WO2010122398A1

WO2010122398A1 - Speakerphone apparatus

Info

Publication number: WO2010122398A1
Application number: PCT/IB2010/000876
Authority: WO
Inventors: Keiichi Yoshida; Minoru Fukushima; Tomohide Furuya; Yukio Okada; Teppei Washi; Yoshihiro Tanigawa; Teppel Washi
Original assignee: Panasonic Electric Works Co., Ltd.
Priority date: 2009-04-21
Filing date: 2010-04-20
Publication date: 2010-10-28
Also published as: WO2010122398A8

Abstract

A speakerphone apparatus includes a microphone and a speaker, a voice switch; an echo canceller, a doubletalk detector, and an echo suppressor. The voice switch includes a transmission side loss insertion unit, a reception side loss insertion unit, and an insertion loss amount control unit, and the insertion loss amount control unit includes a total loss amount calculator and an insertion loss amount allocation processor. Further, the total loss amount calculator has an update mode in which a total sum of the loss amounts inserted in the respective paths is calculated and adaptively updated based on estimations of each feedback gain, and a fixed mode in which a total loss amount is fixed to a initial value, and operates in the fixed mode while the echo canceller does not sufficiently converge after starting communications and in the update mode after the echo canceller fully converges.

Description

SPEAKERPHONE APPARATUS

Field of the Invention

The present invention " relates to a speakerphone apparatus for use in houses or offices.

Background of the Invention

Conventionally, there is known a speakerphone apparatus which realizes handsfree communications without a handset device upon communicating, by magnifying a voice signal by using a speaker and outputting it to a caller far distanced from a communications terminal, and collecting a voice of the caller by using a microphone and transmitting it to a communications terminal of his/her counterpart. Herein, the speaker magnifies the voice signal transmitted from the communications terminal of the counterpart and the microphone collects the voice emitted from the caller.

In such a speakerphone apparatus, an acoustic feedback path may be established by an acoustic coupling between the microphone and the speaker, through which an unpleasant echo (acoustic echo) may be heard. Further, if a closed loop is formed via the feedback path at a frequency component in the communications system and a gain of the loop exceeds 1, there may occur howling at the frequency. Therefore, there is provided an echo canceller and a voice switch in order to prevent the unpleasant echo and howling from occurring.

The voice switch constantly evaluates communications status (transmission state, reception state) and provides a loss into transmission and reception signal paths by properly allocating it based on a result of the evaluation. The echo canceller includes an adaptive filter for identifying an impulse response in the feedback path and estimating suspected echo components in the feedback path from a signal inputted thereto, and a subtracter for subtracting the estimated echo components from the signal inputted thereto through the feedback path. Generally, the adaptive filter needs several seconds of learning time to identify an impulse response of the feedback path. Accordingly, the echo cannot be effectively suppressed for several seconds immediately after starting communications. However, since a closed loop has been established in the communications system, an unpleasant echo or a howling may occur therein. In this regard, there has been proposed a speakerphone apparatus capable of suppressing occurrences of the unpleasant echo or the howling immediately after starting the communications (see, e.g., Japanese Patent Application Publication No. 2002-359580) . In this conventional example, while the echo canceller does not converge right after starting communications, the voice switch operates in a fixed mode in which a total amount of losses (total loss amount) provided in a signal path is fixed at a sufficiently large initial value, thereby suppressing the unpleasant echo or the howling. On the other hand, after the echo canceller fully converges, the voice switch operates in an update mode in which the total loss amount is updated as occasion demands and, accordingly, full duplex communications can be realized.

Meanwhile, if digital communications, e.g., internet protocol (IP) communications or the like, are performed among speakerphone apparatuses, a delay may occur in transmission of voice data due to a packet process or a communications buffer process for jitter suppression. Generally, it is known that, when the delay occurs in the voice data transmission between the speakerphone apparatuses, as the delay time becomes longer, the echo tolerance of a speaker upon communicating gets Worse and, accordingly, even a small echo can make the speaker feel unpleasant.

Such unpleasant echo can be prevented by setting a greater initial value in the fixed mode than that of the conventional analog communications. However, degradation amount in the echo tolerance may exceed several tens decibels, which cannot be covered by a suppression amount of the echo canceller in the update mode and, as a result, the communications are performed in half duplex communications rather than full duplex communications. Furthermore, when the ambient noise level is High, there may occur a problem in which the voice switch continues to switch on only one side of the transmission side and reception side and, accordingly, a voice from the one side is not transmitted to the other side.

On the other hand, when the digital communications, e.g., IP communications or the like are performed between a speakerphone apparatus and a communications apparatus using a handset, a residual, echo occurring in the transmission signal path of the speakerphone apparatus can be attenuated by an echo suppressor equipped therein, thereby realizing full duplex communications. However, if a high level of noise exists in a signal transmitted from the speakerphone apparatus to the counterpart communications apparatus , the echo suppressor attenuates not only the residual echo but momentarily the noise. . Such momentary attenuation of the noise is delayed and perceived by the counterpart through the handset. Therefore, the counterpart using the handset may feel an unpleasant sense of disconnection during communication.

In addition, if the counterpart uses a handset, it is possible to prevent the above mentioned sense of disconnection by setting an attenuation amount of the echo suppressor smaller than that when the counterpart uses a speakerphone apparatus. In this^' case, however, there is a drawback that the residual echo is heard through the handset. Summary of the Invention

In view of the above, the present invention provides a speakerphone apparatus capable of realizing full duplex communications even if a transmission delay occurs in a voice transmitted to a counterpart, and preventing a sense of disconnection from being felt through a handset even when the counterpart uses the handset. In accordance with an aspect of the present invention, there is provided a speakerphone apparatus including a microphone and a speaker; a voice switch adapted to switch over a communications status between transmission and reception by providing a loss in a reception side signal path through which a receiving signal from a communications terminal of the counterpart is transmitted to the speaker and a transmission signal path through which a transmitting signal collected by the microphone is transmitted to the communications terminal of the counterpart; an echo canceller for suppressing acpustic echo caused by acoustic coupling between the microphone and the speaker; and a doubletalk detector for detecting a doubletalk.

The voice switch includes a transmission side loss insertion unit for providing a loss in the transmission signal path; a reception side loss insertion unit for providing a loss in the reception side signal path; and an insertion loss amount control unit for controlling respective amounts of loss provided in the transmission and reception side.

Further, the insertion loss amount control unit includes a total loss amount calculator which is adapted to estimate an acoustic feedback gain of a path fed back to an input terminal of the transmission side loss insertion unit from an output terminal of the reception side loss insertion unit via an acoustic echo path, and calculate a total sum of loss amount provided to the feedback path based on the estimation of the acoustic feedback gain; and a insertion loss amount allocation processor adapted to monitor the transmitting signal and the receiving signal to estimate a communications status, and, based on the estimated communications status and the total sum calculated by the total loss amount calculator, determine respective insertion loss amounts allocated to the transmission side loss insertion unit and the reception side loss insertion unit.

In addition, the total loss amount calculator has an update mode in which a total loss amount is adaptively updated by calculating the total loss amount provided in the feedback path based on respective estimations of the feedback gain, and a fixed mode in which a total loss amount is fixed to an initial value. Thus, the total loss amount calculator operates in the fixed mode while the echo canceller does not sufficiently converge after starting communications and in the update mode after the echo canceller fully converges .

