CN1538718A - Method and system of automatic preventing talking listened by another microphone - Google Patents

Abstract

Self-adapting noise controller is utilized to reduce volume of voice sent from first location by someone, and the voice can be heard at second location without intention. At first location, referenced mike picks up voice and feeds the voice back to a controller. Through self-adapting grid type filtering algorithm, the controller generates sound attenuation signal. The sound attenuation signal causes speaker at third location to send second sound, which cancels out voice of the person at second location. Error mike picks up residual noise of first and second voices at second location. The error microphone provides error signal as feedback for the controller. Volume of second voice can be picked up by referenced mike at first location.

Description

The method and system that the active speech is offset

Technical field

The present invention relates to the initiatively method and system of speech counteracting.The present invention is specially adapted to, but is not limited to, and the active speech in the zones such as office is offset, and someone call can be heard in described zone, perhaps sneaks in the call of other people in the equivalent environment.

Background technology

When people talk each other or face-to-face or by equipment such as phones.The problem that runs into during speech is that speech is not unidirectional, and often spreads on all directions.Therefore, people often in the face of its speech by around other people in straitened circumstances situation of being not intended to hear, do not wish that its speech is not intended to hear by other people this moment or do not wish the interference of being talked as other people.Typical example is in open working environment, and in one of them compartment someone can not avoid being not intended to hear the dialogue of next door compartment, no matter whether the compartment owner wishes so.In less formal environment, for example in public's traffic environment, identical problem is arranged also, one of them passenger uses mobile phone that its passenger on every side is fed up with.

Initiatively noise cancellation is known technology, and its target is by introducing next " counteracting " the undesirable sound of sound field that additional electronics mode produces.This technology can limit the sound wave in certain frequency range, perhaps weakens the sound wave in a certain distance.Though before the thinking of eliminating the noise can be traced back to 60 years, yet coml realizes in fact only limiting to the very special situation that some only needs to offset single-frequency or narrow-band noise.These are used and mainly concentrate on man-made noise control field.For example, people know that use has the 50Hz narrow band signal counteracting noise of transformer of harmonic wave.

Summary of the invention

In this specification and claims, term " comprises " or similar terms is used to represent comprising of non-exclusionism, makes the method or the device that comprise one group of unit not only comprise those unit, and can comprise other unlisted unit.

According to an aspect of the present invention, provide a kind of method, be used to reduce level from first sound primary importance, that can on the second place, hear away from primary importance.This method comprises: produce the base sound signal and produce the noise elimination signal according to the base sound signal in primary importance according to first sound, send in the 3rd position corresponding to second sound of noise elimination signal and at the combination results error voice signal of the second place according to first and second sound that receive.Also produce the noise elimination signal, make and offset first sound that receives in the second place at least in part at second sound of second place reception according to the error voice signal.First sound is the wideband sound of frequency range in 50 to 10000HZ, and preferably speech or music.

According to a second aspect of the invention, provide a kind of method, be used to reduce level from first sound primary importance, that can hear on away from the second place of primary importance.This method comprises: produce the noise elimination signal in the 3rd position and send second sound corresponding to the noise elimination signal; At the combination results error voice signal of the second place according to first and second sound that receive.The generation of giving birth to the noise elimination signal makes and offsets first sound that receives in the second place at least in part at second sound of second place reception based on the error voice signal.First sound is the wideband sound of frequency range in 50 to 10000HZ, and preferably speech or music.

According to a third aspect of the invention we, provide a kind of system, be used to reduce level from first sound primary importance, that can hear on away from the second place of primary importance.This system comprises: be used for receiving the base sound signal and producing the voice offset controller of noise elimination signal according to the base sound signal according to the benchmark microphone of the first sound generating base sound signal with from the benchmark microphone in primary importance.This system also comprises from the voice offset controller and receives the noise elimination signal and send control loudspeaker corresponding to second sound of noise elimination signal in the 3rd position.In addition, it is included in the error microphone of the second place according to the combination results error voice signal of first and second sound that receive.The voice offset controller is configured to produce the noise elimination signal according to the error voice signal, makes to offset first sound that receives in the second place at least in part at second sound of second place reception.First sound is the wideband sound of frequency range in 50 to 10000HZ, and preferably speech or music.

