numeral 11 is a low-pass filter that calculates and outputs the level of voice input from the microphone 2, and noise level detection filter 112 is a low-pass filter that calculates and outputs the noise level of ambient noise picked up by the microphone 2. .

In the voice detection circuit 110, the voice detection threshold is determined from the ambient noise level so that noise picked up by the microphone is not mistaken for voice, and the presence or absence of voice detection is determined based on this value. That is, the comparator 113 compares the input voice level (output P of the voice level detection filter 111) with the ambient noise level (output N of the noise level detection filter × multiplication coefficient b of the multiplier 114). Comparator 11
If P ≧ bN, 3 is determined to be “with voice” and is “1”.
Is output, and if P <bN, it is determined that “no sound” and “0” is output. Here, b is a margin for preventing noise from being mistaken for voice.

FIG. 4 shows a configuration example of the audio level detection filter 111. In the figure, 1111 is a short time integrator, 1112
Is a comparator, 1113 is a switch, 1114 is a shift register for delaying a signal for a certain time, 1115 is a multiplier with a multiplication coefficient of 1-α, 1116 is a multiplier with a multiplication coefficient of α, 1
117 is an adder. The short-time integrator 1111 is for suppressing the amplitude fluctuation of the input signal, and outputs a value X obtained by accumulating the amplitude value (absolute value or square value) of the input signal for several sampling periods. The integration time required to suppress the amplitude fluctuation to a practical level is approximately 1 to 2 ms.

The comparator 1112 is a short time integrator 1111.
Output X of the shift register 1114 is compared, and if X ≧ Z, P = Z, and if X <Z, the switch 1113 is switched such that P = X. Control P.

The voice level detection filter 111 needs to have a sharp rising time constant capable of quickly detecting an input and a short falling time constant such that the output does not become silent for at least one pitch of the vowel even if the input is lost. . this is,
The delay time of the shift register 1114 and the multiplier 111
This is possible by selecting the multiplication coefficient of 5, 1116 to an appropriate value.

FIG. 5 shows a configuration example of the noise level detection filter 112. In the figure, 1121 is a short-time integrator, 1122
Is a comparator, 1123 is a switch, 1124 and 1125 are counters, 1116 is a shift register for delaying a signal for a predetermined time, and 1127 is an adder. The noise level detection filter 112 is configured as a low-pass filter having a long rise time constant and a short fall time constant, and has a characteristic that it does not easily rise even when a voice is input, and immediately returns to the noise level when the voice disappears. ing.

The noise level detection filter 112 operates as follows. The short time integrator 1121 is the same as the short time integrator 1111 of the audio level detection filter 111. The comparator 1122 outputs the output X of the short-time integrator 1121.
And the output M of the shift register 1126 are compared, and X ≧
If M, the count value of the counter 1124 is incremented by 1 and X <M
If so, the count value of the counter 1125 is controlled to be incremented by one. Then, β is selected when the counter 1124 reaches a predetermined value, -β when the counter 1125 reaches the predetermined value, and 0 is selected when neither of the counters 1124 and 1125 reaches the predetermined value. By controlling the switch 1123, the selected value is passed through the adder 1127 to M
Is added to. In order to delay the rise and the fall of the noise level detection file 112, a large predetermined value may be set for the counter 1124 and a small predetermined value may be set for the counter 1125.

FIG. 6 shows a configuration example of the reverberation monitoring filter 120. In the figure, 121 is a short time integrator, 122 is a comparator,
Reference numerals 123 and 128 are switches, 124 is a shift register for delaying a signal for a predetermined time, 125 is a multiplier having a multiplication coefficient of 1-α, 126 is a multiplier having a multiplication coefficient of α, and 127 is an adder. The reverberation monitoring filter 120 has basically the same configuration as the voice level detection filter 111, but differs in the following points. The reverberation monitoring filter 120 controls the switches 128 and 123 using the judgment value R of the reception voice detection by the voice detection circuit 110 (r). That is, the switch 128 is fixed to the contact b side when R = 0 (no received voice), and fixed to the contact a side when R = 1 (with received voice). Further, the switch 123 is fixed to the contact d side regardless of the output of the comparator 122 when R = 0,
When R = 1, it is controlled by the output of the comparator 122.
Therefore, when R = 1, the operation is exactly the same as that of the audio level detection filter 111.