The speakerphone apparatus further includes an echo suppressor adapted to attenuate a residual echo by providing a certain attenuation amount in the transmission signal path in conjunction with the voice switch and/or the doubletalk detector.

With the above configuration, even if a transmission delay occurs in a voice transmitted between the communications apparatuses, the echo suppressor effectively attenuates the residual echo occurring in the transmission signal path due to the transmission delay, thereby realizing a full duplex communications.

The echo suppressor may insert the certain attenuation amount into the transmission signal path if the insertion loss amount allocation processor estimates that the voice switch is in a reception state, and not if otherwise.

Therefore, it is possible to suppress an inflection occurring in the voice and attenuate only an unpleasant echo upon communicating. As a result, pleasant communications can be achieved.

The echo suppressor may not insert the attenuation amount in the transmission signal path while the doubletalk detector is detecting a doubletalk, and do if otherwise. Therefore, it is possible to suppress an occurrence of inflection in a voice due to 'an erroneous attenuation to the voice to be transmitted and attenuate only an unpleasant echo upon communicating. As a result, pleasant communications can be achieved.

The echo suppressor may insert the certain attenuation amount in the transmission signal path if the doubletalk detector does not detect a doubletalk and the insertion loss amount allocation processor estimates that the voice switch is in the reception state.

By doing so, it is possible to prevent that the echo suppressor does not provide the attenuation amount by mistake and an unpleasant residual echo occurs, and, accordingly, to accurately attenuate only such an unpleasant echo during communication.

Preferably,- while the echo canceller does not sufficiently converge after starting communications, the echo suppressor inserts the certain attenuation amount in the transmission signal path only if the insertion loss amount allocation processor estimates that the voice switch is in the reception state. Generally, the doubletalk detector cannot detect a doubletalk while the echo canceller does not sufficiently converge. Therefore, the echo suppressor inserts the certain attenuation amount in the transmission signal path regardless of the detection result from the doubletalk detector if the insertion loss amount allocation processor estimates that the voice switch is in the reception state,

_ O _ thereby attenuating the unpleasant echo while the echo canceller does not sufficiently converge after starting communications .

While the echo suppressor is inserting the certain attenuation amount in the transmission signal path, it may stop the insertion of the attenuation amount into the transmission signal path in a moment if _^ the insertion loss amount allocation processor estimates that the voice switch is not in the reception state or the doubletalk detector detects the doubletalk.

By doing so, it is possible to prevent an occurrence of inflection in the voice due to the erroneous attenuation of the transmission signal by the echo suppressor.

The echo canceller further may include an adaptive filter for adaptively identifying characteristics of an echo path, a subtracter for subtracting an output of the adaptive filter from a near-end signal, and the doubletalk detector

^■for detecting a doubletalk. 'Further, the echo canceller may not update a coefficient of the adaptive filter when the doubletalk detector detects a doubletalk.

With this configuration, it, is possible to achieve simplification of configuration and lowering of cost by using the doubletalk detector provided in the echo canceller.

Respective processes in the echo canceller, the voice switch, and the echo suppressor are preferably performed in that order in the transmission signal path. With this configuration, since the echo suppressor is disposed after the voice switch in the transmission signal path, it is possible to prevent an inflection from occurring in the voice switch even if . the echo suppressor attenuates the transmission signal by mistake, thereby achieving a pleasant communications environment .

The echo suppressor may adjust the attenuation amount based on an acoustic feedback gain estimated by the total loss amount calculator. With this configuration, even if a voice emitted from the near-end speaker is erroneously attenuated by the echo suppressor, it is possible to suppress occurrence of the voice inflection.

The echo suppressor may adjust the attenuation amount based on the signal level average value of the reception signal .

With this configuration,, even if a voice emitted from the near-end speaker is erroneously attenuated by the echo suppressor, it is possible to suppress occurrence of the voice inflection.

The echo suppressor may fix the attenuation amount inserted in the transmission signal path to a normal value when the signal level of the reception signal is less than a threshold, and adjust the attenuation amount in a range from an upper limit greater than the normal value to the normal value when the signal level of the reception signal is the threshold or more .

With this configuration, it is possible to suppress occurrence of the voice inflection.

The echo canceller may further include a signal level average (SLA) operation unit for calculating a signal level average value of the reception signal, an adaptive filter for adaptively identifying characteristics of the echo path, and a subtracter for subtracting an output of the adaptive filter from the near-end signal. Further, the echo suppressor may adjust the attenuation amount based on the signal level average value calculated by the signal level average operation unit.

This makes it possible to achieve simplification of configuration and lowering of cost. The echo suppressor may insert a certain attenuation amount in the transmission signal path only if the doubletalk detector does not detect a doubletalk and the insertion loss amount allocation processor estimates that the voice switch is in the reception state, and monotonously decrease the attenuation amount inserted in the transmission signal path when the insertion loss amount allocation processor estimates that, the voice switch is neither in the reception state nor in the transmission state, or when the insertion loss amount allocation processor estimates that the voice switch is in the ^' reception state and the doubletalk detector detects a doubletalk. With this configuration, even if a transmission delay occurs in a voice transmitted to the communications apparatus of the counterpart, the echo suppressor can effectively attenuate the residual echo occurring in the transmission signal path due to the transmission delay, thereby realizing full duplex communications. Besides, while the transmission signal is not transmitted to the communications apparatus using a handset from the speakerphone apparatus, the echo suppressor monotonously decreases the attenuation amount inserted in the transmission signal path. This makes it possible to prevent the speaker at the communications apparatus using a handset from feeling an unpleasant sense of disconnection because the ambient noise is not momentarily attenuated. The echo suppressor may decrease the attenuation amount to zero immediately when the communications status estimated by the insertion loss amount allocation processor becomes the transmission state while the attenuation amount is being monotonously decreased. By doing so, it is possible to prevent that a voice emitted from the near-end speaker (transmission voice) is erroneously attenuated by the echo suppressor and an inflection occurs in the voice that the counterpart speaker hears . While the attenuation amount inserted in the transmission signal path is being monotonously decreased, the echo suppressor may insert the certain attenuation amount in the transmission signal path again when the doubletalk detector does not detect a doubletalk and the communications status estimated by the insertion loss amount allocation processor is the reception state.

With this configuration, it is possible to prevent that the echo suppressor does not provide the attenuation amount by mistake and an unpleasant residual echo occurs, and to accurately attenuate only an unpleasant echo during communication.

The echo suppressor may decrease the attenuation amount inserted in the transmission signal path at a constant decrease ratio when the communications status estimated by the insertion loss amount allocation processor is neither the reception state nor the transmission state, or when the communications status is ,the reception state and the doubletalk detector detects a doubletalk.

In this case, e.g., when compared to a stepwise decrease, a time period (transition time period) which is needed for the attenuation amount to be decreased to zero can be relatively shortened.