According to a forth aspect of the invention, provide a kind of system, be used to reduce system from the level of first sound primary importance, that can hear on away from the second place of primary importance.This system comprises the voice offset controller that is used to produce the noise elimination signal and receives the noise elimination signal and send control loudspeaker corresponding to second sound of noise elimination signal in the 3rd position from the voice offset controller.It also is included in the error microphone of the second place according to the combination results error voice signal of first and second sound that receive.The voice offset controller can produce the noise elimination signal according to the error voice signal, makes to offset first sound that receives in the second place at least in part at second sound of second place reception.First sound is the wideband sound of frequency range in 50 to 10000HZ, and preferably speech or music.

According to a fifth aspect of the invention, provide a kind of adaptive voice offset controller.Controller comprises filter with filtering algorithm and the adaptive algorithm processor of adjusting filtering algorithm.This controller also comprises: be used to receive the benchmark input from the reference signal that derives from first sound of primary importance; The output offset signal is to produce the counteracting output of offsetting sound, to pass through filter filtering input reference signal to produce offseting signal; Import with receiving from first sound and the error of the error signal of the combination derivation of offsetting sound.The adaptive algorithm processor can be adjusted filtering algorithm according to reference signal and error signal.First sound is the wideband sound of frequency range in 50 to 10000HZ, and preferably speech or music.

According to a sixth aspect of the invention, provide a kind of software that comprises code, in a single day this code is loaded into processor and just allows processor to serve as the controller of the 5th embodiment.

For example, the present invention has covered a kind of like this self-adapting noise controller, and it is used to reduce the volume of speech that someone sends in primary importance, that can be heard unintentionally in the second place.A benchmark microphone picks up speech and is fed to controller in primary importance.Controller produces the noise elimination signal by adaptive lattice type filtering algorithm, and the noise elimination signal causes the 3rd locational loud speaker to send will be used for offsetting someone speech of the second place second sound.The error microphone picks up the residual noise of first and second sound on the second place, provides error signal as feedback to controller.The volume of second sound also can be picked up by the benchmark microphone of primary importance.

Description of drawings

Understand for convenience and enforcement the present invention, describe exemplary preferred embodiment referring now to accompanying drawing, wherein:

Fig. 1 is based on the schematic diagram of structure of the active voice bucking-out system of one embodiment of the present of invention;

Fig. 2 is the system block diagram of Fig. 1 active voice bucking-out system;

Fig. 3 is the schematic diagram of Fig. 2 system part;

Fig. 4 is the detailed maps of Fig. 3 filter;

The flow chart of Fig. 5 shows the generation that filtering returns signal;

Fig. 6 represents to use different initiatively algorithms to carry out the comparative result that speech is offset;

Fig. 7 represents the active counteracting of multitone; And

The figure of Fig. 8 represents the active counteracting of speech.

Embodiment

In institute's drawings attached, same Reference numeral is used to represent identical unit.

With reference to Fig. 1, wherein diagram shows the schematic diagram based on the structure of the embodiment of the invention.Facing to the speech of the phone on its desktop, produce first spoken sounds 14 (in the conversation on course he this side) the people 10 (first sound source) of primary importance 12 therebetween.Lineup 16 at the second place 18 places is positioned at the H.D of the sound that can hear the people 10 who talks facing to phone.Second sound source of control loudspeaker 20 forms is positioned on the 3rd position 22, and produces the second counteracting sound 24 of the guiding second place 18.3 positions away from each other, though second compare more approaching each other with primary importance with the 3rd position.