When the received voice is interrupted and becomes silent and changes from R = 1 to R = 0, the switch 128 is switched to the contact b side. Therefore, the reverberation monitoring filter 120 detects the noise level of the voice detection circuit 110 (s). Output N of filter 112
s is input, and the switch 123 is fixed to the contact d side. Therefore, the output P'of the reverberation monitoring filter 120 after this switching falls to the noise level Ns with the next attenuation characteristic. P ′ = P′α + Ns (1-α) If the near-end speaker is not speaking, immediately before this switching,
It is considered that the echo following the ending of the suspended received voice was input to the reverberation monitoring filter 120. Naturally, this echo gradually decreases, but if the multiplication coefficient α is set so that the attenuation characteristic expressed by the above equation and the indoor echo attenuation characteristic are the same, The change in the output P'simulates a decreasing echo level change.

If the output P'of the reverberation monitoring filter 120 at this time is used, the following control becomes possible. That is, when the received voice is interrupted and becomes silent, and R = 1 is changed to R = 0, the determination circuit 100 determines that the attenuation circuit 20 on the transmission side.
By reducing the insertion attenuation amount of the signal in accordance with the attenuation of the output P ′ of the reverberation monitoring filter 120, the echo sneaking into the microphone 2 can be suppressed to a constant level.

The effect of such echo suppression control will be described with reference to FIGS. 7 and 8. In each figure, (a) changes in the received voice on the near end side, (b) changes in the transmitted voice on the near end side, and (c) changes in the received voice on the far end side are shown on the same time axis. . FIG. 7 shows a case where the near-end speaker does not speak.
During the utterance of the far-end speaker, the echo of the received voice that wraps around the microphone 2 is detected as the transmitted voice on the near-end side. The reverberation monitoring filter 120 outputs the smoothed variation of the transmitted voice. The determination circuit 100 increases or decreases the insertion attenuation amount of the attenuation circuit 20 according to the change of the output P ′ of the reverberation monitoring filter 120. As a result, the received voice on the far end side is kept constant at a level slightly below the voice detection threshold (ambient noise level). At time A in the figure, the received voice sent from the far end side is interrupted and becomes silent, and then an echo that gradually decreases is input to the microphone 2. This echo attenuation is simulated by the output P ′ of the reverberation monitoring filter 120, and the decision circuit 100 reduces the insertion attenuation amount of the attenuation circuit 20 accordingly. Therefore, the received voice on the far end side is kept at a constant level below the voice detection threshold.

FIG. 8 shows a case where the near-end speaker starts speaking at time B after the reception of the received voice on the near-end side. The transmitted voice on the near-end side after the point of time B is the echo of the near-end speaker superimposed on the gradually decreasing echo. Since the insertion attenuation amount of the attenuation circuit 20 is such that the echo is slightly below the voice detection threshold value, the voice of the near-end speaker superimposed on the echo is not attenuated below the voice detection threshold value and is detected as a received voice at the far-end side. To be done. As a result, the sound of the near-end speaker is output from the speaker on the far-end side above the ambient noise level and is audible to the far-end speaker, but since the echo is output below the ambient noise level, howling and strong noise occur. There is no reverberation.

Since the signal delay occurring in the switch control circuit 10 is due to the short-time integration performed in the voice level detection filter 111, the delay circuits 30 and 31 cause the same amount of transmitted voice as the integration interval (about 1 to 2 m). By delaying the received voice, it is possible to prevent the beginning of the voice from being cut off.

