JP4838282B2 - Hands-free call device and hands-free call method - Google Patents

Description

  The present invention relates to a technology of a hands-free call device and a hands-free call method capable of making a call with a call partner even if the caller does not hold it in his hand.

  2. Description of the Related Art Hands-free voice communication in which communication is performed using a speakerphone that combines a speaker and a microphone without using a handset as a handset in a communication device that performs a videophone call has been reviewed. In the past, for example, a video telephone system called “PICSEND-R” (see FIG. 3) developed by NTT Corporation has been mainly used to make a call using a built-in handset while watching a TV screen. However, as seen in the “Flet's Phone VP1000” developed by the NTT Group (see FIG. 4), the speaker and microphone are formed integrally with the TV screen, so hands-free without holding the handset in hand. Videophones that can be used to make calls are emerging. By eliminating the need for the handset, the appearance of the communication device itself is better than before, releasing the caller from holding the handset and realizing a free call environment.

  Such a call device capable of hands-free calling has a function of reproducing the voice of the other party from the speaker and the microphone picking up the voice of the other party, but the voice of the other party released from this speaker. Is sent back to the other party through the microphone, and there is a case where a delayed own voice is given to the other party. In addition, when both the call partner and the caller use hands-free calling, a phenomenon called acoustic coupling occurs, and in an extreme case, howling (oscillation noise) occurs. In order to suppress such a feedback phenomenon, generally, a processing circuit called an echo canceler EC or a voice switch VOX is incorporated between a microphone and a speaker, and the voice emitted from the speaker is transmitted to the other party. It is designed to prevent return and to block acoustic coupling.

  By the way, regarding the hands-free call, there is a factor that hinders the spread while having the above-described convenience. Specifically, there is a possibility of eavesdropping by a third party or leakage of privacy due to leakage of call voice. In other words, since the conversation between the caller's voice and the other party's voice emitted from the speaker can be heard at the same time, when a third party is in the same room or the sound in the room is recorded It is very difficult to protect the privacy of the content of calls.

  Therefore, as one method for extremely preventing such a problem, it is conceivable to generate an interfering sound that masks the call voice so that it cannot be heard by others. In European offices, there is the idea that office telephones belong to individuals, and some hands-free users use telephones while listening to music and ambient noise from speakers such as radio and stereo. However, although such a method can be expected to have a certain effect as a countermeasure against eavesdropping, the radio and stereo operate independently of the hands-free call device, and are equivalent to the caller's real voice when viewed as a sound source. That is, since there is no restriction by the above-described echo canceler EC or voice switch VOX, music and ambient noise are transmitted as they are to the other party, making clear voice call difficult. Further, if the sound pressure level of the music or ambient noise is high, annoying disturbing sound is given to a bona fide third party around the caller.

  As described above, in order to prevent the music and ambient noise from being transmitted to the calling party and to obtain the effect as the disturbing sound, a method of generating and canceling the disturbing sound itself with the hands-free call device is conceivable. A technique for realizing such a method with a casing or handset of a mobile phone is disclosed in Patent Document 1.

  FIG. 1 of Patent Document 1 shows an adaptive noise canceller having a function similar to an echo canceller after placing a disturbing sound source in the vicinity of a transmission microphone of a handset and picking up a caller's real voice and disturbing sound with a microphone. A technique is described in which the interference sound is canceled by ANC and only the voice of the caller is output. However, since the disturbing sound that wraps around the caller's own ear through the case of the handset is not taken into consideration, it becomes very difficult to talk with the other party.

  FIG. 5 of the patent document describes a technique for realizing the same technique as described above with a hands-free call device. However, since the receiving speaker and the interfering sound generating speaker are separately arranged, the housing of the apparatus becomes large depending on the arrangement of the speakers.

  Furthermore, in this patent document, since the sound pressure level of the disturbing sound is not considered, there is a high possibility that excessive noise will be scattered around the surroundings, and it is difficult to say that the practical use is considered.

  Furthermore, although a specific circuit configuration for realizing FIG. 5 of the patent document is not described, if estimated from the circuit diagram of FIG. 4 of the patent document, the masking sound generated from the noise generator is collected by a microphone. It is expected that the circuit configuration includes an adaptive filter ANC for canceling the sound, and an echo canceller EC or a voice switch VOX for realizing a hands-free call using a speaker and a microphone. . Since the adaptive filter ANC and the echo canceller EC have substantially the same circuit configuration, the circuit configuration becomes redundant.

Furthermore, the purpose of producing the disturbing sound in the patent document is different. The caller's own voice was sufficiently leaked even when using a handset so far. Considering this as a premise, the first purpose of originally suppressing wiretapping in hands-free calling is the voice of the other party. The second purpose is to prevent eavesdropping of the caller's own voice. The technique disclosed in this patent document is intended to achieve the second object, but is not specialized for achieving the first object.
Japanese Patent Laid-Open No. 5-22391

  To summarize the above, in the conventional hands-free call device, there has been the following problem regarding generation of disturbing sound that suppresses the voice of the other party being wiretapped by a third party.