The first benchmark microphone 26 is placed in primary importance 12, in this case, is positioned at phone inside, though also can be positioned at other position, for example on the table.It produces reference signal according to first sound 14.This reference signal is received by adaptive voice offset controller 28.The second error microphone 30 is placed in the second place 18, and according to speech that receives in the second place 18 and the combination results error signal of offsetting sound.Error signal also is passed to adaptive voice offset controller 28.Loud speaker 20 produces according to the noise elimination signal and offsets sound 24, and wherein said noise elimination signal is produced according to benchmark and error signal by adaptive voice offset controller 28.

In this structure, from people 10 the spoken sounds 14 arrival second places 18, the others through air, echo and environment in stroke are attenuated.The target of system is in the second place 18 counteractings or reduces this sound, makes crowd 16 can not hear it, perhaps makes reduced by its degree of bothering.This identical spoken sounds 14 is also picked up in unattenuated mode by benchmark microphone 26 at primary importance 12 places, and is passed to the adaptive voice offset controller 28 that produces the noise elimination signal.This causes loud speaker 20 to send counteracting sound 24, offsets sound 24 and also arrives the second place 18, and be attenuated betwixt.

At the second place 18 places, the speech of error microphone 30 receiving attenuations and counteracting sound, and according to residual noise generation error signal.This is fed back to adaptive voice offset controller 28, offsets so that obtain optimum speech and adaptive voice offset controller 28 is adjusted the noise elimination signal.Ideally, offset sound and spoken sounds and can on the second place 18, offset fully each other, thus not from error microphone 30 to adaptive voice offset controller 28 return error signals.

Fig. 2 is the block diagram of system of the active voice bucking-out system of Fig. 1, and as shown in Figure 2, this structure of Fig. 1 can be regarded as the some signals that pass different passages by approximate, and these passages are figure signal by different way.

Spoken sounds 14 can be considered the first signal s (n) with time index n that people 10 produces.The decay of spoken sounds 14 is equivalent to first signal and passes the first sonic propagation passage 32 between first and second positions, and its effect is equivalent to have impulse response function H ₁(z) first filter.So the spoken sounds that arrives the second place is elementary noise, elementary noise can be considered elementary noise signal d (n), wherein

d(n)＝H ₁(z)*s(n)

(* is a convolutional calculation).

Offset sound 24 and can be considered the secondary signal c (n) that loud speaker 20 sends.The 3rd and the second place between offset sound 24 decay be equivalent to secondary signal and pass the second sound wave propagation ducts 34, its effect is equivalent to have impulse response function H ₂(z) second filter.So the counteracting sound that arrives the second place is secondary noise, secondary noise can be considered secondary noise signal g (n), wherein:

g(n)＝H ₂(z)*c(n)

Elementary noise and secondary noise are picked up by error microphone 30 and are primary and secondary noise signal d (n) and g (n), to produce error signal e (n), wherein:

e(n)＝d(n)+g(n)＝H ₁(z)*s(n)+H ₂(z)*c(n)

Be fed to controller 28.

Benchmark microphone 26 also picks up the spoken sounds 14 of first signal s (n) form.Yet it is not the sound of unique arrival benchmark microphone 26.In addition, be similar to the mode that spoken sounds 14 arrives the second place 18, some counteracting sound from loud speaker 20 arrives primary importance 12 and is picked up by benchmark microphone 26.

The 3rd and primary importance between offset sound 24 decay be equivalent to secondary signal and pass the 3rd sonic propagation passage 36, its effect is equivalent to have impulse response function H ₃(z) the 3rd filter.So the counteracting sound that arrives the 3rd position is third level noise, third level noise can be considered third level noise signal t (n), wherein:

t(n)＝H ₃(z)*c(n)

Spoken sounds and third level noise are picked up as the first signal s (n) and third level noise signal t (n) by benchmark microphone 26 respectively, to produce reference signal x (n), wherein:

x(n)＝s(n)+t(n)＝s(n)+H ₃(z)*c(n)

Be forwarded to controller 28.Because it is the attenuated versions to the response of the attenuated versions of s (n), the level of t (n) is usually much smaller than the level of s (n).