Next, switching control of the voice switch 1 of this embodiment will be described. The switching control of the voice switch 1 is basically the same as that of the conventional voice switch shown in FIG. That is, the call direction is determined based on the presence or absence of voice detection, the switching rule based on the present priority is applied, and the receiving / transmitting blocking is avoided by providing a transitional state in which the present is not prioritized. And

FIG. 9 shows a state transition diagram of the voice switch 1. In the figure, S = 0/1 and R = 0/1 represent the presence / absence of detection of transmitted voice / received voice. The transition condition between each state is basically the same as in the case of the conventional example of FIG. However, the duration of the receive-priority / sending-priority state is determined by the constants T _R , T _S , and T _T determined by taking into account the length of the silent period before the onset of the consonant or the consonant and the reverberation time of the room. It is prescribed.

If the received voice is interrupted and becomes silent when the near-end side is in the receiving state, only the echo of the received voice remains and R
= 0 and S = 1, the state shifts to the receiving priority state (the communication direction is the same as the transmitting state). As described above, since the control for suppressing the echo of the received voice is performed when the state shifts to the receiving priority state, S is used within the time T _R corresponding to the reverberation time of the room.
When R = 1, the receiving state is restored regardless of the value of. If S = 1 is detected after the lapse of time T _{R, the} state transits to the transmitting state.

Further, when the near-end side is in the transmitting state, even if a silence period occurs in the voice of the near-end speaker, the far-end side performs control to suppress the echo of the received voice, so that the echo is near. It never returns to the edge. Therefore, when the received voice from the far end side is detected during this silent period, it can be considered that it is due to the utterance of the far end speaker. The following three cases can be considered for the silent period that occurs when the speaker's voice is interrupted. (A) 100 ms that occurs before the onset of explosive consonants and fricatives
~ 200 ms silent period (b) Short breathing period of about 500 ms corresponding to punctuation (c) Long silent period following one sentence corresponding to punctuation

When the silent period due to the interruption of the transmitted voice of the near-end speaker is (a), the far-end speaker rarely speaks during that period, and even if there is, it is light enough to be a coward. Probability is high. Further, it may be considered that the noise is not the voice of the far-end speaker but the noise at the far-end side (door open / close sound, etc.).

In this way, when the near-end speaker's voice has a silent period during the transmitting state, and the received voice from the far-end side is detected during that period: -In the transmitting state, based on the idea of receiving priority. Even if there is a received voice, it immediately shifts to the receiving state and switches the call direction. Or, based on the idea of the present priority, the received voice is ignored and the transmitting state is maintained. The control is considered. In this embodiment, these ideas are eclectic and the following transmission priority state is provided.

If the received voice is detected when the transmitted voice is interrupted in the transmitting state, the state shifts to the transmitting priority state (the communication direction is the same as the receiving state). If after that if is detected the received voice is continuously, the (a) within the time T _T corresponding to silence section maintains the transmit preference, regardless of whether one sending condition, the time T _T is elapsed Even if the received voice is detected, the state changes to the receiving state. Further, when the received voice is interrupted and becomes silent within the time T _T , the echo of the received voice is attenuated until it is attenuated to the ambient noise level, that is, the reverberation time T _S is the same as in the receive priority state. Performs echo attenuation control. When the transmitted voice is detected after the reverberation time T _S has elapsed, the state returns to the transmitted state.

Next, with reference to FIG. 1, description will be given of the amount of attenuation to be inserted in the closed loop of the hands-free communication system.
The purpose of howling prevention is achieved by selecting an insertion attenuation amount equal to the gain of the closed loop, but this attenuation amount is not sufficient from the viewpoint of suppressing reverberation. That is, in this case, the sound produced by the near-end speaker with the volume of SdB returns to the position of the microphone 2 on the near-end side at the volume of SdB and returns to the position of the microphone 2 on the near-end side at the volume of SdB. Will feel a strong reverberation. Therefore, in this embodiment, it is assumed that the amount of attenuation that suppresses the echo to be equal to or lower than the ambient noise level is the lower limit of the insertion attenuation amount, and the insertion attenuation amounts of the attenuation circuits 20 and 21 are controlled according to the call state of the voice switch 1. ing.

