CN102957819B - Method and apparatus for processing audio signals - Google Patents
Method and apparatus for processing audio signals Download PDFInfo
- Publication number
- CN102957819B CN102957819B CN201210368101.5A CN201210368101A CN102957819B CN 102957819 B CN102957819 B CN 102957819B CN 201210368101 A CN201210368101 A CN 201210368101A CN 102957819 B CN102957819 B CN 102957819B
- Authority
- CN
- China
- Prior art keywords
- shaper
- audio signal
- echo
- received
- equipment
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04M—TELEPHONIC COMMUNICATION
- H04M9/00—Arrangements for interconnection not involving centralised switching
- H04M9/08—Two-way loud-speaking telephone systems with means for conditioning the signal, e.g. for suppressing echoes for one or both directions of traffic
- H04M9/082—Two-way loud-speaking telephone systems with means for conditioning the signal, e.g. for suppressing echoes for one or both directions of traffic using echo cancellers
Abstract
The utility model discloses a method, a device, and a computer program product which are used for processing audio signals at the device department. The device comprises an audio frequency output component for outputting audio signals. The audio signals are received at the multiple microphones of the device. At least one characteristic of the audio signals received by the multiple microphones is measured. A beam shaper applies the beam shaper coefficient over the received audio signals, thereby generating the output of the beam shaper. An echo-removing device is applicable to the output of the beam shaper, thereby preventing the echoes resulting from the audio signals output by the audio frequency output component. The operating parameters of the echo removing device are controlled based on at least one of the measured characteristics of the audio signals received by the multiple microphones.
Description
Technical field
The audio signal that the present invention relates to receiving at equipment place processes.
background technology
Equipment can have the audio frequency entering apparatus of such as microphone and so on, and it can be used to from surrounding environment received audio signal.Such as, the microphone of subscriber equipment can receive audio signal (such as from the speech of user) and other audio signal.Other audio signal described may be the interference tones signal received at the microphone place of equipment, and may receive from interference source, or may be the self noise of ambient background noise or microphone.Interference tones signal may disturb the audio signal received at equipment place.Institute's received audio signal can be used for many different purposes by equipment.Such as, when received audio signal is the voice signal from user's reception, this voice signal can be undertaken processing to use in communication event by equipment, such as, by this voice signal being sent to through network another equipment that can be associated with another user of communication event.Alternatively or in addition, the audio signal received can be used to other purposes as known in the art.
In order to improve, received audio signal (such as, from user receive the voice signal for using in a call) quality, need to suppress the interference tones signal received at the microphone place of subscriber equipment (such as, background noise and the interference tones signal that receives from interference tones source).
The use that stereophony microphone and wherein multiple microphone carry out as single audio frequency entering apparatus other microphone array operated becomes more prevalent.Except the information extracted the audio signal that the multiple microphone of equipment use makes except receiving from single microphone, the spatial information extracted from institute's received audio signal can also be used.When using such equipment, a kind of for suppressing the method for interference tones signal to be the audio signal application of beam former received multiple microphone.Beam forming carries out concentrating to make the specific audio signal received from one or more desired position (that is, direction and distance) at microphone array compare the process promoted to some extent at the remainder of microphone institute received audio signal to the audio signal that microphone array receives by application signal transacting.In order to briefly, be described to the situation only with single desired direction here, but identical method can be applied when there is more how interested direction.Before beam forming row relax, can determine or be arranged on the angle (and/or distance) that microphone array receives desired audio signal, i.e. so-called arrival direction (DOA) information.Estimation due to arrival direction may be complicated, so it can be favourable for being set to by desired arrival direction fixing.But in the situation substituted, the condition making desired arrival direction be adapted to change can be favourable, and therefore when using beam-shaper, to perform the estimation of desired arrival direction in real time may be favourable.Adaptive beamforming device applies multiple weight (or " beam-shaper coefficient ") to received audio signal.These weights can be able to adaptation and process to be formed " wave beam " to consider DOA information to the audio signal that multiple microphone receives, from desired position (namely high-gain is applied to microphone thus, desired direction and distance) the desired audio signal that receives, low gain is then applied to the direction going to other (such as, disturbing) signal source any.Can carry out with regard to adaptive meaning with regard to the suppression of interference source, beam-shaper can be " adaptive ", and the selection of desired source/view direction then needs not to be can be adaptive.