  First, if the sound pressure level of the interfering sound is increased in order to reliably prevent eavesdropping by a third party, the caller himself / herself will also hear the interfering sound, so the call should be made in a noisy environment. There was a first problem that it was forced to make calls very difficult. Further, in the case where noise equivalent to the disturbing sound exists in the surrounding area, the noise acts as the disturbing sound, so that there is a second problem that the effect of newly generating the disturbing sound is reduced. Furthermore, since the disturbing sound is heard by people in the surroundings, there is a third problem that in the case of generating noise exceeding a certain wiretapping suppression effect, the surrounding people are unpleasantly affected. . In addition, both the speaker for disturbing sound that emits the disturbing sound and the speaker for the other party that emits the voice of the other party are required, so that not only the size of the device casing is increased, but also the interference sound is canceled out. The adaptive filter ANC and the echo canceller EC or voice switch VOX for realizing the hands-free configuration have a multi-stage configuration, and there is a fourth problem that the processing becomes complicated.

  The present invention has been made in view of the above, and it is a first problem to generate a disturbing sound with a simple configuration and appropriately adjust the level of the disturbing sound according to the ambient noise. Accordingly, it is an object of the present invention to provide a hands-free call device and a hands-free call method whose second problem is to appropriately adjust the level of the interference sound according to the situation.

  The present invention according to claim 1 is a hands-free call device comprising a speaker and a microphone and capable of making a call with a call partner even if the caller does not hold it in his / her hand. A disturbing signal generating means for generating a disturbing signal that interferes with listening to a person other than the caller; and adding the disturbing signal to a sound signal transmitted from the call partner communication device, Interference signal addition means for emitting sound based on the sound and the interference sound based on the interference signal from the speaker, and a sound collection signal based on the sound collected by the microphone are input, and corresponds to a non-speech time period that is not a call voice Noise level measuring means for measuring the magnitude of the collected sound signal as a noise level; and the collected sound signal corresponding to a voice time interval that is a voice signal by inputting the collected sound signal A call sound level measuring means for measuring the volume as a call sound level; a difference calculating means for calculating a difference between the noise level with respect to the call sound level; and a magnitude of the jamming signal when the difference is smaller than a predetermined value. And a disturbance signal level control means for increasing the magnitude of the disturbance signal when the difference is larger than the predetermined value.

  According to the present invention, a disturbing signal is generated to prevent the other party's voice from being heard from a speaker from being heard by a speaker, and the disturbing signal is transmitted to the other party's speech device. Since a signal is added and the sound based on the speech signal of the other party and the interference sound based on the interference signal are emitted from the speaker, a hands-free communication device that generates the interference sound with a simple configuration can be provided. Specifically, since the speaker that emits the voice of the other party and the speaker that emits the disturbing sound are combined into one, it is possible to prevent an increase in the size of the hands-free talking device and The arrangement of the speaker is not required, and the present invention can be easily applied to the case of a conventional hands-free call device.

  Further, in the present invention, a sound pickup signal based on a sound picked up by a microphone is input, and the magnitude of the sound pickup signal corresponding to a non-voice time interval that is not a call voice is measured as a noise level. Similarly, a sound pickup signal based on the sound picked up by the microphone is input, and the magnitude of the sound pickup signal corresponding to the voice time interval that is the call voice is measured as a call sound level. In order to reduce the magnitude of the interfering signal when the difference is smaller than the predetermined value, and to increase the magnitude of the interfering signal when the difference is larger than the predetermined value, the level of the interfering sound is adjusted according to the ambient noise. It is possible to provide a hands-free communication device that is appropriately adjusted. Specifically, the ambient noise is evaluated in a time interval in which the caller or the call partner does not make a sound, and if a sufficient noise level is present relative to the sound level, the sound pressure level of the interference sound is lowered. When the noise level is low, the sound pressure level of the interfering sound is increased, so that the level of the interfering sound can be appropriately adjusted according to the surrounding noise conditions.

  According to a second aspect of the present invention, there is provided the voice component of the call partner and the jamming signal collected by the microphone after being emitted from the speaker using the voice signal of the call partner to which the jamming signal is added. A removal means for removing a sound component from the collected sound signal, and the collected sound signal from which the voice component and the disturbing sound component of the other party are removed are input, and corresponds to a voice time interval that is the voice of the caller A caller voice change amount measuring means for measuring a time change amount of the collected sound signal as a caller voice change amount, and a voice signal transmitted from the call partner's call device are input, and the voice of the call partner is input. Further comprising: a call partner voice change amount measuring unit that measures a time change amount of the size of the voice signal corresponding to the voice time interval as a call partner voice change amount, wherein the disturbing signal level control unit includes: When the speaker voice change amount and / or the other party voice change amount is larger than a predetermined value, the magnitude of the disturbing signal is increased, and the talker voice change amount and / or the other party voice change amount The gist is to reduce the size of the interference signal when the value is smaller than a predetermined value.

  In the present invention, the sound collection signal from which the voice component and the interference sound component of the other party are removed is input, and the amount of time change in the magnitude of the sound collection signal corresponding to the voice time interval that is the voice of the caller Is measured as the amount of change in the caller's voice, and the voice signal transmitted from the other party's call device is input, and the amount of time change in the volume of the voice signal corresponding to the voice time interval that is the voice of the other party is called When the caller voice change amount and / or the call partner voice change amount is larger than a predetermined value, the size of the disturbing signal is increased and the caller voice change amount and / or the call partner voice is measured. When the amount of change is smaller than the predetermined value, it is possible to provide a hands-free call device that appropriately adjusts the level of the disturbing sound according to the call situation in order to reduce the magnitude of the disturbing signal. More specifically, the amount of change in the voice spectrum is evaluated in a pseudo manner from the viewpoint of the amount of information transmitted between the callers, and if the change in the voice spectrum decreases, the amount of information transmitted between the two parties will be reduced, making the call difficult. Since the sound pressure level of the interfering sound is lowered by assuming that it has become, it is possible to appropriately adjust the size of the interfering sound according to the situation of the call.