The lower limit of the insertion attenuation amount is the following four types. When the received voice is silent (only the ambient noise on the far end side) regardless of the call state, the insertion attenuation amount A _R of the attenuation circuit 21 on the reception side is L _N equal to the acoustic coupling gain on the near end side. As a result, the possibility that howling will occur due to the ambient noise sent from the far end side can be suppressed, and if there is a voice received from the far end speaker, it can be output from the speaker 3. When the reception voice is detected in the transmitting state, the insertion attenuation amount A _R of the attenuation circuit 21 on the reception side is the attenuation amount L _R at which the reception voice leaking from the speaker 3 is equal to or lower than the ambient noise level at the position of the microphone 2, that is, _Let L _R = N _S / (P _R K). Here, N _S is the ambient noise level, P _R is the output of the voice level detection filter 111 of the voice detection circuit 110 (r), and K is a constant (for example, 1.8). This allows
The received voice output from the speaker 3 on the near end side is suppressed to the ambient noise level, the gain of the closed loop becomes less than 1, and howling is prevented.

The insertion attenuation amount A _S of the attenuation circuit 20 on the transmitting side in the reception state is the attenuation amount L _E that makes the echo of the reception voice sneaking into the microphone 2 equal to or lower than the ambient noise level, that is, L _E = N and _S / (P _E K). Here, P _E is the output of the reverberation monitoring filter 120 (= continuous echo level change). As a result, the received voice is not detected on the far end side, and the attenuation circuit 21 maintains the state in which the attenuation amount L _N is inserted. Therefore, the echo sneaking into the microphone 2 on the near end side is L _E L _N
Is output from the speaker 3'on the far end side.
Its size is equal to the ambient noise level on the far end side, and there is little deterioration in call quality.

When the receiving voice is interrupted in the receiving state to be silent and only the echo remains, the attenuation circuit 20 on the transmitting side is provided.
The insertion attenuation amount A _S is set to an attenuation amount L _V that is controlled so as to be gradually reduced in accordance with the attenuation of the echo, that is, L _V = N _S / (P _V K). Here, P _V is the output of the reverberation monitoring filter 120 (= changing level of echoes). If there is no utterance from the near-end speaker during this period, the received voice is not detected on the far-end side as in the case. Further, if the near-end speaker speaks during this period, the received voice is detected on the far-end side, and the echo and the near-end speaker's voice superimposed on it are output from the speaker 3 ', but the echo is ambient noise. Since it is attenuated to the level, only the voice of the near-end speaker can be heard. In addition, except for the above four cases, the insertion attenuation amounts A _R and A _S are set to 1 and the received / transmitted voice is not attenuated.

Next, switching control by the switch control unit 10 will be described with reference to the flowcharts of FIGS.
FIG. 10 is a flowchart of a process of detecting a voice signal by the voice detection circuit 110 shown in FIG. That is, in the voice detection circuit 110, the comparison unit 113 causes the voice level P (output of the voice level detection filter 111) and the ambient noise level D (output N of the noise level detection filter 112).
X multiplication coefficient b of the multiplier 114 is compared (step S
1), if P> D, it is determined that there is an audio signal and "1"
Is output (step S2), and if P ≦ D, it is determined that there is no audio signal and “0” is output (step S2).
3).

The voice detection circuit 110 constantly monitors the voice signal by repeatedly executing this processing. However,
10 (a) is a process receiving side of the voice detection circuit 110 (r) outputs a determination value R to determine the presence or absence of received voice P _R, FIG. 10 (b) transmitting side speech detection circuit 110
(S) is a process of outputting a decision value S to determine the presence or absence of the reception voice P _S.