Controller 28 produces noise elimination signal y (n), and noise elimination signal y (n) is issued as offsetting sound 24, is equivalent to secondary signal c (n).Owing to do not have other input of arriving loud speaker 20, therefore

c(n)＝y(n)

When producing noise elimination signal y (n), the purpose of controller 28 is the elementary noise signal d (n) that offset the 3rd position.So the sound that arrives the 3rd position from loud speaker should be inverted (reverse) of d (n) with identical magnitude.Therefore, elementary noise signal d (n) can be known as desired signal.If realize eliminating the noise fully, then error signal e (n) can be zero.Yet,, when signal changes in time, can not reach fully and eliminate the noise, as in dialogue owing to when controller receives and handle reference signal x (n), always have some time lags.Therefore target is to keep e (n) minimum.

In Fig. 3, illustrate in greater detail controller 28.It comprises sef-adapting filter 42, and this filter 42 is its input with reference signal x (n) and has impulse response H _c(z).Reference signal x (n) also is diverted in the forward path model 44, and the effect of forward path model 44 is equivalent to have impulse response function H ' ₂(z) filter, impulse response function H ' ₂(z) be similar to impulse response function H in the forward path 34 ₂(z).Therefore the output p (n) of forward path model 44 is:

p(n)＝H′ ₂(z)*x(n)

It is imported into adaptive algorithm processor 46.Error signal e (n) also is imported into processor 46, and processor 46 produces adaptive to the filter factor in the sef-adapting filter 42 according to p (n) and e (n).

Usually at the problem definition sef-adapting filter such as the active noise cancellation, its median filter output is the estimation to desired signal.Yet in control was used, sef-adapting filter served as controller, and control comprises the dynamical system of exciter and amplifier etc.In this case, estimation (antinoise or rp-wave) can be considered from dynamical system, promptly from the output signal g (n) of forward path 34.

Present embodiment uses lattice type adaptive IIRs (impulse response immediately) filtering algorithm.Its advantage is at forming noise by the arrowband component with more high-power difference (for example voice signal), can obtaining restraining relatively fast.The normal gradients lattice type algorithm (for example: P.A.Regalia describes at " the Adaptive IIR Filtering in SignalProcessing and Control " that published by Marcel Dekker company) that is used for adaptive lattice type IIR filtering is subject to large-scale calculations and memory load requirement, and this makes that the realization in real-time system is difficult and expensive more.

In traditional lattice type iir filter, suppose that the filtering regression coefficient that obtains corresponding to rotation parameter needs M additional lattice filter, then for calculating and storage, the magnitude of complexity is M ²Yet other possible active noise controlling algorithm (for example: FXLMS, FULMS and FVLMS) all only needs the M rank to calculate and storage, and is easy to realize by dsp program.

Along tap parameter { v with optimization _kThe reduction error level check that the filtering corresponding to rotation parameter returns signal { θ _kCharacteristic after, can derive the part gradient algorithm of M rank complexity.Derivation is similar to the derivation of finding in the works of above-mentioned Regalia, does not wherein consider and offsets the path transfer function.

Therefore, present embodiment uses the shortcut calculation of adaptive lattice type form IIR filter structure, and its convergence result is better than traditional FULMS and FVLMS, and increases calculating and memory load.

Fig. 4 illustrates in greater detail the filter 42 of Fig. 3.Though normally M rank filter shown in the figure is M=3 in this embodiment.Check-in signal passes through some convolutional filters in proper order, and then returns in opposite direction.When returning, between convolutional filter, extract signal, and signal parallel is by some tap filters, the accumulation result signal is to provide output y (n).