  Further, in the present invention, using the voice signal of the other party to which the disturbing signal is added, the voice component and the disturbing sound component of the other party collected by the microphone after being emitted from the speaker are extracted from the collected signal. Therefore, a hands-free communication device with simplified processing can be provided. Specifically, a function (a function corresponding to the above-described echo canceler EC) for preventing the voice of the other party from the speaker from being sent back to the other party, and a disturbance emitted from the speaker. It is possible to share a function (a function corresponding to the above-described adaptive filter ANC) that prevents a sound from reaching the microphone and being sent back to the other party.

  According to a third aspect of the present invention, there is provided a hands-free call method that is provided with a speaker and a microphone, and is performed by a hands-free call device capable of making a call with a call partner without being held by a caller. Generating a jamming signal that prevents the voice of the other party from being heard by a person other than the calling party; and adding the jamming signal to the voice signal transmitted from the other party's calling device; A voice signal based on the voice signal and a jamming sound based on the jamming signal are emitted from the speaker, and a voice collecting signal based on the sound picked up by the microphone is input to correspond to a non-speech time period that is not a call voice Measuring the magnitude of the collected sound signal as a noise level, and inputting the collected sound signal, and the collected sound signal corresponding to a voice time interval that is a call voice Measuring the magnitude as a call sound level; calculating a difference between the noise level with respect to the call sound level; and reducing the magnitude of the jamming signal when the difference is smaller than a predetermined value; And a step of increasing the magnitude of the interference signal when the difference is larger than the predetermined value.

  According to a fourth aspect of the present invention, there is provided the voice component of the call partner and the jamming signal collected by the microphone after being emitted from the speaker using the voice signal of the call partner to which the jamming signal has been added. Removing the sound component from the collected sound signal; inputting the collected sound signal from which the voice component of the other party and the disturbing sound component have been removed; and corresponding to a voice time interval that is the voice of the caller Measuring the amount of time change in the magnitude of the collected sound signal as the amount of change in the caller's voice, and inputting the audio signal transmitted from the call device of the other party, corresponding to the voice time interval that is the voice of the other party Measuring the amount of time change in the size of the voice signal as a call partner voice change amount, and when the caller voice change amount and / or the call partner voice change amount is larger than a predetermined value, A step of increasing the size of the jamming signal and reducing the size of the jamming signal when the amount of change in voice of the caller and / or the amount of change in voice of the other party of the call is smaller than the predetermined value. This is the gist.

  According to the present invention, it is possible to provide a hands-free call device and a hands-free call method capable of generating a disturbing sound with a simple configuration and appropriately adjusting the level of the disturbing sound in accordance with ambient noise and a call situation. it can.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  FIG. 1 is a functional block diagram showing functional blocks of the hands-free call device according to the present embodiment. This hands-free call device includes a speaker 7 and a microphone 8 and can talk to a call partner even if the caller does not hold it in his / her hand. In FIG. 1, the voice input terminal 1 is a call device (FIG. This is an audio terminal for inputting an audio signal transmitted from a device (not shown). The audio output terminal 2 is an audio terminal that outputs a collected sound signal based on a voice of a caller picked up by the microphone 8 or a noise such as noise generated around the caller to a call device of a call partner. . The disturbing signal generation source 3 generates a disturbing signal as a disturbing sound that prevents the other party from listening to the other party's voice emitted from the speaker 7. Examples of the interfering sound include white noise and sound having a spectrum structure different from human-generated sound such as brook noise. The switch 4 is a switch for turning on / off the input of the interference signal generated by the interference signal generation source 3 to the mixer 5 at the subsequent stage.

  The mixer 5 is a terminal that mixes each signal from the voice input terminal 1 and the interference signal generation source 3, and more specifically, a voice signal transmitted from the other party's call device via the voice input terminal 1. In addition, the interference signal generated by the interference signal generation source 3 is added to the speaker 7 to emit the sound based on the voice signal of the communication partner and the interference sound based on the interference signal. The amplifier 6 amplifies the magnitude of the voice signal and jamming signal output from the mixer 5. The speaker 7 emits the voice and jamming sound of the other party based on the voice signal and the jamming signal amplified through the amplifier 6 to the caller. The microphone 8 collects sound such as noise of the caller and the noise generated around the caller.

  The echo canceller 9 is a circuit that cancels out the voice and disturbance sound of the other party that is emitted from the speaker 7 and input to the microphone 8. More specifically, the echo canceller 9 is a circuit that is output from the mixer 5 and added with the disturbance signal. The other party's voice signal is inputted, and the voice component and the disturbing sound component of the other party who are picked up by the microphone 8 after being emitted from the speaker 7 are removed from the collected sound signal.