The echo suppressor may insert a certain attenuation amount in the transmission signal path when the doubletalk detector does not detect a doubletalk and the communications status estimated by the insertion loss amount allocation processor is the reception state, estimate a level of an ambient noise superimposed in the transmission signal, and reduce the attenuation amount inserted in the transmission signal path as the ambient noise level becomes higher based on the estimated ambient noise level. With this configuration, even if a transmission delay- occurs in a voice transmitted to the communications apparatus of the counterpart, the echo suppressor can effectively attenuate the residual echo occurring in the transmission signal path due to the transmission delay, thereby realizing full duplex communications. Besides, since the attenuation amount inserted in the transmission signal path is reduced by the echo suppressor as the noise level becomes higher, it is possible to suppress a momentary attenuation of the noise, thereby making it difficult to feel a sense of disconnection. In this case, if the echo suppressor decreases the attenuation amount, attenuation effect of the residual echo is lowered. However, the residual echo is masked by the ambient noise and does not make a trouble in communications in a practical use. The echo suppressor may compare an echo amount estimated from the reception signal level with an attenuation amount depending on an ambient noise level superimposed in the transmission signal, and insert one having a relatively small magnitude¹ among the echo amount and the attenuation amount in the transmission signal path. This makes it possible to sufficiently attenuate the residual echo and, at the same time, to prevent a sense of disconnection from being felt.

When the ambient noise level exceeds a certain level, the echo suppressor may adjust the attenuation amount to a constant value based on the noise level.

With this configuration, it is possible to prevent the residual echo from not being attenuated when the ambient noise level is greatly high.

Brief Description of the Drawings,

The objects and features of the present invention will become apparent from the following description of preferred embodiments, given in conjunction with the accompanying drawings, in which:

Fig. 1 shows a block diagram of a first embodiment in accordance with the present invention;

Fig. 2 presents_, a flow chart for explaining an operation of a voice switch shown in Fig. 1; Fig. 3 represents a flow chart for explaining an operation of an echo suppressor shown in Fig. 1;

Fig. 4 illustrates a block diagram of a second embodiment in accordance with the present invention;

Fig. 5 is a flow chart for explaining an echo suppressor shown in Fig. 4;

Fig. 6 shows a block diagram of a third embodiment in accordance with the present invention;

Fig. 7 presents a flow chart for explaining an operation of an echo suppressor depicted in Fig. 6;

Fig. 8 is a flow chart for explaining an operation of the echo suppressor in a modified example of the third embodiment ;

Fig 9 shows a block .diagram of a fourth embodiment in accordance with the present invention;

Fig. 10 depicts a flow chart for explaining an operation of an echo suppressor in the fourth embodiment;

Fig. 11 presents a flow chart for explaining an operation of the echo suppressor in a fifth embodiment in accordance with the present invention;

Fig. 12 illustrates an attenuation amount of the echo suppressor with time in the fifth embodiment of the present invention;

Fig. 13 represents a flow chart for explaining an operation of the echo suppressor in a sixth embodiment in accordance with the present invention; Fig. 14 is a flow chart for explaining an operation of the echo suppressor in the sixth embodiment; and

Fig. 15 illustrates an attenuation amount of the echo suppressor with time in the sixth embodiment.

Detailed Description of the Preferred Embodiments Hereinafter, embodiments of the present invention will be described in more detail with reference to accompanying drawings which form a part hereof .

First Embodiment Referring to Fig. 1, a speakerphone apparatus of the present embodiment includes a microphone 1; a speaker 2; a voice switch 10 which changes over a communications status between transmission and reception, by inserting losses in a reception signal path (shown by a dotted line A in Fig. 1) through which a reception signal coming from a communications terminal of a counterpart is transmitted to the speaker 2 and a transmission signal path (shown by a dotted line B in Fig. 1) through which a transmission signal collected by the microphone 1 is transmitted to the communications terminal of the counterpart; an echo canceller 20 suppressing an acoustic echo caused by an acoustic coupling between the microphone 1 and the speaker 2; and a microphone amplifier Gl for amplifying an output signal (transmission signal) of the microphone 1; a speaker amplifier G2 for amplifying an input signal (reception signal) of the speaker 2.

Further, the speakerphone apparatus includes an amplifier G3 for adjusting a transmission volume which is interposed between the echo canceller 20 and the voice switch 10 in the transmission signal path; and an echo suppressor 30 which attenuates a residual echo by inserting a certain attenuation amount.-in the transmission signal path. The echo canceller 20 has an adaptive filter 21 and a subtracter 22 which are well known. The echo canceller 20 adaptively identifies an impulse response of a feedback path (acoustic echo path) Hac formed, by an acoustic coupling between the speaker 2 and the microphone 1 by the adaptive filter 21, and suppresses an echo component (acoustic echo) estimated from a reference signal (an input signal to the speaker amplifier G2) by subtracting the estimated acoustic echo from an output signal of the microphone amplifier Gl by the subtracter 22.

The voice switch 10 includes a transmission signal attenuator 11 for inserting a loss in the transmission signal path, a reception signal attenuator 12 for inserting a loss in the reception signal path, and an insertion loss amount controller 13 for controlling an amount of losses inserted by respective transmission and reception signal attenuators 11, 12.

The insertion loss amount controller 13 has a total loss amount calculator 14 which estimates an acoustic feedback gain α of a path (hereinafter, referred to as an "acoustic feedback path) C fed back to an input point Tin of the transmission signal attenuator 11 from an output point Rout of the reception signal , attenuator 12 via the acoustic echo path Hac, and calculates a total sum of the loss amounts (a sum of insertion loss amounts of the transmission signal attenuator 11 and the reception signal attenuator 12) to be inserted based an estimation α' of the acoustic feedback gain α.

Further, the insertion loss amount controller 13 includes an insertion loss amount (ILA) allocation processor 15 , which estimates a communications status by monitoring the transmission and the reception signal, and, based on the estimated communications status and . the calculated total loss amount, determines an allocation of respective insertion loss amounts to the transmission signal attenuator 11 and the reception signal attenuator 12.

Further, the echo canceller 20 and the voice switch 10 may be realized by controlling hardwares, e.g., DSP (digital signal processor), by softwares, e.g., programs for the echo canceller and the voice switch. Therefore, signals

(transmission signals and reception signals) inputted to and outputted from the voice switch 10 and the echo canceller 20 are sampled and quantized at a certain sampling period by an

A/D converter (analog to digital converter) not shown. The total loss amount calculator 14 estimates respective average powers of a signal inputted to the transmission signal attenuator 11 and a signal outputted from the reception signal attenuator 12 for short time periods by using a rectifier, a low pass filter, and/or the like. The total loss amount calculator 14 also obtains an average power estimation of the output signal of the reception signal attenuator 12 at . a maximum delay time expected in the acoustic echo path Hac . Then, the total loss amount calculator 14 sets an estimation α' of the acoustic feedback gain to a value obtained by dividing the average power estimation of the output signal of the reception signal attenuator 12 at a maximum delay time by the ^■ average power estimation of the input signal of the transmission signal attenuator 11. '

Further, the total loss amount calculator 14 calculates a total loss amount Lt by using the estimation α' of the acoustic feedback gain α and a desired gain margin MG, and outputs the total loss amount Lt to the insertion loss amount allocation processor 15.

The insertion loss amount allocation processor 15 monitors respective input/output signals of the transmission signal attenuator 11 and the reception signal attenuator 12 and determines a communications status (reception state, transmission state) based on a size relationship in power levels of these input/output signals and a presence of a voice signal. Further, the insertion loss amount allocation processor 15 adjusts insertion loss amounts of respective attenuators 11 and 12 such that the total loss amount Lt can be allocated to the transmission signal attenuator 11 and the reception signal attenuator 12^' proportionally based on the determined communications status.