Algorithm in the adaptive algorithm processor 46 is determined filter factor, constantly adjusts and upgrades them.Related algorithm is:

v _k(n+1)＝v _k(n)-μe(n)·B _k(z)H′ ₂(z)×(n)，k＝0，1，...，M

θ _k(n+1)＝θ _k(n)-μ _e(n)·γ _kB _k-1(z)H _c(z)H′ ₂(z)×(n)k＝0，1，...，M

Ideally, can use H ₂(z), but actual so impossible, so use H ' ₂(z).

Fig. 5 shows the flow chart that the necessary filtering of explanation returns the generation of signal, and corresponding algorithm is as follows.

Input parameter:

The rank M of lattice filter

Upgrade step size mu

Suppose the counteracting path transfer function H of iir filter form ' ₂(z) has feedforward weight a _W0..., a _WNWith feedback weight b _W1..., b _WL

Initialization:

All filter factors and state are configured to 0.

Can obtain at time n:

Filter factor:

Tap parameter v _k(n), k=0,1 ..., M;

Rotation parameter θ _k(n), k=1 ..., M.

Filter status:

b _k(n-1)，k＝0，1，...，M-1；

Post-filter states:

x(n-k)，k＝1，...，N；

c(n-k)，k＝1，...，M；

y _c(n)；

b _ck(n-1)，k＝0，1，...，M-1；

b _yk(n-1)，k＝0，...，M-1.

New data:

X (n) (reference signal)

E (n) (error signal)

Lattice filter calculates:

Make f _M(n)=x (n)

To k=M, M-1 ..., 1 does 0

$\left[\begin{array}{c}{f}_{k-1}\left(n\right)\\ b_k\left(n\right)\end{array}\right]=\left[\begin{array}{cc}{\cos \mathrm{\θ}}_k\left(n\right)& {-\sin \mathrm{\θ}}_k\left(n\right)\\ {\sin \mathrm{\θ}}_k\left(n\right)& {\cos \mathrm{\θ}}_k\left(n\right)\end{array}\right]\left[\begin{array}{c}{f}_k\left(n\right)\\ b_{k-1}(n-1)\end{array}\right]$

Finish

·b ₀(n)＝f ₀(n)，

Lattice filter output:

$y\left(n\right)=\underset{k=0}{\overset{M}{\mathrm{\Σ}}}{b}_k\left(n\right){\mathrm{\υ}}_k\left(n\right)$

Postfilter calculates:

Make f _CM(n)=c (n), wherein

$c\left(n\right)=\underset{i=0}{\overset{N}{\mathrm{\Σ}}}{a}_{\mathrm{wi}}\left(n\right)x(n-i)+\underset{j=1}{\overset{L}{\mathrm{\Σ}}}{b}_{\mathrm{wi}}\left(n\right)c(n-j)$

To k=M, M-1 ..., 1 does

$\left[\begin{array}{c}{f}_{\mathrm{ck}-1}\left(n\right)\\ b_{\mathrm{ck}}\left(n\right)\end{array}\right]=\left[\begin{array}{cc}{\cos \mathrm{\θ}}_k\left(n\right)& {-\sin \mathrm{\θ}}_k\left(n\right)\\ {\sin \mathrm{\θ}}_k\left(n\right)& {\cos \mathrm{\θ}}_k\left(n\right)\end{array}\right]\left[\begin{array}{c}{f}_{\mathrm{ck}}\left(n\right)\\ b_{\mathrm{ck}-1}(n-1)\end{array}\right]$

Finish

·b _c0(n)＝f _c0(n)。Corresponding to v _kFiltering to return signal be b _Ck(n).

Filter returns:

Make γ M=1.