First audio power calculation unit 10 receives the collected sound signal collected by the microphone 8, to calculate the sound power a _n for time intervals with the waveform of the sound collection signal, the speech power calculated The decibel value _pn is calculated from the logarithmic value. First, the speech power _{a n,} for example, waveform sampling frequency 8kHz of the speech signal, A / D (analog → digital) with 16bit linear waveform converted data (discrete sampling _{waveform) {w m} (m =} 1 , 2,..., 8000) can be obtained by dividing the waveform into units of frames each having a frame length of 1 second and a frame period of 0.5 seconds, squared, and summing up the waveforms (formula (1)).

  Here, f is the number of samplings in one frame (corresponding to the above 8000), and n is an arbitrary frame number.

Next, the decibel value p _n of the audio power a _n, for example, determine the logarithm to the base 10 with respect to speech power a _n, it can be obtained by multiplying 10 (the formula (2) ).

  The second voice power calculation unit 11 and the third voice power calculation unit 18 have the same function, but the signal input to the second voice power calculation unit 11 is sent to the voice component of the other party by the echo canceller 9. And the collected sound signal from which the disturbing sound component has been removed, and the signal input to the third voice power calculation unit 18 is a voice signal transmitted from the other party's call device. Is different.

Non-speech time interval determining section 12, whether not it contains the audio collected sound signal collected by the microphone 8 using the decibel value p _n of the calculated voice power with the first audio power calculation unit 10 performs discrimination, discriminates the non-speech time interval does not include audio outputs only decibel value p _n of the sound power corresponding to the non-speech time interval the average non-voice power calculating section 14 of the subsequent frame by frame . Here, as a method for determining the presence or absence of speech, for example, the power spectrum of a signal of 100 Hz to 300 Hz corresponding to the pitch frequency of speech is observed, and this observation value corresponds to the entire band (corresponding to the sampling frequency described above, If it has a certain magnitude with respect to the voice power of 100 Hz to 4 kHz here, it is determined that the time is the voice time interval in which the voice is input, and has the magnitude. If not, it is possible to use a method for determining that the point in time is a non-speech time period that is not speech. Further, as a more advanced estimation method, cepstrum analysis is performed on each voice, a high cefency region corresponding to a component from 100 Hz to 300 Hz is observed, and voiced sound (voice time interval) and unvoiced sound (non-voice time interval) It is also possible to use a method of distinguishing between

  On the other hand, the first voice time interval discriminating unit 13 uses the decibel value of the voice power calculated by the second voice power calculation unit 11 and the echo canceller 9 removes the voice component and the disturbing sound component of the other party. Discriminates whether or not the sound signal contains sound, determines the sound time interval in which the caller's sound is included, and only the decibel value of the sound power corresponding to this sound time interval is shown in frame units. To the average voice power calculation unit 15 and the first voice dynamic information calculation unit 17. Further, in the second voice time interval discriminating unit 19, voice is included in the voice signal transmitted from the other party's call device using the voice power decibel value calculated by the third voice power calculation unit 18. Whether or not the voice of the other party is included is discriminated, and only the decibel value of the voice power corresponding to this voice time interval is determined in frame units as the second second voice dynamic information. It is sent to the calculation unit 20.

  The average non-speech power calculation unit 14 accumulates the decibel value of the speech power corresponding to the non-speech time interval transmitted from the non-speech time segment determination unit 12 for a certain number of frames, and obtains the average value and outputs it as a noise level. To do. For example, as described above, when the audio power output from the first audio power calculator 10 is a value in units of 0.5 second frame period, it can be obtained as an average value of audio power for 15 seconds, that is, 30 frames. it can. Further, the average voice power calculation unit 15 calculates the average value using the decibel value of the voice power corresponding to the voice time interval transmitted from the first voice time interval determination unit 13 and obtains it as the call sound level. Output. The non-speech time interval determination unit 12 and the first sound time interval determination unit 13 distribute the non-speech time interval and the sound time interval, respectively, so the average non-speech power calculation unit 14 and the average speech power calculation unit 15 Then, there is a possibility that the sound power for 30 frames may not be accumulated reliably. In that case, the sound power accumulated and calculated before that may be used instead.

  The difference calculation unit 16 calculates a difference between the noise level output from the average non-speech power calculation unit 14 and the call sound level output from the average sound power calculation unit 15, that is, the difference between the noise level with respect to the call sound level. This is calculated and sent to the interference signal level control unit 22 in the subsequent stage.

  The first voice dynamic information calculation unit 17 determines how much the decibel value of the voice power corresponding to the voice time interval transmitted from the first voice time interval determination unit 13 varies in units of frames. Calculate as For example, a logarithmic voice power regression coefficient known as “delta power” can be used as the voice feature quantity representing such a change amount. The delta power becomes a small value when the sound changes gently, and becomes a large value when the sound changes drastically (the calculation formula of the delta power will be described later). The first voice dynamic information calculation unit 17 sends a value obtained by squaring this value and calculating the sum to the interference signal level control unit 21 at the subsequent stage. Further, the second voice dynamic information calculation unit 20 calculates the call partner voice change amount using the decibel value of the voice power corresponding to the voice time interval transmitted from the second voice time interval determination unit 19. .