Equipment not only has the multiple microphones for received audio signal, can also have the audio output device (such as, comprising loud speaker) for output audio signal.Such equipment is such as useful when audio signal such as will be output to equipment user and receive from equipment user during communication event.Such as, equipment can be the subscriber equipment of such as phone, computer or TV, and can comprise allow user participate in the necessary device of videoconference.
Audio output device is comprised (such as at equipment, comprise loud speaker) and audio frequency entering apparatus is (such as, microphone) the two when, then in institute's received audio signal, there is echo is frequent existing problems, and wherein echo results from and to export and in the audio signal of microphones from loud speaker.Echo Canceller can be used to the echo in the audio signal eliminated microphone place and receive.Echo is suppressed and echo reduction is the two kinds of methods implementing Echo Canceller.Such as, Echo Canceller can implement echo suppressor, and it is used to suppress the echo in the audio signal received at microphone.The propagation path of audio signal from loud speaker to microphone is referred to as echo path, and echo suppressor can estimate echo path gain as the function of time and the echo power using it to estimate in received audio signal.Echo power in institute's received audio signal estimates the level that can be used to echo being suppressed in received signal be made them can not be noted when there is any near end signal (not deriving from the signal component of loud speaker in the audio signal that microphone receives).Echo power in the audio signal received estimates it is the model carrying out loud speaker-closed-microphone (loudspeaker-enclosure-microphone) system operated based on Echo Canceller wherein.This model is often linear at least in part, but in some cases, this model may be nonlinear.Mixing Echo Canceller cuts down device by the echo of applying with hierarchal manner and echo suppressor formed.By using mixing Echo Canceller, cutting down device by echo and achieving both-end voice (doubletalk) transparency increased, and if need, then obtain extra echo by echo suppressor and suppress gain.
General Requirements for the Optimum Operation of echo cancellor is:
Echo path changes relatively lentamente, this is because otherwise echo path gain estimates will be inaccurate very soon;
System is enough linear, to carry out modeling by linear echo model; With
Echo path gain should not be underestimated, this is because underestimate and then echo power also can be caused to be underestimated.This will cause Echo Canceller to apply too small suppression and let slip very important residual echo thus.
Beam-shaper is implemented to institute received audio signal and Echo Canceller is not inappreciable task.In fact, when combining adaptive microphone beam former (such as, in conference call application), need to carry out paying close attention to make the performance of Echo Canceller can not reduce by the adaptation in beam forming.
In the first system implemented together with echo cancellor by beam forming, before execution beam forming, independent Echo Canceller is applied to each microphone signal.But this first system is computationally very complicated due to the operation of the multiple Echo Cancellers for multiple microphone signal.In addition, microphone signal is used to the beam forming process of Echo Canceller (particularly using the Echo Canceller implemented echo and suppress) meeting interfering beam former.
In the second system implemented together with echo cancellor by beam forming, Echo Canceller is applied to the output of beam-shaper.In this second system; the performance of data adaptive beam-shaper is preferably constrained to passes in time and changes very lentamente; this is because otherwise; when Echo Canceller is attempted to estimate to regulate to echo path in response to the change of beam-shaper performance, the accuracy that the echo path used in the echo suppression that Echo Canceller is implemented is estimated will affect adversely.In addition, in this second system, beam-shaper is preferably confined to linear so that the echo preventing Echo Canceller from implementing suppresses the unfavorable decline of attainable echo cancellation performance in near-end transparency.Some beam-shapers are linear, but some are not then linear, thus the selection of beam-shaper is restricted (being restricted to linear wave beam shaper) in second system.
Therefore, the first system described above and second system all have problems.
Summary of the invention
Embodiments of the invention allow microphone beam former to use together with Echo Canceller, and and never in any form on the type of beam-shaper or performance limits and do not affect the performance of Echo Canceller.Embodiments of the invention calculate Echo Canceller performance based on microphone signal the strongest before beam-shaper, and export application Echo Canceller to beam-shaper.