In the present embodiment, the total loss amount calculator 14 has an update mode in which a total sum of the loss amounts to be inserted is calculated and adaptively updated based on the estimation α' of the acoustic feedback gain α, and a fixed mode in which the total loss amount is fixed to a initial value. The total loss amount calculator 14 operates in the fixed mode before the echo canceller 20 fully converges after starting communications and then in the update mode after the echo canceller fully converges .

That is, once the value of the estimation α' of the acoustic feedback gain α remains below a threshold ε (e.g., a value 10 dB^' ~ 15 dB less than the estimation α' upon starting communications)^' for a certain time period, e.g., several hundreds of milliseconds, or more after starting communications, the total loss amount calculator 14 considers that the echo canceller 20 has sufficiently converged and switches to operate in the update mode thereafter and, until' then, it operates in the fixed mode. Further, in the fixed mode, the initial value is set to a value sufficiently larger than the total loss amount updated as occasion demands in the update mode..

Thus, while the echo canceller 20 does not sufficiently converge right after starting communications, the total loss amount calculator 14 operates in the fixed mode and sets a total loss amount to an initial value having a sufficiently large value and, accordingly, it is possible to suppress an occurrence of the unpleasant echo (acoustic echo) or the howling and realize a stable half duplex communications

Further, when the echo canceller 20 fully converges after the lapse of a reasonable time from the time when the communications start, an operation mode of the total loss amount calculator 14 changes to the update mode from the fixed mode and the total loss amount to be inserted decreases to a value sufficiently lower than the initial value and, therefore, it is possible to realize the full duplex communications.

Hereinafter, a specific operation of the total loss amount calculator 14 in the update mode will be explained with reference to Fig. 2.

The total loss amount calculator 14 carries out an estimation process of the acoustic feedback gain α in a certain sampling period from a time point when the fixed mode changes to the update mode and calculates an estimation α' thereof (step 1) . From the estimation α' and a gain margin MG, a desired total loss mount Lr(n) is calculated by using a following formula (step 2) such that a gain margin MG[dB] of the closed loop can be maintained:

Lr(n) = 20 log I α' (n) | + MG[dB]

Herein, α' (n) and Lr(n) are ^■ the estimation a^r of the acoustic feedback gain α and the desired total loss amount, respectively, which are calculated by an nst sampling from the time point when the fixed mode changes to the update mode. Further, when the desired total loss amount Lr(n) is greater than a previous total loss amount Lt(n-l) determined at an (n-l)st sampling and actually inserted, the total loss amount calculator 14 adds a minute increment Δi [dB] to the previous total loss amount Lt(n-l) (step 3 and 4). That is, the total loss amount calculator 14 sets a total loss amount Lt (n) to Lt (n-l)+Δi.

Further, when the desired total loss amount Lr(n) is smaller than the previous total loss amount Lt(n-l), the total loss amount calculator 14 subtracts a minute decrement

Δd [dB] from the previous total loss amount Lt(n-l) and sets the total loss amount thereto (step 5 and 6) . That is,

Lt (n)=Lt (n-1) -Δd.

As above, since the total loss amount calculator 14 increases and decreases a total loss amount by minute values Δi and Δd, respectively, it is possible to eliminate uncomfortable sense in hearing even while an acoustic feedback gain α is wildly fluctuating because the echo canceller 20 is actively updating a coefficient (filter coefficient) to converge, like as right after starting communications with the communications terminal of the counterpart .

As described above with regard to the conventional example, if digital communications such as IP communications are performed among speakerphone apparatuses, a delay may occur in transmission of voice data due to a packet process or a communications buffer process for jitter suppression. Generally, it is known that, when the delay occurs in the voice data transmission between the speakerphone apparatuses, as the amount of the delay time gets longer, the echo tolerance of a speaker upon communicating becomes worse and, accordingly, the speaker can perceive even a small echo to thereby feel unpleasant . Such unpleasant echo can be prevented by setting a greater initial value for the fixed mode than that of the conventional example . However, degradation amount in the echo tolerance may sometimes reach several tens decibel, which cannot be fully covered by a suppression amount of the echo canceller in the update mode. In this case, the communications are performed in half duplex other than full duplex. Furthermore, when the noise level is high, there may occur a problem in which the voice switch continues to switch on only one side of the transmission side and reception side and, accordingly, a voice from one side is not transmitted to the other side. In this regard, there is additionally provided the echo suppressor 30 after the echo canceller 20 and the voice switch 10 in the transmission signal path, in the present embodiment. The echo suppressor 30 attenuates an acoustic echo which should have been suppressed by the echo canceller 20 (hereinafter, referred to as a 'residual echo') . The echo suppressor 30 needs to effectively attenuate the residual echo while transmit the voice signal (transmission signal) without attenuating.

The echo suppressor 30 inserts an attenuation amount into the transmission signal path in conjunction with the voice switch 10 and, more specifically, operates as illustrated in the flowchart shown in Fig. 3. That is, the echo suppressor 30 continues to monitor (step 11) a status of the voice switch 10 (an estimated result of the communications status (reception state or transmission state) by the insertion loss amount allocation processor 15) . When the voice switch 10 is in the reception state, the echo suppressor 30 deems that no voice signal is transmitted to the transmission signal path, attenuates and outputs (step 12) a signal inputted from the voice switch 10 by multiplying it by a certain attenuation coefficient.

Further, when the voice switch 10 is in the transmission state, the echo suppressor 30 deems that there is no residual echo to be eliminated or a voice signal (transmission signal) to be transmitted, and outputs the signal inputted from the voice switch 10 as it is without attenuating it (step 13) .

With the first embodiment in accordance with the present invention, even if a transmission delay occurs in a voice transmitted between the communications apparatuses, the echo suppressor effectively attenuates the residual echo occurring in the transmission signal path due to the transmission delay, thereby realizing a full duplex communications .

At this point, when the echo suppressor 30 inserts an attenuation amount in the transmission signal path while the voice switch 10 is, e.g., in the transmission state other than in the reception state, a voice emitted from a speaker at the near-end side (a speaker communicating by using a speakerphone apparatus in the present embodiment) is attenuated, which is undesirable, and may be heard by a speaker at the counterpart communications apparatus like inflection of the voice in which a volume thereof is made lower or higher. However, in the present embodiment, the echo suppressor 30 inserts an attenuation amount in the transmission signal path when the voice switch 10 is in the reception state and does not when the voice switch 10 is in the transmission state. Therefore, it is possible to attenuate only an unpleasant echo (residual echo) without an occurrence of the voice inflection.

Further, if the echo suppressor 30 is disposed before the voice switch 10 in the transmission signal path and erroneously attenuates the transmission signal, a voice inflection may occur in the voice switch 10 due to an erroneous attenuation of the transmission signal by the echo suppressor 30. However, in the present embodiment, since the echo suppressor 30 is arranged after the echo canceller 20 and the voice switch 10 in the transmission signal path and respective processes are performed in that order, an occurrence of the voice inflection can be suppressed in the voice switch 10 even when the echo suppressor 30 erroneously attenuates the transmission signal. Therefore, it is possible to provide a pleasant environment for communications .