  The jamming signal level control unit 21 controls the jamming signal level controller 22 based on outputs from the difference calculation unit 16, the first voice dynamic information calculation unit 17, and the second voice dynamic information calculation unit 20. Specifically, when the difference calculated by the difference calculation unit 16 is smaller than a predetermined value, the size of the jamming signal is reduced, and when the difference is larger than the predetermined value, the size of the jamming signal is increased. . Further, the jamming signal level control unit 21 increases the magnitude of the jamming signal when the caller voice change amount and / or the call partner voice change amount is larger than a predetermined value, and the caller voice change amount and / or the call. When the other party voice change amount is smaller than the predetermined value, the size of the interference signal is reduced.

  The interference signal level controller 22 is an attenuator connected between the switch 4 and the mixer 5, and outputs an interference signal generated by the interference signal generation source 3 based on the control from the interference signal level control unit 21. It is possible to change the gain.

  Here, the relationship between each functional block of the hands-free call device described in the present embodiment and each means of the hands-free call device described in the claims will be described. The interference signal generation source 3 corresponds to the interference signal generation means described in the claims, and the mixer 5 corresponds to the interference signal addition means described in the claims. The voice time interval discriminating unit 12 and the average non-speech power calculating unit 14 correspond to the noise level measuring means described in the claims. The second voice power calculating unit 11, the first voice time interval discriminating unit 13, the average voice The power calculation unit 15 corresponds to a call sound level measurement unit described in the claims, and the second voice power calculation unit 11, the first voice time interval determination unit 13, and the first voice dynamic information calculation unit 17 are patented. The third voice power calculation unit 18, the second voice time interval determination unit 19, and the second voice dynamic information calculation unit 20 correspond to the caller voice change amount measuring unit described in the claims. Caller voice change listed The difference calculation unit 16 corresponds to the amount measurement unit, the difference calculation unit 16 corresponds to the difference calculation unit described in the claims, and the interference signal level control unit 21 corresponds to the interference signal level control unit described in the claims. The echo canceller 9 corresponds to the removing means described in the claims.

  In the present embodiment, the second voice power calculation unit 11 and the first voice time interval determination unit 13 are common as the call sound level measurement unit and the caller voice change amount measurement unit described in the claims. Of course, it is possible to provide the second voice power calculation unit 11 and the first voice time interval discrimination unit 13 respectively corresponding to these two means.

  Next, the operation of the hands-free call device having the above function will be described. FIG. 2 is a flowchart showing a processing flow of the hands-free call device according to the present embodiment. Note that it is assumed that the voice input terminal 1 and the voice output terminal 2 are connected to the other party's communication device, and the gain of the interference signal level controller 22 is not reduced (step S100).

  The audio signal transmitted from the call device of the other party is released from the speaker 7 through the mixer 5 and the amplifier 6 to the caller's environment as the other party's voice. Now, when the caller turns on the switch 4, the disturbing signal generated from the disturbing signal generation source 3 passes through the disturbing signal level controller 22, the mixer 5, and the amplifier 6 as a disturbing sound from the speaker 7. To be released. Here, if the caller does not intentionally speak, the other party's voice and disturbance sound emitted from the speaker 7 are affected by the acoustic characteristics and ambient noise in the caller's environment and are applied to the microphone 8. Entered. If the waveform of the collected sound signal picked up by the microphone 8 is not extremely distorted, the synthesized sound of the other party's voice and the disturbing sound is convoluted by the digital filter into the output from the mixer 5 in the echo canceller 9. It is possible to cancel the sound to a certain level by the reverse phase synthesized sound obtained by performing the above. Sounds that cannot be canceled out include the voice of the caller, the ambient noise around the caller, the prediction error of the synthesized sound caused by the accuracy of the digital filter of the echo canceller, and the electricity generated when the microphone 8 collects the sound. This is a distortion component due to noise and excessive volume. Here, it is assumed that electrical noise and distortion components can be ignored at a much lower level than the caller's real voice and ambient noise. Then, the output of the echo canceller 9 is mainly a synthesized sound of the caller's real voice and ambient noise. The output of the echo canceller 9 is sent to the voice output terminal 2 as it is, and the other party may hear some feedback of his / her voice, but mainly the voice of the caller and the ambient noise of the caller are relatively clear. You can listen to audio. Hereinafter, the collected sound signal before being input to the echo canceller 9 is referred to as a pre-removed collected signal including a voice component, a noise component, a caller's voice component, and a caller's ambient noise component. The collected sound signal after being input to the canceller 9 is referred to as a post-removal collected sound signal including the voice component of the caller and the ambient noise component of the caller, from which the voice component and noise component of the other party have been removed. .

At this time, the first sound power calculation unit 10 uses the pre-removed sound collection signal before being input to the echo canceller 9, for example, as frame data with a frame length of 1.0 seconds and a frame period of 0.5 seconds. The voice power is calculated and its decibel value is obtained (step S101). Here, the time calculated by the first sound power calculation unit 10, that is, the decibel value of the sound power at the frame _n is defined as pn.

Then, the non-speech time interval determining section 12, only the p _n from p _n determined in a non-speech time interval frames discriminated, and sends the average non-voice power calculating section 14 (step S102).

The average non-speech power calculation unit 14 uses the _pn sent from the non-speech time interval determination unit 12 to calculate an average non-speech power P having a fixed time length for the non-speech time interval (step S103). That is, _assuming that _pn is a logarithmic power obtained every frame period of 0.5 seconds, a method for calculating the average non-speech power P for the non-speech time interval calculated by the average non-speech power calculation unit 14 is, for example, If the average value in units of 30 frames and 15 seconds is to be calculated, it can be calculated using equation (3).