FIG. 11 and FIG. 12 are flowcharts of the processing for controlling the call state transition and the insertion attenuation amount in the determination circuit 100. The determination circuit 100 constantly monitors the call state by repeatedly executing this process. The symbols used in this flow chart have the following meanings. W: Variable representing the call state with the following value R ₀ = Reception state S ₀ = Transmission state R _T = Reception priority state S _T = Transmission priority state R ・ ・ ・ Judgment value indicating presence / absence of reception voice 0 = No received voice 1 = Received voice S ... Judgment value indicating presence or absence of transmitted voice 0 = No transmitted voice 1 = Presented voice A _S ... Insertion attenuation amount by the attenuation circuit 20 on the transmitting side A _r: Insertion attenuation amount by the reception side attenuation circuit 21 N _S: Ambient noise level P _E : Output of the reverberation monitoring filter 120 P _V: Echo of attenuation that occurs after interruption of reception voice level L _R · · · reception voice equal attenuation to the attenuation amount L _N · · · acoustic coupling gain decays to the ambient noise level _{T R, T S, T T} ··· reception / transmission priority state duration constant defining the _{C R, C S, C T} ··· T R, T S, and measures an elapsed time corresponding to T _T Counter of

This processing will be described in the order of the numbers shown in the state transition diagram of FIG. (1) Priority on the present condition of the receiving state If there is a receiving voice, step S10 → step S17 →
Since the insertion attenuation amount A _S = N _s / (P _E K) on the transmitting side is passed through the processing path of step S18, the echo of the received voice is suppressed to the ambient noise level or lower. Also, the received voice,
If there is no transmitted voice, step S10 → step S
17 → step S19 → step S20 → step S2
Insertion attenuation amount A _R = L _N on the receiving side through the processing path of 1,
Since the insertion attenuation amount A _S on the transmitting side is 1, the voice uttered by the near-end speaker is sent to the far-end side without being attenuated, and howling is generated by ambient noise sent from the far-end side. Possibility is suppressed.

(2) Transition from the receiving state to the receiving priority state When the receiving state (W = R ₀ ) occurs, the state of no receiving voice (R = 0) and with transmitting voice (S = 1) occurs, By going through the processing route of step S10 → step S17 → step S19 → step S20 → step S22, the state transits to the receiving priority state (W = _RT ), and further step S2
3 → step S25 → step S26 → step S18
Insertion attenuation amount on the transmitting side A _S = N _s /
(P _V K). Since the attenuation amount A _{S in} this case decreases in accordance with the attenuation of the echo, only the echo is suppressed below the ambient noise level, and the transmitted voice superimposed on the echo is detected at the far end side.

(3), (4) Transition from the receiving priority state After the transition to the receiving priority state, the period until the echo of the received voice is attenuated and reaches the ambient noise level (= reverberation time T _R ).
The current state is maintained regardless of the presence or absence of transmitted voice, and the time is T
After the elapse of _R , it is necessary to switch to another state depending on the presence / absence of the received / transmitted voice. Whether or not the time T _R has elapsed is determined by comparing the constant T _R with the counter value C _R. The following control is performed based on this determination.

After the transition to the receiving priority state, the above steps S10 to S18 are repeated, and the counter C _R is incremented by one each time, and it is determined in step 26 that C _R > T _R. Then, the process branches to step S27 to check the presence / absence of transmitted voice. As a result, if there is no transmitted voice (S = 0), the current state is maintained (step S1).
8) Further, if there is a transmitted voice (S = 1), the state changes to the transmitted state (W = S ₀ ), and the insertion attenuation amount A _S on the transmitting side becomes 1 (step S39). Is sent to the far end without being attenuated. Further, when the presence of the received voice (R = 1) is detected in step S24 before C _R > T _R , the receiving state (W = W) is obtained via the processing route of step S24 → step S32 → step S33 → step S18. R ₀ ), and the insertion attenuation amount AR on the receiving side becomes A _R = 1 and the insertion attenuation amount on the transmitting side becomes A _S = N _s / (P _E K). Therefore, the reception voice is not attenuated and output from the speaker. Thus, the echo of the received voice is suppressed below the ambient noise level.