Term " the strongest " audio signal can represent the institute's received audio signal passed in time and have most high-average power.But " the strongest " audio signal can represent the audio signal received of passing in time and having the highest absolute value by mean value, or instantaneous there is the most superpower audio signal.
According to a first aspect of the invention, provide a kind of method processed audio signal at equipment place, this equipment comprises the audio output device for output audio signal, and the method comprises: at multiple microphone places received audio signal of this equipment; Measure the characteristic of at least one in the audio signal that multiple microphone receives; Beam-shaper, to received audio signal application of beam former coefficient, generates beam-shaper thus and exports; And application echo cancellor device is exported to this beam-shaper, the echo of suppressing the audio signal exported from this audio output device to cause from this beam-shaper exports thus, measured by least one wherein in the audio signal that receives based on multiple microphone of the operating parameter of this echo cancellor device, characteristic controls.
Advantageously, the computation complexity of the embodiment of the present invention is low, and reason is that echo cancellor device is applied to each the accepted audio signal before beam-shaper output instead of beam-shaper.Therefore required Echo Canceller is less.In a preferred embodiment, only use an Echo Canceller, it is applied to beam-shaper and exports.In addition, the performance of beam-shaper can't be subject to the adverse effect of echo cancellor device, this is because echo cancellor device did not change institute's received audio signal before beam-shaper is to institute's received audio signal application of beam former coefficient.In addition, the performance of beam-shaper is not restricted because beam-shaper exports the requirement of slowly change, this is because before beam-shaper application of beam former coefficient, the operating parameter of echo cancellor device controls based on the characteristic of one or more institutes received audio signal.Like this, the operating parameter of echo cancellor device is not subject to the impact of beam-shaper performance, and beam-shaper output is not required in order to echo cancellor device correctly works and slowly changes like this.Similarly, because the operating parameter of echo cancellor device controlled based on the characteristic of at least one institute's received audio signal before beam forming process, so the beam-shaper parameter applied of beam-shaper need not be confined to be linear (although beam-shaper coefficient can be linear in certain embodiments).
In a preferred embodiment, comprised the strongest the audio signal received according at least one carrying out measurement characteristics in the audio signal that multiple microphone receives, and the method can comprise the step determining that the audio signal which receives is the strongest audio signal received.Can be comprised by according at least one carrying out measurement characteristics the institute's received audio signal being different from the strongest audio signal in the audio signal that multiple microphone receives.
In addition, at least one that can comprise the audio signal received from multiple microphone of step measuring the characteristic of at least one audio signal measures echo path gain.The method may further include and uses described measured echo path gain to determine that the echo power in beam-shaper output is estimated, wherein said operating parameter is estimated to control based on the echo power in beam-shaper output.This operating parameter can comprise echo cancellor device (such as, by the echo suppressor part of echo cancellor device) and beam-shaper be exported to the echo suppression level applied.Advantageously, in a preferred embodiment, estimate to ensure can not to be underestimating echo power in beam-shaper output according to the echo power in exporting from the determined beam-shaper of the echo path gain measured by the strongest audio signal received, thus ensure that echo cancellor device can not export the too small echo of application to beam-shaper and suppress.
In a preferred embodiment, the method comprises the beam-shaper gain determining that beam-shaper is applied from the audio signal that the Main way of beam-shaper receives multiple microphone further, wherein said beam-shaper gain be used in beam-shaper export in describedly determining of estimating of echo power.Determined echo power estimates the upper limit that can represent echo power in beam-shaper output.
The beam-shaper coefficient of analysis to beam-shaper that the method may further include based on institute's received audio signal carries out adaptation.Such as, can to beam-shaper coefficient carry out adaptive with in the beam forming constraint of beam-shaper by the minimum power in beam-shaper output.The beam forming of beam-shaper retrains audio signal constraint of distortion not due to application of beam former coefficient that such as can comprise multiple microphone and receive in the Main way of beam-shaper.Beam-shaper coefficient can describe linear function or the nonlinear function of received audio signal.Such as, beam-shaper can be the undistorted response of minimum variance (MVDR) beam-shaper.