Similarly, the second sound power calculation unit 11 calculates the sound power using the post-removal collected sound signal after being input to the echo canceller 9, and obtains the decibel value (step S104). Here, the time calculated by the second audio power calculator 11, that is, the decibel value of the audio power at the frame _n is defined as q _n .

Then, the first audio time interval discriminating unit 13 discriminates only q _n of the frame in the non-audio time interval from the obtained q _n and sends it to the average audio power calculating unit 15 and the first audio dynamic information calculating unit 17. (Step S105).

The average speech power calculation unit 15 uses the q _n sent from the first audio time section determination unit 13 calculates the average sound power Q of a certain time length for the speech time interval (step S106). The average voice power Q can be calculated using Expression (4) in the same manner as Expression (3).

  Note that the average non-speech power P and the average speech power Q pass through the process of discriminating the non-speech time interval and the speech time interval, and therefore cannot be obtained simultaneously in time. Here, the average non-speech power P in the non-speech time interval calculated by the average non-speech power calculation unit 14 can be regarded as a noise level generated by the ambient noise and the disturbing sound of the caller. Further, the average voice power Q in the voice time interval calculated by the average voice power calculation unit 15 can be regarded as a call sound level generated by a real voice uttered by the caller.

Thereafter, the difference calculation unit 16 calculates the difference value D using Expression (5) (step S107).

  As already described, the average non-voice power P, the average voice power Q, and the difference value D are all decibel values. In a general speech recognition apparatus, when the difference between the average speech input level and the average ambient noise level is about 10 decibels or less, the speech recognition rate is significantly reduced. When considered in a real environment, if the voice input level to the microphone by the real voice has a magnitude equivalent to the equivalent noise level of 76 dBA, the noise in the surrounding environment corresponds to 66 dBA (= 76 dBA-10 dBA). It can be seen that the noise is considerably large because it is almost equivalent to the noise regulation level in a road having two lanes. If this is assumed to be the same as the sound pressure level at which eavesdropping becomes difficult, that is, if the difference value D is a small value of 10 decibels or less, it is difficult for a third party to eavesdrop on, but also adversely affects the surroundings. Suppresses the disturbing sound as an enormous sound.

Next, the first audio dynamic information calculation unit 17, the delta power r _n at frame n is calculated using equation (6).

The delta power r _n is the amount of change talker speech understanding, large value changes more significant is obtained. On the other hand, when the sound approaches a steady sound, even if the sound has a high sound pressure level, it tends to be a small numerical value.

Further, the first audio dynamic information calculation unit 17, the squared value of the delta power r _n, determine the long-term average value R of the predetermined time (step S108). Assuming that the fixed time is the average value in units of 15 seconds as described above, the long-time average value R can be obtained by Expression (7).

Similarly, the third voice power calculating section 18 calculates the speech power by using the audio signal of the other party, obtain the decibel value s _n (step S109).

Similarly, the second audio time segment discriminating unit 19 discriminates only s _n frames of the speech time interval from the logarithmic power obtained is sent to the second audio dynamic information calculation section 20 (step S110).

Then, second audio dynamic information calculation section 20 is calculated using equation (8) the delta power u _n at frame n.

Furthermore, the second audio dynamic information calculation unit 20, the squared value of the delta power u _n, determining the long-term average value U for a predetermined time (step S111).

  The long-time average value R and the long-time average value U are sent to the interference signal level control unit 21 in units of 15 seconds when the average value is calculated. Then, the interference signal level control unit 21 squares the difference value D obtained from the difference calculation unit 16 and the delta power obtained from the first sound dynamic information calculation unit 17 and the second sound dynamic information calculation unit 20. Using the long-time average value R and the long-time average value U as the sum, for example, the following control is performed in units of 15 seconds.

  First, a case where the difference value D is 10 decibels or less will be described. As already described, since the difference value D corresponds to the decibel value of the signal-to-noise ratio between speech and ambient noise, a small value of the difference value D means that there is considerable ambient noise. is doing. Therefore, for example, when the difference value D is 10 decibels or less (step S112), the interfering signal level control unit 21 decreases the gain of the interfering signal level controller 22 in units of −3 decibels, for example, from the interfering signal generation source 3. The output of the disturbing signal to be emitted is lowered (step S113).

  Next, a case where the difference value D is larger than 10 decibels but the sum of the long time average value R and the long time average value U is small will be described. The sum of the long-time average value R and the long-time average value U is compared with the threshold value X. Here, the threshold value X is a positive value, and is a value derived from analysis conditions such as the analysis frame length and frame period of the audio signal. The small value of the sum means that in the conversation between the caller and the other party, the clarity of the voice has decreased, neither of them has been actively discussing, or one of them has spoken unilaterally. It is thought that it corresponds to either. The clarity of speech may be attributed to the acoustic characteristics of the caller's voice, but the disturbing sound may hinder the thinking of the caller and the other party. Therefore, in order to eliminate such a possibility, when the difference value D is smaller than the threshold value X (step S114), the interfering signal level control unit 21 performs the interfering signal level in units of, for example, -3 decibels as in step S113. The gain of the controller 22 is reduced, and the output of the interference signal emitted from the interference signal generation source 3 is lowered.