Meanwhile, there has been proposed a speakerphone apparatus in which an attenuator for inserting a loss is provided after the voice switch in the transmission signal path, by the present inventors (see Japanese Patent Application Publication No. 2003-324371) . However, in this document, the attenuator inserts a certain amount of loss in the transmission signal path only when a momentary power estimation of the transmission signal remains below a threshold for a certain time period. Therefore, if the echo has a large volume higher than the threshold, the attenuator does not insert a loss and an unpleasant echo can be transmitted to the counterpart .

Further, even if a power level of the voice emitted from a speaker at the near-end side is lower than the threshold, the attenuator may accidentally attenuate the voice signal by inserting a loss. In the speakerphone apparatus in accordance with the embodiment, however, the echo suppressor 30 operates to insert an attenuation amount in conjunction with the voice switch 10, thereby preventing the above from occurring.

-27- i Second Embodiment

Fig. 4 schematically shows a block diagram of the speakerphone apparatus in accordance with a second embodiment of the present invention. The speakerphone apparatus of the second embodiment is different from that of the first embodiment in that the echo suppressor 30 provides an attenuation amount in the transmission signal path in conjunction with the echo canceller 20 and the others are same as those in the first embodiment. Therefore, same components are designated by same numerals and description thereof will be omitted.

In this embodiment, the echo canceller 20 includes a doubletalk detector 23 in addition to the adaptive filter 21 and the subtracter 22. The doubletalk detector 23 detects a doubletalk, i.e., a state in which the near-end speaker and the far-end speaker are substantially simultaneously talk to each other, based on whether or not a signal outputted from the microphone amplifier Gl includes a level of signal interfering convergence of the adaptive filter 21. When the doubletalk detector 23 detects the doubletalk, the adaptive filter 21 does not update a filter coefficient and keeps the previous value .

Next, an operation of the echo suppressor 30 will be explained with reference to the flowchart shown in Fig. 5. The echo suppressor 30 continues to monitor the doubletalk detector 23 of the echo canceller 20 (step 21) . When the doubletalk detector 23 does not detect a doubletalk, the echo suppressor 30 decides that a residual echo to be eliminated exists or no voice signal to be transmitted is in the transmission signal path, and attenuates an input signal by multiplying it by a certain attenuation coefficient and outputs it (step 22) .

Further, when the doubletalk detector 23 detects a doubletalk, the echo suppressor 30 decides that there is a voice signal to be transmitted, and outputs an input signal without attenuating it by not multiplying it by a certain attenuation coefficient (step 23) .

With the second embodiment, even if a transmission delay occurs in a voice transmitted between communications apparatuses, the echo suppressor can effectively attenuate the residual echo occurring in the transmission signal path due to the transmission delay, thereby realizing a full duplex communications .

In this embodiment, the echo suppressor 30 does not inserts an attenuation amount in the transmission signal path while the doubletalk detector 23 is detecting the doubletalk, and inserts the attenuation amount in the transmission signal path when the doubletalk detector 23 does not detect the doubletalk.' Therefore, it is possible to prevent a voice emitted from a speaker at the near-end side from being erroneously attenuated and an inflection from occurring in the voice. Further, since an unpleasant echo can be attenuated upon communicating, it is possible to provide a pleasant environment for communications .

Further, since a doubletalk detection means (doubletalk detector 23) provided in the echo canceller 20 can be used in this embodiment, simplification of configuration, ' e.g., decrease in amounts of programs executed in the DSP, lower specifications in the DSP, reduction in capacity of ROM for storing the programs , and the like, and a cost reduction can be achieved. Third Embodiment

Fig. 6 illustrates a block diagram of a speakerphone apparatus in accordance with a third embodiment of the present invention. A basic configuration of the present embodiment is same as that of the first and second embodiments except that the echo suppressor 30 inserts an attenuation amount in the transmission signal path in conjunction with both of the voice switch 10 and the echo canceller 20. Therefore, same components are designated by same numerals and description thereof will be omitted. Hereinafter, operation of the echo suppressor 30 will be explained with reference to Fig. 7. The echo suppressor 30 always monitors a status of the voice switch 10 (step 31) and, when the voice switch 10 is in the reception state, determines whether or not the doubletalk detector 23 detects a doubletalk (step 32) . When the doubletalk detector does not detect a doubletalk, the echo suppressor 30 deems that there is a residual echo to be eliminated or no voice signal to be transmitted, and . attenuates an input signal by- multiplying it by a certain attenuation coefficient to output it (step 33) . When the voice switch 10 is not in the reception state, or when the doubletalk detector .23 detects a doubletalk while the voice switch 10 is in' the reception state, the echo suppressor 30 determines that there is no residual echo or a voice signal to be transmitted, and outputs the input signal as it is without attenuating (step 34) .

With this embodiment, it is possible to prevent a voice emitted from a speaker at the near-end side from being erroneously attenuated and an inflection from occurring in the voice. Further, the echo suppressor 30 can attenuate only an unpleasant echo while an unpleasant residual echo does not occur because the echo suppressor 30 does not accidentally insert an attenuation amount, thereby providing a pleasant environment for communications .

Herein, the echo suppressor 30 may adjust an attenuation amount inserted in the transmission signal path based on an estimation α' of the acoustic feedback gain α which is estimated by the total loss amount calculator 14. Next, description will be made on an operation of the echo suppressor 30 when the echo suppressor 30 adjusts an attenuation amount inserted in the transmission signal path, with reference to Fig. 8. The echo suppressor 30 monitors a state of the voice switch 10 all the time (step 41) . When the voice switch 10 is in the reception state, the echo suppressor 30 determines whether or not the doubletalk detector 23 detects a doubletalk (step 42) . When the voice switch 10 is not in the reception state, or when the doubletalk detector detects a doubletalk while the voice switch 10 is in the reception state, the echo suppressor 30 deems that there is no residual echo or a voice signal to be transmitted, and outputs the input signal as it is without attenuation it (step 47) .

On the other hand, when the voice switch 10 is in the reception state and the doubletalk detector 23 does not detect a doubletalk, the echo suppressor 30 sets an attenuation coefficient which is multiplied to the input signal to a normal value SUP_MIN (step 43) . Thereafter, when a sum of the estimation a' of the acoustic feedback gain α estimated by the voice switch 10 and the t.otal loss amount Lt (n) is a threshold TH or more (step 44) , the echo suppressor 30 changes the attenuation coefficient from the normal value SUP_MIN to an upper limit SUP_MAX greater than the normal value (step 45) .

Further, when a sum 'of the estimation α' of the acoustic feedback gain α estimated by the voice switch 10 and the total loss amount Lt (n) is less than the threshold value TH (step 44) , the echo suppressor 30 maintains the normal value SUP_MIN as the attenuation coefficient without change. Finally, the echo suppressor 30 attenuates the input signal by multiplying it by the decided attenuation coefficient (normal value SUP_MIN or UPPER LIMIT SUP_MAX) and outputs the result (step 46) .