To k=M, M-1 ..., 1 does

Corresponding to θ _kFiltering return signal:

b _θk(n)＝-γ _kb _yk-1(n)

γ _k-1＝γ _kcosθ _k(n)

Finish

Filter factor upgrades:

v _k(n+1)＝v _k(n)-μe(n)b _ck(n)

θ _k(n+1)＝θ _k(n)-μe(n)b _θk(n)

Test:

To k=1 ..., M does

If | θ _k(n+1) |＞pi/2 is established θ _k(n+1)=θ _k(n)

Finish

Postfilter calculates:

$y_c\left(n\right)=\underset{k=0}{\overset{M}{\mathrm{\Σ}}}{b}_{\mathrm{ck}}\left(n\right){\mathrm{\υ}}_k\left(n\right)$

Make f _YM(n)=y _c(n), wherein

To k=M, M-1 ..., 1 does

$\left[\begin{array}{c}{f}_{\mathrm{yk}-1}(n+1)\\ b_{\mathrm{yk}}(n+1)\end{array}\right]=\left[\begin{array}{cc}{\cos \mathrm{\θ}}_k(n+1)& {-\sin \mathrm{\θ}}_k(n+1)\\ {\sin \mathrm{\θ}}_k(n+1)& {\cos \mathrm{\θ}}_k(n+1)\end{array}\right]\left[\begin{array}{c}{f}_{\mathrm{yk}}\left(n\right)\\ b_{\mathrm{yk}-1}(n-1)\end{array}\right]$

Finish

·b _y0(n)＝f _y0(n)。

Though this is the algorithm of revising at adaptive lattice type iir filter, yet can use canonical algorithm, calculation requirement is higher.

A possibility of adaptive algorithm is to use reticle type form adaptive IIR filtering algorithm.Another possibility is by the algorithm based on LMS, and the LMS algorithm of filtering x for example, this algorithm are suitable for ACTIVE CONTROL and use.It is from the LMS algorithm development, dynamic system model between its median filter output y (n) and the estimation g (n), be between the algorithm of forward path 34 coefficient vector that is introduced in input signal x (n) and matched filter 42, wherein the filter 42 for the LMS algorithm of filtering x is FIR filters.

This algorithm does not use convolutional filter.Filter factor upgrades just:

vk(n+1)＝vk(n)+μp(n)e(n)

Because forward path model 44 uses the impulse response estimation of forward path, the output p (n) of forward path model 44 is approximations, and the difference between forward path estimation and the true forward path influences the stability and the rate of convergence of algorithm.Yet algorithm is strong to the error of forward path estimation.

Can introduce on the basic frequency time delay corresponding to forward path.

Optional algorithm based on LMS comprises FULMS and FVLMS.Yet lattice type algorithm is preferred.Fig. 6 provides the simplification gradient lattice type algorithms (rolling off the production line) of suggestion, the simulation comparative result between FVLMS algorithm (center line) and the FVLMS algorithm (reaching the standard grade).The x axle represents to calculate number of iterations, and the y axle is represented the dB amplitude.

Result according to Fig. 6 can find, simplifies the convergence of algorithm of gradient lattice type and is better than traditional adaptive IIR filtering algorithm.Note, simplify gradient lattice type algorithm and not only when control and treatment begins, restrain fast, and when control and treatment restrains, have more excellent offset result.

Fig. 7 shows the offset result of multi-tone signal (having big eigenvalue distribution).Row (1) is an initialize signal, and row (2) is to use the offset result of simplifying lattice type gradient algorithm, and go (3) be to use the offset result of FVLMS algorithm.Fig. 8 shows the offset result of voice signal (having big eigenvalue distribution).Row (1) is an initialize signal, and row (2) is to use the offset result of simplifying lattice type gradient algorithm, and go (3) be to use the offset result of FVLMS algorithm.In Fig. 7 and Fig. 8, the x axle is represented the iterations that calculates, and the y axle represents to return the amplitude of instrument absolute value.According to the result, when reference signal had big eigenvalue distribution, the problem of slow convergence appearred in traditional adaptive control algorithm, and the derivation of this and those algorithm quite meets.Yet because the prewhitening characteristic of input lattice structure, lattice type gradient algorithm greatly reduces the sensitivity to the reference signal eigenvalue distribution, so convergence rate is very fast.In addition, lattice type gradient algorithm will be calculated and memory load is reduced to the M rank owing to simplify, and then not have too big difficulty aspect the dsp program realization.