(5) Priority of the present condition in the transmitting state If there is no received voice, step S10 → step S11
→ Step S12 → Through the processing route of Step S15,
Inserting attenuation A _R = L _N next to the receiving side, a possibility that howling is generated by ambient noise sent from the far-end side is suppressed. If both the received voice and the transmitted voice are present, step S
10 → step S11 → step S12 → step S1
The insertion attenuation amount A _R = L _R on the receiving side is passed through the processing path of 3 → step S16, and the receiving voice from the far end side is attenuated to the ambient noise level or lower.

(6) Transition from the transmitting state to the transmitting priority state In the transmitting state (W = S ₀ ), there is no transmitting voice (S = 0) and receiving voice (R = 1). When occurring, step S10 → step S11 → step S12 → through the processing path, step S13 → step S14, transition to transmit preference state (W = S _T), further step S34 → step S29 → step S30 → step S31 → The insertion attenuation amount A _S = N _s / (P _E K) on the transmitting side and the insertion attenuation amount A _R = 1 on the receiving side are obtained through the processing path from step S33 to step S18. As a result, the voice received from the far end side is output from the speaker without being attenuated, and the echo of the voice received is suppressed below the ambient noise level. (7), (8) after the transition to the transition transmit preference state from transmit preference state, until the time T _T to be regarded as silent period by explosion consonants like has elapsed to maintain the status quo, after a time T _T It is necessary to switch to the receiving state when the receiving voice is detected. Whether or not the time T _T has elapsed is determined by comparing the constant T _T with the counter value C _T. Further, if the received voice is interrupted before the time T _T has elapsed, the time (= reverberation time T _S ) until the echo of the received voice is attenuated to reach the ambient noise level is irrespective of the presence or absence of the transmitted voice. It is necessary to perform the echo suppression control while maintaining the transmission priority state, and switch to the transmission state when the transmission voice is detected after the time T _S has elapsed. The determination as to whether or not the reverberation time has elapsed is made by comparing the constant T _S and the counter value C _S. The following control is performed based on this determination.

After the transition to the transmission priority state, the reception sound is continuously received.
While voice (R = 1) is detected, step S1
0 → step S11 → step S28 → step S29
→ step S30 → step S31 → step S33 →
The process of step S18 is repeated, and each time
TA C_TIs increasing. Then, in step S31, C _T
> T_TIf it is determined that step S31 → step S3
2 → step S33 → step S18
And the receiving state (W = R₀). At this time the transmitting side
Insertion attenuation A_SAnd the amount of insertion attenuation A on the receiving side_RIs the current situation
Carry.

Also, before C _T > T _T , step S
When the interruption (R = 0) of the received voice is detected in 29, the insertion attenuation amount A _S = on the transmitting side is passed through the processing route of step S35 → step S36 → step S40 → step S18.
_{N s / (P V K)} , the insertion attenuation of the reception side A _R = L _N. As a result, the attenuation amount A _S decreases in accordance with the attenuation of the echo, so that only the echo is suppressed below the ambient noise level, and the transmitted voice superimposed on the echo is detected on the far end side. Further, the possibility of howling occurring due to ambient noise sent from the far end side is suppressed.

Then, unless the presence of the received voice (R = 1) is detected again in step S29, the above step S1 is performed.
The processing from 0 to step S36 to step S18 is repeated, and the counter C _S is incremented each time, and when C _S > T _S is determined in step S36, the process branches to step S37 and the presence or absence of the transmitted voice. Is checked. as a result,
If there is no transmission voice (S = 0), the current state is maintained, and if there is transmission voice (S = 1), the transmission state (W = S
₀ ) and the insertion attenuation amount A _S on the transmitting side becomes 1 (step S39). Therefore, the transmitted voice is not attenuated and is transmitted to the far end side.