  Thereafter, the interference signal level control unit 21 measures the difference value D, the long-time average value R, and the long-time average value U again as a result of lowering the output of the interference signal in step S113, and the difference value D is 10 decibels or more. Or whether the sum of the long-time average value R and the long-time average value U exceeds a certain value X. If no change is observed, step S113 is repeated, and the output of the interference signal emitted from the interference signal generation source 3 is further lowered. On the other hand, when the sum of the long-term average value R and the long-term average value U exceeds a certain value X, the jamming signal level control unit 21 increases the gain of the jamming signal level controller 22 in units of, for example, +3 decibels, The output of the disturbing signal emitted from the disturbing signal generation source 3 is restored (step S115). However, it is assumed that the interference signal level controller 22 does not perform gain amplification exceeding a maximum of 0 dB.

  The hands-free call device can output a disturbing sound having an appropriate volume by repeating Steps S112 to S115.

  Finally, when the caller ends the hands-free call, if the caller turns off the switch 4 (step S116), the gain applied by the interference signal level controller 22 is restored.

  In this embodiment, the dynamic feature amount of the voice is calculated using the delta power. However, if there is a margin in the calculation amount, the delta cepstrum representing the dynamic feature of the voice more accurately than the delta power is used. It is also possible to capture more precise changes in speech.

  Furthermore, although the speaker that emits the voice of the other party is the same as the speaker that emits the disturbing sound, the mixer 5 can be eliminated and the voice and the disturbing sound of the other party can be emitted from different speakers. . This is because, in principle, the echo canceller 9 can cancel both the voice of the communication partner and the disturbing sound that have entered the microphone 8.

  In the present embodiment, the case where the difference value D is 10 decibels or less has been described in the comparison in step S112. However, the comparison value for the difference value D is not limited to 10 decibels, and other comparison values are used. It may be used.

  Furthermore, in the present embodiment, the sum of the long-time average value R and the long-time average value U has been described in the comparison in step S114. However, depending on the value of the constant value X, not only the addition. , Multiplication, division, and subtraction may be used, and a predetermined mathematical formula may be used. Further, either one of the long-term average values may be used.

  Finally, when considering “adjusting the level of the disturbing sound in accordance with the ambient noise” described as the first problem, it is possible to discriminate between the call voice and the ambient noise emitted by the other party and the caller. Therefore, it should be noted that the above problem can be solved without removing the voice of the communication partner by the echo canceller 9, in other words, without having the echo canceller 9.

  According to the present embodiment, a disturbing signal that prevents the other party's voice being heard from the speaker 7 from being heard by a person other than the calling party is generated, and the voice signal transmitted from the other party's calling apparatus is generated. Since the interference signal is added and the sound based on the speech signal of the other party and the interference sound based on the interference signal are emitted from the speaker 7, it is possible to provide a hands-free communication device that generates the interference sound with a simple configuration. Specifically, since the speaker 7 that emits the voice of the call partner and the speaker 7 that emits the interference sound are combined into one, it is possible to prevent an increase in the size of the housing of the hands-free communication device and to generate the interference sound. The arrangement of a new speaker 7 is not required, and it can be easily applied to the case of a conventional hands-free call device.

  Further, according to the present embodiment, a sound collection signal based on the sound collected by the microphone 8 is input, and the magnitude of the sound collection signal corresponding to a non-voice time period that is not a call voice is measured as a noise level. At the same time, a sound pickup signal based on the sound picked up by the microphone 8 is input in the same manner, and the magnitude of the sound pickup signal corresponding to the voice time interval that is the call voice is measured as the call sound level, When the difference from the noise level is smaller than a predetermined value, the size of the interfering signal is reduced, and when the difference is larger than the predetermined value, the size of the interfering signal is increased. It is possible to provide a hands-free communication device that appropriately adjusts the level of the communication. Specifically, the ambient noise is evaluated in a time interval in which the caller or the call partner does not make a sound, and if a sufficient noise level is present relative to the sound level, the sound pressure level of the interference sound is lowered. When the noise level is low, the sound pressure level of the interfering sound is increased, so that the level of the interfering sound can be appropriately adjusted according to the surrounding noise conditions.

  Furthermore, according to the present embodiment, the collected sound signal from which the voice component and the disturbing sound component of the other party are removed is input, and the magnitude of the collected sound signal corresponding to the voice time interval that is the voice of the caller is The time change amount is measured as the caller voice change amount, and the voice signal transmitted from the call device of the other party is input, and the time change of the magnitude of the voice signal corresponding to the voice time interval which is the voice of the other party When the amount of change in the other party's voice is measured and the amount of change in the other party's voice and / or the amount of change in the other party's voice is larger than a predetermined value, the magnitude of the disturbing signal is increased, and Since the magnitude of the disturbing signal is reduced when the amount of change in the other party's voice is smaller than the predetermined value, it is possible to provide a hands-free call device that appropriately adjusts the level of the disturbing sound according to the call situation. More specifically, the amount of change in the voice spectrum is evaluated in a pseudo manner from the viewpoint of the amount of information transmitted between the callers, and if the change in the voice spectrum decreases, the amount of information transmitted between the two parties will be reduced, making the call difficult. Since the sound pressure level of the interfering sound is lowered by assuming that it has become, it is possible to appropriately adjust the size of the interfering sound according to the situation of the call.