Of the present invention extensive aspect, controller can be the feedback or feedforward controller, and can be the numeral or the simulation.It can be programmable, for example is by using the DSP of software implementation algorithm and processing.Advantageously, can for voice signal, be generally 50 to 10,000Hz stablizing in the frequency range on a large scale.

Optional embodiment can be provided, wherein not have the benchmark microphone, so signal x (n) is not fed to controller 28.Provide the approaching of elementary noise signal d (n), rather than reference signal x (n) is provided to sef-adapting filter 42 and forward path model 44 as input with input as sef-adapting filter 42 and forward path model 44.Deriving this elementary noise according to the error signal e (n) that adds up with approaching of secondary noise signal g (n) or secondary noise signal approaches.This secondary noise is approached and can be caused secondary signal c (n) shunting and by forward path model (being similar to forward path model 44).

When the present invention was used to environment such as open office, each compartment can be equipped with a benchmark microphone, an error microphone and a loud speaker.Be determined by experiment between each compartment and the frequency response between error microphone and the loud speaker in each compartment.Do not need various parts are installed in compartment, can on ceiling they be installed, this can reduce the possibility that people hear dialogue each other, even people wish so.If compartment or zone are subjected to the influence from more than noise source of other compartment, the offseting signal that produces can be added together.

In places such as public building or public transport, microphone and loud speaker can be provided with the interval of expectation.Can be determined by experiment frequency response, for example, according to the different degree of crowdings.If locate pick-up noise, can around the calling party, produce suitable noise removing a benchmark calling party.

If people are just using mobile phone, can be with its microphone as the benchmark microphone.Yet may need the voice signal from phone is provided to local noise canceling system, and determine the position of phone at concrete intra-zone easily.

Reduce the possibility that is not intended to hear speech though described embodiment is mainly used in, yet the present invention also can be used for reducing the noise pollution of music and other any wideband sound source of not expecting (for example having the frequency range in 50 to 10000Hz).

Preceding detailed description only provides preferred embodiment, and scope of the present invention, applicability or structure are not produced any restriction.The front detailed description of the preferred embodiment is just in order to enable those skilled in the art to realize the preferred embodiments of the present invention.Should be appreciated that under the prerequisite that does not depart from aim of the present invention that claims limit and scope, can on the function of unit and layout, carry out various changes.

Claims (24)

1. a method is used to reduce the level from one first a sound primary importance, that can hear on the second place away from this primary importance, and this method comprises:

In described primary importance according to the described first sound generating base sound signal;

Produce the noise elimination signal according to described base sound signal;

Send one second sound corresponding to the noise elimination signal in one the 3rd position; With

At the combination results error voice signal of the described second place according to first and second sound that receive;

Wherein also produce described noise elimination signal, make and offset first sound that receives in this second place at least in part at second sound of described second place reception according to the error voice signal; And

Described first sound is the wideband sound of frequency range in 50 to 10000Hz.

2. use adaptive filter algorithm to produce the noise elimination signal according to the process of claim 1 wherein.

3. according to the method for claim 2, wherein adaptive filter algorithm has self adaptation tap and coefficient of rotary.

4. according to the method for claim 2, wherein adaptive filter algorithm is an adaptive lattice type filtering algorithm.

5. according to the method for claim 2, wherein adaptive filter algorithm is the IIR filtering algorithm.

6. can hear second sound according to the process of claim 1 wherein in primary importance; And according to the second sound generating base sound signal of hearing in primary importance.

7. in phone, produce the base sound signal according to the process of claim 1 wherein.

8. according to the process of claim 1 wherein that described first sound is music and/or voice.

9. according to the process of claim 1 wherein that first, second and the 3rd position are in the office environment.

10. a method is used to reduce the level from one first a sound primary importance, that can hear on the second place away from this primary importance, and this method comprises:

Produce the noise elimination signal in one the 3rd position and send second sound corresponding to the noise elimination signal; With

At the combination results error voice signal of the described second place according to first and second sound that receive;

Wherein produce the noise elimination signal, make and offset first sound that receives in this second place at least in part at second sound of second place reception according to the error voice signal; And

Described first sound is the wideband sound of frequency range in 50 to 10000Hz.