The configuration of the voice switch described above is based on the premise that the call terminal equipped with the voice switch is connected to the 4-wire type transmission line and the call terminal on the opposite side is also equipped with the voice switch of the same configuration. I am trying. However, when the call terminal is connected to a two-wire transmission line, the transmitted voice sent from the near-end side to the far-end side circulates around the protection side sound circuit on the transmission line and echoes to the near-end side. May come back as. Also,
Even when the far-end side terminal uses the voice switch having the conventional configuration, the transmitted voice sent from the near-end side may return to the near-end side as an echo through the acoustic coupling on the far-end side.
In order to be able to handle these cases, it is necessary to perform echo attenuation control on the transmitting side in the voice switch on the receiving side as well.

FIG. 13 shows an example of the structure of a voice switch that can handle such a case. In this configuration, the voice switch 1 of the embodiment shown in FIG. 2 is further provided with a reverberation monitoring filter 120 (r) on the receiving side. When the decision circuit 100 detects the received voice immediately after the transmitted voice is interrupted and becomes silent, it is determined that it is an echo sent back from the far end side or a side tone that wraps around the protection side tone circuit. Judgment is made, and the insertion attenuation amount of the attenuation circuit 21 on the receiving side is controlled so as to suppress the echo or stimulus sound to a constant level below the ambient noise level during the period corresponding to the reverberation time. The control content in this case is the same as that performed on the transmitting side, and therefore its description is omitted.

Further, in the voice switch 1 having the configuration of FIG. 13, while the attenuation amount of the attenuation circuit 20 on the transmitting side is gradually reduced according to the echo attenuation characteristic, the attenuation circuit 21 on the receiving side is inserted. By increasing the attenuation amount in accordance with the decrease, it is possible to perform control such that the attenuation amount to be inserted into the closed circuit is kept above a certain value.

[0080]

As described above, according to the present invention, in a call terminal of a hands-free call system, howling and reverberation are suppressed, and the call direction is changed at high speed without causing reception blocking or transmission blocking. It is possible to switch.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration example of a hands-free call system to which the present invention is applied.

FIG. 2 is a diagram showing a configuration of a voice switch (corresponding to a 4-wire type transmission line) as an embodiment of the present invention.

FIG. 3 is a diagram showing a configuration of a voice detection circuit in the voice switch of the embodiment.

FIG. 4 is a diagram showing a configuration example of an audio level detection filter in an audio detection circuit.

FIG. 5 is a diagram showing a configuration example of a noise level detection filter in a voice detection circuit.

FIG. 6 is a diagram showing a configuration example of a reverberation monitoring filter in the voice switch of the embodiment.

FIG. 7 is a diagram for explaining the effect of echo suppression control (when there is no near-end speaker's voice).

FIG. 8 is a diagram for explaining the effect of echo suppression control (when there is a voice of a near-end speaker).

FIG. 9 is a state transition diagram of the voice switch of the embodiment.

FIG. 10 is a flowchart showing a voice detection process in the voice detection circuit.

FIG. 11 is a flowchart (part 1) of control by the determination circuit.

FIG. 12 is a flowchart (part 2) of control by the determination circuit.

FIG. 13 is a diagram showing a configuration of a voice switch (corresponding to a two-wire transmission line) as another embodiment of the present invention.

FIG. 14 is a diagram showing a closed loop constituted by a hands-free call system.

FIG. 15 is a diagram showing a configuration example (1) of a conventional voice switch.

FIG. 16 is a diagram showing a configuration example (2) of a conventional voice switch.

FIG. 17 is a state transition diagram of a conventional voice switch (No. 2).

[Explanation of symbols]

 1 Voice Switch 2 Microphone 3 Speaker 10 Switch Control Circuit 100 Judgment Circuit 111 Voice Level Detection Filter 1111 Short Time Integrator 1112 Comparator 1113 Switch 1114 Delay Shift Register 1115, 1116 Multiplier 1117 Adder 112 Noise Level Detection Filter 1121 Short Time Integrator 1122 Comparator 1123, 1124 Counter 1123 Switch 1126 Delay shift register 1127 Adder 113 Comparator 114 Multiplier 110 Voice detection circuit 120 Reverberation monitoring filter 121 Short time integrator 122 Comparator 123, 128 Switch 124 Delay shift Registers 125 and 126 Multiplier 127 Adder 20 Attenuation circuit on the transmitting side 21 Attenuation circuit on the transmitting side 30 Delay circuit on the transmitting side 31 Delay circuit on the transmitting side