According to a second aspect of the invention, provide a kind of equipment for the treatment of audio signal, this equipment comprises: audio output device, for output audio signal; Multiple microphone, for received audio signal; Beam-shaper, it is configured to received audio signal application of beam former coefficient, generates beam-shaper thus and exports; Echo cancellor device, it is configured to be applied to this beam-shaper and exports, and exports the echo of suppressing the audio signal exported from this audio output device to cause thus from this beam-shaper; And processing apparatus, for measuring in audio signal that multiple microphone receives the characteristic of at least one, and control for the operating parameter of characteristic to this echo cancellor device measured by least one in the audio signal that receives based on multiple microphone.
This echo cancellor device can comprise the echo suppressor part being arranged to and being applied to beam-shaper and exporting.This echo cancellor device may further include to be arranged to cuts down device at beam-shaper to the echo being applied to each audio signal that multiple microphone receives before received audio signal application of beam former coefficient.
According to a third aspect of the invention we, provide a kind of computer program for processing audio signal at equipment place, this equipment comprises for multiple microphone of received audio signal and the audio output device for output audio signal, and the computer-readable medium that this computer program is recorded in non-momentary performs step when being configured to perform on the processor of this equipment with box lunch: measure the characteristic of at least one in the audio signal that multiple microphone receives; Use beam-shaper to received audio signal application of beam former coefficient, generate beam-shaper thus and export; And application echo cancellor device is exported to this beam-shaper, the echo of suppressing the audio signal exported from this audio output device to cause from this beam-shaper exports thus, the characteristic measured by least one wherein in the audio signal that receives based on multiple microphone of the operating parameter of this echo cancellor device controls.
Accompanying drawing explanation
In order to understand the present invention better and illustrate how the present invention can come into force, now will by exemplary reference the following drawings, wherein:
Fig. 1 shows the schematic diagram of the equipment according to preferred embodiment;
Fig. 2 shows the system according to preferred embodiment;
Fig. 3 shows the functional block diagram of the parts of the equipment according to the first prior art systems;
Fig. 4 shows the functional block diagram of the parts of the equipment according to the second prior art systems;
Fig. 5 shows the functional block diagram of the parts of the equipment according to preferred embodiment;
Fig. 6 shows the flow chart of the process for the treatment of audio signal according to preferred embodiment.
Embodiment
To be described the preferred embodiments of the present invention by means of only example now.Such as, may wish to implement beam-shaper at equipment place with both Echo Cancellers to use with conference call application.In following examples of the present invention, beam-shaper type or beam-shaper performance not to be limited together with Echo Canceller uses allowing beam-shaper and the technology not affecting the performance of Echo Canceller is described.
Beam-shaper can carry out modeling as follows.The purposes of microphone beam former merges some microphone signals to export to produce beam-shaper.Beam-shaper exports and can be made up of many signals, but in order to briefly, in following described preferred embodiment, hypothesis beam-shaper is produced one and export separately.There is many different operable beamforming algorithm and universal model for describing beamforming algorithm is
Wherein y (t) is that beam-shaper exports, y
nt () is the n-th input signal from the n-th microphone in microphone array, N is the sum of beam-shaper input signal, and f () is beam-shaper function.Beam-shaper function f () can adopt many one of multi-form and can be linear or nonlinear.For delay stack (Delay-and-sum) beam-shaper, beamforming algorithm provides as follows:
For MVDR beam-shaper, beamforming algorithm provides as follows:
Wherein g
nt () is the weight factor of the n-th input signal.
The Echo Canceller carrying out operating the echo of eliminating in single microphone input signal y (t) can carry out modeling as follows.Echo path is the path that loudspeaker signal x (t) carries out to microphone signal y (t) propagating.The Echo Canceller using echo to suppress can pass through for time t and frequency f estimated echo path gain
, and use the estimation of echo path gain
estimate the echo power in institute's received audio signal
operate.This estimation of echo power can perform according to the linear model of following equation based on the loud speaker-closed-microphone system operating Echo Canceller thereon:
Wherein Y (t, f) be for microphone signal Frequency Estimation (such as, digital fourier transformation), X (t, f) is the Frequency Estimation for loudspeaker signal, S (t, f) be Frequency Estimation for echo signal, H (t, f) is the frequency response of echo path, and N (t, f) is all near-end sound, the Frequency Estimation of microphone noise and modeling error (they are all the audio signals outside received echo).In order to echo cancellor model is effective, suppose that the conversion estimated for calculated rate all has enough rank (sufficient orders).Because what signal understanding will be exported from its loud speaker by equipment, so be known for the Frequency Estimation X (t, f) of loudspeaker signal.Therefore, in order to determine echo power
, should the frequency response H (t, f) in estimated echo path.