11. a system is used to reduce the level from one first a sound primary importance, that can hear on the second place away from this primary importance, this system comprises:

At the benchmark microphone of described primary importance according to the described first sound generating base sound signal; With

Receive the base sound signal and produce the voice offset controller of noise elimination signal according to described base sound signal from the benchmark microphone;

Receive the noise elimination signal and send control loudspeaker in one the 3rd position from the voice offset controller corresponding to one second sound of noise elimination signal; With

At the error microphone of the second place according to the combination results error voice signal of first and second sound that receive;

Wherein the voice offset controller is configured to produce the noise elimination signal according to the error voice signal, makes to offset first sound that receives in this second place at least in part at second sound of second place reception; And

Described first sound is the wideband sound of frequency range in 50 to 10000Hz.

12. according to the system of claim 11, wherein the voice offset controller is configured to use adaptive filter algorithm to produce the noise elimination signal.

13. according to the system of claim 12, wherein adaptive filter algorithm has self adaptation tap and coefficient of rotary.

14. according to the system of claim 12, wherein adaptive filter algorithm is an adaptive lattice type filtering algorithm.

15. according to the system of claim 12, wherein adaptive filter algorithm is the IIR filtering algorithm.

16., wherein can hear second sound in described primary importance according to the system of claim 11; And the benchmark microphone can be according to described first sound of hearing in described primary importance and the described second sound generating base sound signal.

17. according to the system of claim 11, wherein the benchmark microphone is in phone.

18. according to the system of claim 11, wherein said first sound is music and/or voice.

19. according to the system of claim 11, wherein first, second is in the office environment with the 3rd position.

20. a system is used to reduce the level from one first a sound primary importance, that can hear on the second place away from described primary importance, this system comprises:

Produce the voice offset controller of noise elimination signal;

Receive the noise elimination signal and send control loudspeaker in one the 3rd position from the voice offset controller corresponding to second sound of noise elimination signal; With

At the error microphone of the described second place according to the combination results error voice signal of first and second sound that receive;

Wherein said voice offset controller can produce the noise elimination signal according to the error voice signal, makes to offset described first sound that receives in this second place at least in part at second sound of described second place reception; And

Described first sound is the wideband sound of frequency range in 50 to 10000Hz.

21. the adaptive voice offset controller comprises:

Filter with filtering algorithm;

Be used to adjust the adaptive algorithm processor of described filtering algorithm;

Receive benchmark input according to the reference signal that derives from first sound of a primary importance;

The output offset signal wherein passes through filter filtering input reference signal to produce offseting signal to produce the counteracting output of offsetting sound; With

Reception is imported according to the error of the error signal that the combination of first sound and counteracting sound is derived;

Wherein said adaptive algorithm processor can be adjusted filtering algorithm according to reference signal and error signal; And

Described first sound is the wideband sound of frequency range in 50 to 10000Hz.

22. according to the controller of claim 21, its median filter is an adaptive lattice type iir filter.

23. comprise the software of code, in a single day described code is loaded into processor, just allows processor to serve as controller, this controller comprises:

Filter with filtering algorithm;

Adjust the adaptive algorithm processor of described filtering algorithm;

Receive benchmark input according to the reference signal that derives from first sound of a primary importance;

The output offset signal wherein passes through filter filtering input reference signal to produce offseting signal to produce the counteracting output of offsetting sound; With

Reception is imported according to the error of the error signal that the combination of first sound and counteracting sound is derived;

The adaptive algorithm processor can be adjusted filtering algorithm according to reference signal and error signal; And

First sound is the wideband sound of frequency range in 50 to 10000Hz.

24. according to the software of claim 23, its median filter is an adaptive lattice type iir filter.

Images