  Furthermore, according to the present embodiment, the voice component and the disturbing sound component of the call partner collected by the microphone 8 after being emitted from the speaker 7 using the voice signal of the call partner to which the jamming signal has been added. Since it is removed from the collected sound signal, a hands-free communication device with simplified processing can be provided. Specifically, the function of preventing the other party's voice emitted from the speaker 7 from turning around the microphone 8 and being sent back to the other party, and the disturbing sound emitted from the speaker 7 wraps around the microphone 8. Thus, it is possible to share the function of preventing the call from being sent back to the other party.

It is a functional block diagram which shows the functional block of the hands-free call apparatus which concerns on this Embodiment. It is a flowchart which shows the processing flow of the hands-free call apparatus which concerns on this Embodiment. It is a perspective view of the conventional videophone. It is a perspective view of the present videophone.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Audio | voice input terminal 2 ... Audio | voice output terminal 3 ... Interference signal generation source 4 ... Switch 5 ... Mixer 6 ... Amplifier 7 ... Speaker 8 ... Microphone 9 ... Echo canceller 10 ... 1st audio | voice power calculation part 11 ... 2nd audio | voice power calculation Unit 12: Non-speech time interval discriminating unit 13 ... First voice time segment discriminating unit 14 ... Average non-speech power calculation unit 15 ... Average speech power calculation unit 16 ... Difference calculation unit 17 ... First voice dynamic information calculation unit 18 ... Third voice power calculation unit 19 ... second voice time interval determination unit 20 ... second voice dynamic information calculation unit 21 ... jamming signal level control unit 22 ... jamming signal level controller S100 to S116 ... step

Claims (4)

In a hands-free call device that has a speaker and a microphone, and that allows the caller to talk to the other party without holding it,
Interference signal generating means for generating an interference signal that prevents the voice of the other party released from the speaker from being heard by a person other than the speaker;
An interference signal adding means for adding the interference signal to the audio signal transmitted from the communication device of the other party, and emitting the sound based on the audio signal of the other party and the interference sound based on the interference signal from the speaker;
A noise level measurement means for inputting a collected sound signal based on the sound collected by the microphone and measuring the magnitude of the collected sound signal corresponding to a non-voice time interval that is not a call voice as a noise level;
A call sound level measuring unit that inputs the sound pickup signal and measures the magnitude of the sound pickup signal corresponding to a voice time interval that is a call voice as a call sound level;
A difference calculating means for calculating a difference between the noise level and the call sound level;
A jamming signal level control means for reducing the magnitude of the jamming signal when the difference is smaller than a predetermined value and increasing the magnitude of the jamming signal when the difference is larger than the predetermined value;
A hands-free communication device comprising: Using the voice signal of the other party to which the jamming signal has been added, the voice component and the jamming sound component of the other party that have been emitted from the speaker and collected by the microphone are removed from the collected signal. Removal means;
The collected sound signal from which the voice component and the disturbing sound component of the other party are removed is input, and the time change amount of the collected signal corresponding to the voice time interval that is the voice of the caller is determined. Caller voice change amount measuring means for measuring the voice change amount;
The other party who inputs the voice signal transmitted from the other party's calling device and measures the amount of time change of the voice signal corresponding to the voice time interval which is the voice of the other party as the other party voice change. Voice change amount measuring means,
The jamming signal level control means increases the magnitude of the jamming signal when the caller voice change amount and / or the call partner voice change amount is larger than a predetermined value, and the caller voice change amount and / or The hands-free call device according to claim 1, wherein when the call partner voice change amount is smaller than the predetermined value, the interference signal is reduced. In a hands-free call method that is performed with a hands-free call device that includes a speaker and a microphone and can be called with a call partner even if the caller does not hold it in his / her hand,
Generating a jamming signal that prevents the voice of the other party emitted from the speaker from being heard by a person other than the caller;
Adding the jamming signal to a voice signal transmitted from the call device of the other party, and emitting sound based on the voice signal of the other party and jamming sound based on the jamming signal from the speaker;
Inputting a collected sound signal based on the sound collected by the microphone, and measuring the magnitude of the collected sound signal corresponding to a non-voice time interval that is not a call voice as a noise level;
Inputting the collected sound signal and measuring the magnitude of the collected sound signal corresponding to a voice time interval that is a call voice as a call sound level;
Calculating a difference between the noise level relative to the call sound level;
Reducing the magnitude of the jamming signal when the difference is smaller than a predetermined value, and increasing the magnitude of the jamming signal when the difference is larger than the predetermined value;
A hands-free calling method characterized by comprising: Using the voice signal of the other party to which the jamming signal has been added, the voice component and the jamming sound component of the other party that have been emitted from the speaker and collected by the microphone are removed from the collected signal. Steps,
The collected sound signal from which the voice component and the disturbing sound component of the other party are removed is input, and the time change amount of the collected signal corresponding to the voice time interval that is the voice of the caller is determined. Measuring the amount of change in voice;
Inputting a voice signal transmitted from the call partner's call device, and measuring a time change amount of the voice signal corresponding to a voice time interval that is the voice of the call partner as a call partner voice change amount; ,
When the caller voice change amount and / or the call partner voice change amount is larger than a predetermined value, the size of the jamming signal is increased, and the caller voice change amount and / or the call partner voice change amount is Reducing the magnitude of the jamming signal if less than the predetermined value;
The hands-free call method according to claim 3, further comprising:

Images