Under the hypothesis that X (t, f) and N (t, f) are irrelevant, echo path gain
can estimate according to the model of the function as Y (t, f) and X (t, f).In addition, can based on estimating that the ratio of echo and near-end is to estimating speed or estimates that confidence level regulates, the possible accuracy that the restriction of this ratio is estimated, namely main estimate echo power in microphone signal by force time upgrade estimating.A kind of operable method of estimation is least square estimation method.
Echo path gain is estimated
be used to use above model to carry out estimated echo power
, that is:
The General Requirements of the Optimum Operation of echo cancellor is:
Echo path changes relatively lentamente, this is because otherwise echo path gain estimates will be inaccurate very soon;
System is enough linear to carry out modeling by above linear echo model; With
Echo path gain should not be underestimated, namely
, this is because underestimate and then also can cause echo power
underestimated.This is too small known and let slip very important residual echo thus by causing Echo Canceller to be applied.
With reference now to Fig. 1, which illustrates the explanatory view of equipment 102.Equipment 102 can be fixing or mobile device.Equipment 102 comprises CPU 104, what be attached thereto is microphone array 106 for received audio signal, for the audio output device 110 of output audio signal, such as exporting the display 112 of the screen of vision data and the memory 114 for storing data to the user of equipment 102.
Referring now to Fig. 2, which illustrates the example context 200 that equipment 102 carries out operating wherein.
The microphone array 106 of equipment 102 is from environment 200 received audio signal.Such as, as shown in Figure 2, microphone array 106 receives from user 202(as the d in Fig. 2
1represented) audio signal, from another user 204(as the d in Fig. 2
2represented) audio signal, from fan 206(as the d in Fig. 2
3represented) audio signal, and from loud speaker 210(as the d in Fig. 2
4represented) audio signal.The audio output device 110 of equipment 102 comprises audio frequency output processing device 208 and loud speaker 210.Audio frequency output processing device 208 carries out operating to send audio output signal to export from loud speaker 210 to loud speaker 210.Audio frequency output processing device 208 as the software that CPU 104 performs or can operate as the hardware in equipment 102.It is evident that for those skilled in the art, microphone array 106 can receive other shownschematically outer audio signal of Fig. 2.In the situation shown in Fig. 2, the audio signal from user 202 is desired audio signal, and is interference tones signals in other audio signals all that microphone array 106 receives.In other embodiments, the more than one audio signal received at microphone array 106 can be considered to " desired " audio signal, but in order to briefly, in described embodiment, only there is the audio signal desired by (it is the audio signal from user 202) and other audio signal is all considered to interference here.Other source of less desirable noise signal such as can comprise air-conditioning system, play the equipment of music, and in environment 200, such as leave the reverberation (reverberance) of audio signal of wall.
Referring now to Fig. 3, which illustrates the functional representation of the parts of the first prior art equipment.Microphone array 106 comprises multiple microphone 302
1, 302
2with 302
3.First prior art equipment comprises beam-shaper 304 further.In microphone array 106, the output of each microphone is coupled to multiple acoustic echo suppressor 306
1, 306
2with 306
3corresponding input.By this way, for each received signal (y represented in Fig. 3
1(t), y
2(t) and y
3(t)) there is an Echo Canceller 306.The output of each Echo Canceller 306 is couple to the corresponding input of beam-shaper 304.Those skilled in the art will appreciate that and need multiple input to implement beam forming.As shown in Figure 3, microphone 302
1, 302
2with 302
3the echo audio signal receiving near-end audio signal and export from loud speaker 210.
In the first prior art equipment, before beam-shaper 304 performs beam forming, independent Echo Canceller 306 is applied to each received audio signal y (t).Therefore, as described above, the first prior art equipment shown in Fig. 3 is computationally complicated owing to implementing multiple Echo Canceller 306.In addition, introduce Echo Canceller 306 and can cause interference to the beam forming process of adaptive beamforming device 304.