With this configuration, even if a voice emitted from a speaker at the near-end side is erroneously attenuated by the echo suppressor 30, a voice inflection can be prevented. Meanwhile, if a speaker is in the environment where an ambient noise is normally high level, the adaptive filter 21 of the echo canceller 20 does not update a filter coefficient for a significant time period after starting communications. For that reason, the echo canceller 20 does not converge and the doubletalk detector 23 cannot correctly detect a doubletalk. In such case, even when the voice switch 10 is in the reception state and the doubletalk detector 23 does not detect a doubletalk, the echo suppressor 30 may output the input signal as it is without attenuating it, which may cause an unpleasant residual echo. In this regard, while the echo canceller 20 does not sufficiently converge after starting communications through the communications terminal of the counterpart, the echo suppressor 30 inserts a certain attenuation amount in the transmission signal path when the insertion loss amount allocation processor 15 estimates that the voice switch 10 is in the reception state. That is, the echo suppressor 30 attenuates an unpleasant residual echo even while the echo canceller 20 does not sufficiently converge after starting communications, thereby achieving a pleasant environment for the speakerphone communications . On the other hand, if different voices are simultaneously inputted from the near-end and the far-end sides while the total loss amount calculator 14 is operating in the fixed mode as right after starting communications, the voice switch 10 alternates between the reception state and the transmission state. For that reason, the echo suppressor 30 does not correctly operate and, accordingly, an inflection may occur in the transmitted voice in which a level of the transmitted voice greatly fluctuates.

In this regard, when the insertion loss amount allocation processor 15 estimates that the voice switch 10 is not in the reception state or the doubletalk detector 23 detects a doubletalk while a certain attenuation amount is being inserted in the transmission signal path, the echo suppressor 30 stops inserting a certain attenuation amount into the transmission signal path in a moment. This makes it possible to prevent an occurrence of the inflection due to the erroneous attenuation of the transmission signal, thereby providing a pleasant condition for the speakerphone communications . Fourth Embodiment

Fig. 9 shows a block diagram of a speakerphone apparatus in accordance with a fourth embodiment of the present invention. In this embodiment, it is characterized that the echo suppressor 30 adjusts an attenuation amount by- using a signal level average value of the reception signal and the other basic configuration of the present embodiment is same as that of the third embodiment. Therefore, same components are designated by same numerals and description thereof will be omitted.

In the present embodiment, the adaptive filter 21 adaptively identifies an impulse response of the acoustic echo path Hac, and estimates an echo component (acoustic echo) g(n) from a signal y(n) inputted from the far-end side, i.e., the reception signal inputted to the speaker amplifier G2. To estimate the echo component g(n), the filter coefficient hp(n) is updated by using a following formula: Hp(n+1) = hp(n) + K-F(Ey(n))

Herein, p is a tap number, n is a sample period, K is a step gain, F(n) is a coefficient update function. Further, Ey (n) is a signal level average value of the input signal (reception signal) y(n), which is calculated by using a below formula in the signal level average (SLA) operation unit 24 provided in the adaptive filter 21. Herein, SPAN is a period for which a magnitude abs[y(n)] of the input signal y(n) is added.

SPAN-I

Ey(X) = 2_^ ^abs[y(n-i)]

J=O Further, the echo component (estimation) g(n) can be calculated by using a following formula:

N-I gCn) = ^hp(ru)y(n-k)

In the present embodiment, the echo suppressor 30 obtains the signal level average value Ey (n) from the SLA operation unit provided in the adaptive filter 21.

Hereinafter, operation of the echo suppressor 30 will be explained with reference to a flow chart shown in Fig. 10. The echo suppressor 30 monitors a status of the voice switch 10 all the time (step 51) . When the voice switch 10 is in the reception state, the echo suppressor 30 determines whether or not the doubletalk detector 23 of the echo canceller 20 detects a doubletalk (step 52) . When the voice switch 10 is not in the reception state, or when the doubletalk detector 23 detects a doubletalk and the voice switch 10 is in the reception state, the echo suppressor 30 deems that there is no residual echo or a voice signal to be transmitted and outputs the input signal as it is without attenuating it (step 58) . On the other hand, when the voice switch 10 is in the reception state and the doubletalk detector 23 does not detect a doubletalk, an attenuation coefficient multiplied to the input signal by the echo suppressor 30 depends on a signal level average value RS_AVE,(n) (=Ey(n)) calculated by the SLA operation unit 24. More specifically, the attenuation coefficient can be determined as a following formula (1) when RS_AVE(n)>TH and a following formula (2) when RS_AVE (n) ≤TH:

RS

ATTENUATION COEFFICIENT = SUP_MAS X ^{AVE(n) ■ ■ ■}- (l)

RS

ATTENUATION COEFFICIENT = SUP_MAX X ^AV∑{Ώ} ~SUP_Mrer ^{• • • •} (2)

^AVEmax

Herein, SUPJYLAX and ,SUP_MIN are attenuation coefficients needed to suppress signal levels corresponding to a loud voice (e.g., QOdB sound pressure or more) and a normal volume (e.g., OO~OOdB sound pressure), respectively, and RS_AVEmax is a signal level average value RS_AVE(n) of the loud voice. Regarding the formula (2), while RS_AVE (n) ≤TH every sampling period, the attenuation coefficient is multiplied by a certain transition coefficient and gradually reduced from SUP_MAXxRS_AVE(n) /RS_AVEmax to SUP_MIN in which SUP_MIN is set to a lower limit of the attenuation coefficient.

More specifically, the .echo suppressor 30 first determines whether or not an attenuation coefficient is set to SUP-MIN (step 53) . When the attenuation coefficient is not set to SUP_MIN, the echo suppressor 30 multiplies it by the transition coefficient, thereby setting a new attenuation coefficient (step 54) . If otherwise, SUP_MIN is used as the attenuation coefficient because it is the lower limit thereof . Next, if RS_AVE (n) >TH, that is, a condition of the formula (1) is satisfied (step 55) , the echo suppressor 30 sets an attenuation coefficient to SUP_MAXxRS_AVE (n) /RS_AVEmax (step 56) and, if otherwise, use one set in a previous step without changing the attenuation coefficient. Finally, the echo suppressor 30 attenuates the input signal to output by multiplying the input signal by the set attenuation coefficient (step 57) .

With the present embodiment, even if a voice emitted from a speaker at the near-end side is accidentally attenuated by the echo- suppressor 30, an occurrence of the voice inflection can be suppressed. Further, since a signal level average value RS_AVE(n) is calculated by using a signal level average operation unit 24 provided in the echo canceller 20, simplification of configuration, e.g., decrease in amounts of programs executed in the DSP, lower specifications in the DSP, reduction in capacity of ROM for storing the programs, and the like, and a cost reduction can be achieved. Fifth Embodiment

Next, a fifth embodiment of the present invention will be explained with reference to drawings . In the present embodiment, the echo suppressor 30 inserts an attenuation amount in the transmission signal path in conjunction with the voice switch 10, the doubletalk detector 23 of the echo canceller 20, and the signal level average (SLA) operation unit 24 provided in the adaptive filter 21 of the canceller 20. The other configuration of the embodiment is same as that of the fourth embodiment. Therefore, same components are designated by same numerals and description thereof will be omitted.

Operation of the echo suppressor 30 will be explained in detail with reference to a flow chart shown in Fig. 11. Steps 61 to 67 in Fig. 11 are same as steps 51 to 57 in Fig. 10 related to the fourth embodiment. Therefore, description thereof will be omitted.