Claims (11)

[Claims] Claims: 1. A transmitting side communication circuit and a receiving side communication circuit are respectively provided with attenuation means for inserting an attenuation amount into the communication circuit. In the transmitting state, a large amount of attenuation is inserted into the receiving side communication circuit, and in the receiving state, transmission is performed. The voice switch used for the hands-free function that suppresses howling by inserting a large amount of attenuation in the talk side call circuit and enables a call using a microphone and a speaker. At the time of the state transition from the state to the transmitting state, the amount of attenuation that the received voice sent to the partner terminal side is below the voice detection level at the partner terminal side is temporarily inserted into the transmitter side call circuit. A voice switch characterized in that 2. An attenuator for inserting an attenuation amount into a speech circuit is provided in each of the speech circuit on the transmitting side and the speech circuit on the receiving side. By inserting a large amount of attenuation in the talk side call circuit, the howling is suppressed, and in the voice switch used for the hands-free function that enables a call with a microphone and a speaker, a hands-free call system with a protection side sound circuit It is used by the received voice that has circulated as an echo to the microphone to temporarily change itself from the received state sent from the receiving side to the transmitting side via the protection side sound circuit of the received voice sent to the other terminal side. The component that has spilled into the receiving side communication circuit is inserted into the transmitting side communication circuit with an attenuation amount that is less than the voice detection level in the receiving side communication circuit. Voice switch, characterized in that it was. 3. The method according to claim 1, wherein the period of inserting the attenuation amount into the transmitting side communication circuit is set until the echo sneaking into the microphone falls below the voice detection level in the transmitting side communication circuit. Voice switch. 4. An attenuation amount to be inserted into the transmitter side speech circuit is gradually reduced in accordance with the reverberation characteristic.
The voice switch according to any one of 3 to 3. 5. A reverberation monitoring filter for observing a power change of an echo that wraps around a microphone is provided, and according to a power change of the echo output by the reverberation monitoring filter,
The voice switch according to any one of claims 1 to 3, wherein an attenuation amount to be inserted into the transmitter side speech circuit is controlled. 6. A size that eliminates the reverberation of the received voice that has circulated as the echo into the receiving side call circuit through the protection side sound circuit while the attenuation amount control according to claim 4 or 5 is being performed. 3. The voice switch according to claim 2, wherein the attenuation amount is inserted into the communication circuit on the receiving side. 7. The voice switch according to claim 6, wherein when the received voice is detected by the receiving side call circuit, the attenuation amount inserted in the receiving side call circuit is immediately removed. 8. When a voice is detected in the receiving side call circuit for a predetermined period immediately after the state transition from the receiving state to the transmitting state,
The voice switch according to any one of claims 1 to 7, which immediately shifts to a receiving state regardless of whether or not a transmitting voice is detected by the transmitting side communication circuit. 9. A transmission state while the transmission side speech circuit continuously detects the voice even if the speech is detected by the reception side speech circuit in the transmission state. The voice switch according to any one of claims 1 to 7, wherein the voice switch is maintained. 10. When a predetermined time, which can be regarded as a temporary silent section of the transmitted voice, elapses after the transmitted voice detected by the transmitting side communication circuit is interrupted and transitions from the transmitted state to the received state. If the transmitting side communication circuit detects the transmitted voice again,
10. The voice switch according to claim 1, wherein the voice switch is immediately returned to the transmitting state. 11. When the received voice detected by the receiving side call circuit is interrupted within a predetermined time which can be regarded as a temporary silent section of the transmitting voice, an echo circling around the microphone is generated in the calling side call circuit. 11. The voice switch according to claim 10, wherein the voice switch is returned to the transmitting state when the voice detection level is lower than the voice detection level.