When the voice switch 10 is not in the reception state, or when the doubletalk detector ^' 23 detects a doubletalk while the voice switch 10 is in the reception state, the echo suppressor 30 further determines whether or not the voice switch 10 is in the transmission state (step 68) . When the voice switch 10 is in the transmission state, the echo suppressor 30 determines that there is no residual echo or a voice signal to be transmitted and outputs the input signal as it is without attenuating it (step 611) . If communications are performed between the speakerphone apparatus in accordance with the present invention and communications apparatus using a handset (hereinafter, referred to as a ^λ handset communications apparatus') and a level of an ambient noise is fairly high in the transmission signal transmitted to the handset communications apparatus from the speakerphone apparatus, the ambient noise is momentarily attenuated as the echo suppressor 30 in the speakerphone apparatus attenuates a residual echo. Such attenuation of the ambient noise is heard by the handset with a delay . and, accordingly, a speaker at the handset communications apparatus may feel an unpleasant sense of disconnection in voice during communications.

In this regard, in the present embodiment, when the voice switch 10 is not in the transmission state in step 68, e.g., the insertion loss amount allocation processor 15 inserts same loss amounts in the transmission and reception signal attenuators 11 and 12 (hereinafter, this state is referred to as a ^λmiddle state' ) , the echo suppressor 30 monotonously attenuates the attenuation amount inserted in the transmission signal path.

More specifically, when the voice switch 10 is not in the transmission state in step 68, the echo suppressor 30 further determines whether or , not the attenuation coefficient is zero (step 69) . If the attenuation coefficient is zero, the process returns to step 61 without changing the attenuation coefficient. If otherwise, the echo suppressor 30 sets a new attenuation coefficient by multiplying the attenuation coefficient by the transition coefficient (step 610) . That is, the echo suppressor 30 gradually decreases (monotonous reduction as shown by a line A in Fig. 12) the attenuation coefficient (attenuation amount) by repeating processes of step 61 or 62 ->step 68->step 69->step 610-^step 61.

In the present embodiment, since the echo suppressor 30 monotonously decreases _'an, attenuation amount inserted in the transmission signal path, the- ambient noise is not momentarily attenuated. This makes it difficult to cause a speaker at the handset communications apparatus to feel an unpleasant sense of disconnection.

Meanwhile, when a communications status of the voice switch 10 change to the transmission state while the attenuation coefficient (attenuation amount) is being monotonously decreased by repeating the processes of step 61 or 62->step 68->step 69->step 610->step 61, the echo suppressor 30 immediately reduces the attenuation coefficient (attenuation amount) to zero (see, e.g., a dotted line B in Fig. 12) . Therefore, it is possible to prevent an inflection from occurring in the voice which is heard by a speaker at the counterpart communications apparatus because a voice (transmission voice) emitted from a speaker at the speakerphone apparatus is accidentally attenuated by the echo suppressor 30.

Further, when a communications status of the voice switch 10 change to the transmission state and the doubletalk detector 23 does not detect a doubletalk while the attenuation coefficient (attenuation amount) is being monotonously decreased by repeating the processes of step 61 or 62->step 68->step 69->step 610->step 61, the echo suppressor 30 inserts a certain attenuation amount in the transmission signal path by performing processes of steps 63 to 67 (see, e.g., a dotted line C in Fig. 12). This makes it possible to prevent the echo suppressor 30 from accidentally not inserting an attenuation amount and an unpleasant residual echo from occurring and to accurately attenuate only an unpleasant echo during communications .

In the present embodiment, a temporal ratio (reduction ratio) at which the echo suppressor 30 monotonously decreases an attenuation amount is constant (see, e.g., a line A in Fig. 12) . When the attenuation amount is linearly attenuated as above, a time period (transition time period) needed to decrease to zero can ■ be shortened compared to a case in which the attenuation amount is attenuated step by step.

Sixth Embodiment

Next, a sixth embodiment in accordance with the present invention will be explained with reference to Figs. 13 to 15. As shown in Fig. 13, the sixth embodiment is different from the fifth embodiment in that step 712 for controlling an attenuation amount based on an ambient noise level is provided between steps 77 and 78. Steps 71 to 79 and steps 710 and 711 are^' same as steps 61 to 69 and steps 610 and 611 described in the fifth embodiment and description thereof will be omitted. In the present embodiment, the echo suppressor 30 further controls an attenuation amount in conjunction with an ambient noise level. A process for controlling an attenuation amount in conjunction with an ambient noise level is illustrated in the flow chart shown in Fig. 14.

That is, an attenuation amount inserted in the transmission signal path by the echo suppressor 30 is reduced as an ambient noise level at a side of the speakerphone apparatus becomes higher. Therefore, it is possible to suppress an instant attenuation of the ambient noise, thereby making it difficult to feel a sense of disconnection. In this case, if the echo suppressor decreases the attenuation amount, an attenuation effect of the residual echo is lowered. However, the residual echo is masked by the ambient noise and does not make a trouble in communications in a practical use.

Hereinafter, the process for controlling an attenuation amount in conjunction with an ambient noise level with reference to the flow chart shown in Fig. 14. The echo suppressor detects (estimates) a level of the ambient noise (ambient noise level) LEVEL_NOISE continuously existing in the transmission signal by using a long term average value of the transmission signal inputted from the input point Tin of the transmission signal attenuator 11, and compares the estimated ambient noise level LEVEL_NOISE with a threshold Nth (step 821) . When the ambient noise level LEVEL_NOISE is the threshold Nth or less, the echo suppressor 30 calculates a normal attenuation coefficient ATT by using a follow forraaula (step 822) :

ATT=αxLEVEL_NOISE+β, α <0 Further, when the ambient noise level LEVEL_NOISE exceeds the threshold Nth, the echo suppressor 30 sets the normal attenuation coefficient ATT to a constant value

ATTmin regardless of the ambient noise level LEVEL_NOISE

(step 823) . Then, the echo suppressor 30' compares the attenuation coefficient (attenuation amount corresponding to the echo amount estimated from the reception signal level) calculated by processes of steps 73 to 76 with the normal attenuation coefficient ATT calculated by processes of steps 821 to 823 (step 824) . When the attenuation coefficient is the normal attenuation coefficient ATT or less, the echo suppressor 30 updates the attenuation coefficient by the normal attenuation coefficient ATT (step 825) to proceed to step 77. Further, when the attenuation coefficient is greater than the normal attenuation coefficient ATT, the echo suppressor 30 proceeds to step 77 without updating the attenuation coefficient .

That is, in the steps 824 and 825, the echo suppressor 30 compares an attenuation amount (see, e.g., a curve D in Fig. 15) corresponding to the echo amount estimated from the voice signal level in the reception signal path with the attenuation amount (see, e.g., a curve E in Fig. 15) determined based on the ambient noise level and then inserts one having a relatively ^'• small magnitude among the attenuation amounts in the transmission signal path. Further, the echo suppressor 30 sets the attenuation amount

(attenuation coefficient) to the constant value ATT based on the ambient noise level LEVEL_NOISE when the ambient noise level LEVEL_NOISE exceeds the threshold Nth. This makes it possible to prevent the residual echo from not being attenuated when the ambient noise level is greatly high.

While the invention has been 'Shown and described with respect to the embodiments, it will be understood by those skilled in the art that various changes and modifications may be made without departing from the scope of the invention as defined in the following claims.

Claims

What is claimed is:

1. A speakerphone apparatus comprising: a microphone; a speaker; a voice switch adapted to switch over a communications status between transmission and reception by providing a loss in a reception signal path through which a receiving signal from a communications terminal of the counterpart is transmitted to the speaker and a transmission signal path through which a transmitting signal collected by the microphone is transmitted to the communications terminal of the counterpart ; an echo canceller for suppressing an acoustic echo caused by an acoustic coupling between the microphone and the speaker; a doubletalk detector for detecting a doubletalk; and an echo suppressor adapted to attenuate a residual echo by inserting a certain attenuation amount in the transmission signal path in conjunction with the voice switch and/or the doubletalk detector, wherein the voice switch includes a transmission side loss insertion unit for inserting a loss in the transmission signal path; a reception side loss insertion unit for inserting a loss in the reception signal path; and an insertion loss amount control unit controlling respective amounts of the losses inserted by the transmission and reception side loss insertion units, the insertion loss amount control unit includes a total loss amount calculator which is adapted to estimate an acoustic feedback gain of a path fed back to an input terminal of the transmission side loss insertion unit from an output terminal of the reception ^' side loss insertion unit via an acoustic echo path, and calculate a total sum of the loss amounts inserted in the paths based on the estimation of the acoustic feedback gain; and an insertion loss amount allocation processor adapted to monitor the transmission signal and the reception signal and estimate a communications status, and, based on a result of the estimation and a value calculated by the total loss amount calculator, determine an allocation of insertion loss amounts to the transmission side loss insertion unit and the reception side loss insertion unit, and the total loss amount calculator has an update mode in which a total sum of the loss amounts inserted in the paths is calculated and adaptively updated based on respective estimations of the feedback gain, and a fixed mode in which a total loss amount is fixed to a initial value, and operates in the fixed mode while the echo canceller does not sufficiently converge after starting communications and in the update mode after the echo canceller fully converges.

2. The speakerphone apparatus of claim 1, wherein the echo suppressor inserts the certain attenuation amount in the transmission signal path if the insertion loss amount allocation processor estimates that the voice switch is in a reception state, and not if otherwise.

3. The speakerphone 'apparatus of claim 1, wherein the echo suppressor does not insert the certain attenuation amount in the transmission signal path while the doubletalk detector is detecting a doubletalk, and do if otherwise.

4. The speakerphone apparatus of claim 3, wherein the echo suppressor inserts the certain attenuation amount in the transmission signal path if the doubletalk detector does not detect a doubletalk and the insertion loss amount allocation processor estimates that the voice switch is in the reception state. ,

5. The speakerphone apparatus of claim 4, wherein, while the echo canceller does not sufficiently converge after starting communication, the echo suppressor inserts the certain attenuation amount in the transmission signal path only if the insertion loss amount allocation processor estimates that the voice switch is in the reception state.

6. The speakerphone apparatus of any one of claims 3 to 5 , wherein, during inserting the certain attenuation amount in the transmission signal path, the echo suppressor immediately stops insertion of the attenuation amount into the transmission signal path if the insertion loss amount allocation processor estimates that the voice switch is not in the reception state or the doubletalk detector detects a doubletalk.

7. The speakerphone apparatus of any one of claims 1 to 6, wherein the echo canceller includes an adaptive filter for adaptively identifying characteristics of an echo path, a subtracter for subtracting an output of the adaptive filter from the transmission signal, and the doubletalk detector for detecting a doubletalk, and the echo canceller does not update a coefficient of the adaptive filter when the doubletalk detector detects a doubletalk.

8. The speakerphone apparatus of any one of claims 1 to 7, wherein there is provided the echo canceller, the voice switch, and the echo suppressor in that order in the transmission signal path, and respective processes thereof are performed.

9. The speakerphone apparatus of any one of claims 1 to 8 , wherein the echo suppressor adjusts the attenuation amount based on the acoustic feedback gain estimated by the total loss amount calculator.

10. The speakerphone apparatus of any one of claims 1 to 6, wherein the echo suppressor adjusts the attenuation amount based on the signal level average value of the reception signal .

11. The speakerphone apparatus of claim 10, wherein the echo suppressor sets the attenuation amount inserted in the transmission signal path to a certain normal value when the signal level of the reception signal is less than a threshold, and adjusts the attenuation amount in a range from an upper limit greater than the certain normal value to the certain normal value when the signal level is the threshold or more.

12. The speakerphone apparatus of claim 10, wherein the echo canceller includes a signal level average (SLA) operation unit for calculating signal level average value of the reception signal, an adaptive filter for adaptively identifying characteristics of the echo path, and a subtracter for subtracting an output of the adaptive filter from the transmission signal, and the echo suppressor adjusts the attenuation amount by using the signal level average value calculated by the signal level average value operation unit.

13. The speakerphone apparatus of claim 1, wherein the echo suppressor inserts a certain attenuation amount in the transmission signal path only if the doubletalk detector does not detect a doubletalk and the insertion loss amount allocation processor estimates that the voice switch is in the reception state, and monotonously decreases the attenuation amount inserted in the transmission signal path when the insertion loss amount allocation processor estimates that the voice switch is neither in the reception state nor in the transmission state, or when the insertion loss amount allocation processor estimates that the voice switch is in the reception state and the doubletalk detector detects a doubletalk.

14. The speakerphone apparatus of claim 13, wherein the echo suppressor decreases the attenuation amount to zero immediately when the communications status estimated by the insertion loss amount allocation processor becomes the transmission state while the attenuation amount is being monotonously decreased.

15. The speakerphone apparatus of claim 13 or 14, wherein, while the attenuation amount inserted in the transmission signal path is being monotonously decreased, the echo suppressor inserts the certain attenuation amount in the transmission signal path again when the doubletalk detector does not detect a doubletalk and the communications status estimated by the insertion loss amount allocation processor is the reception state.

16. The speakerphone apparatus of any one of claims 13 to 15, wherein the echo suppressor decreases the attenuation amount inserted in the transmission signal path at a constant decrease ratio when the communications status estimated by the insertion loss amount allocation processor is neither the reception state nor the transmission state, or when the communications status is the reception state and the doubletalk detector detects a doubletalk.

17. The speakerphone apparatus of claim 1, wherein the echo suppressor inserts the certain attenuation amount in the transmission signal path only when the doubletalk detector does not detect a doubletalk and the communications status estimated by the insertion loss amount allocation processor is the reception state, estimates a level of an ambient noise superimposed in the transmission signal, and reduces the attenuation amount inserted in the transmission signal path as the ambient noise level becomes higher, based on an estimated result of the ambient noise level.

18. The speakerphone apparatus of claim 17, wherein the echo suppressor compares an echo amount estimated from the signal level of the reception signal with an attenuation amount depending on an ambient noise level superimposed in the transmission signal, and inserts one having a relatively small magnitude among the echo amount and the attenuation amount in the transmission signal path.

19. The speakerphone apparatus of claim 18, wherein, when the noise level exceeds a certain level, the echo suppressor adjusts the attenuation amount to a constant value based on the ambient noise level .