Referring now to Fig. 4, which illustrates the functional representation of the parts of the second prior art equipment.Similar with the first prior art equipment shown in Fig. 3, the second prior art equipment has and comprises multiple microphone 302
1, 302
2with 302
3microphone array 106 and beam-shaper 404.But in the second prior art equipment, the output of each microphone in microphone array 106 is coupled to the corresponding input of beam-shaper 404, and the output of beam-shaper 404 is coupled to the input of acoustic echo suppressor 404.By this way, beam-shaper 404 application of beam before Echo Canceller 406 carries out echo cancellation process is shaped and processes.Second prior art equipment eliminates the needs to multiple Echo Canceller, but the second prior art equipment has defect.Such as; the action need of beam-shaper 404 carries out regulating slowly to change in the second prior art equipment because otherwise Echo Canceller 406 cannot fast enough for beam-shaper performance change carry out regulating to provide satisfied suppression to the echo in institute's received audio signal.In addition, in order to Echo Canceller 406 proper operation, beam-shaper 404 preferably needs to be linear.Which has limited the selection of beam-shaper 404 execution mode.
Referring now to Fig. 5, which illustrates the functional representation of the parts of equipment 102 according to the preferred embodiment of the invention.Microphone array 106 comprises multiple microphone 302
1, 302
2with 302
3.Equipment 102 comprises beam-shaper 504, acoustic echo suppressor 506 and processing block 508 further.Beam-shaper 504 can be such as the undistorted response of minimum variance (MVDR) beam-shaper.Implement in the software that beam-shaper 504, Echo Canceller 506 and processing block 508 can perform on CPU 104 or the hardware in equipment 102.In microphone array 106, the output of each microphone is coupled to the corresponding input of beam-shaper 504.In microphone array 106, the output of each microphone is also coupled to the corresponding input of processing block 508.The output of beam-shaper 504 is coupled to the input of Echo Canceller 506.The output of processing block 508 is coupled to the input of Echo Canceller 506 to provide side information to Echo Canceller 506 thus.One of ordinary skill in the art would recognize that, need multiple input to implement beam forming.Microphone array 106 is shown to have three microphones (302 in Figure 5
1, 302
2with 302
3), but will be appreciated that, the quantity of this microphone is only example and not limits by any way.
Beam-shaper 504 comprises the microphone 302 for receiving and process from microphone array 106
1, 302
2with 302
3audio signal y
1(t), y
2(t) and y
3the device of (t).Such as, beam-shaper 504 can comprise voice activity detector (VAD) and DOA estimation block (not shown).In operation, beam-shaper 504 determine the audio signal that microphone array 106 receives attribute and based on estimate as VAD and DOA block the text hegemony of quality that detects, determine one or more Main way of (one or more) main loudspeaker.In other embodiments, (one or more) Main way of (one or more) main loudspeaker can pre-set to make beam-shaper 304 concentrate on fixed-direction.In the illustrated example shown in fig. 2, the direction (d of the audio signal received from user 202
1) be confirmed as Main way.Beam-shaper 504 can use DOA information (or use to pre-set simply fixedly check direction for beam-shaper 504) to carry out audio signal by forming wave beam, described wave beam desired by the reception of microphone array 106 place signal from Main way (d
1) direction in there is high-gain, and at direction (such as, the d of any other signal
2, d
3and d
4) in there is low gain.
Beam-shaper 504 can also determine the interference radiating way (d arrived
2, d
3and d
4), and advantageously, the adaptive audio signal received with the arrival direction making gain low especially be applied to disturbing from those can be carried out, to suppress the audio signal disturbed to the performance of beam-shaper 504.Simultaneously, be described above, beam-shaper 504 can determine the Main way of any amount, the quantity of determined Main way affects the attribute of beam-shaper, such as, for a large amount of Main way, compared with only determining the situation of single Main way, the decay that beam-shaper 504 will be less to the signal application received from other (less desirable) direction at microphone array.The output of beam-shaper 504 is provided to Echo Canceller 506 with the form will carrying out the single passage processed.Such as, more than one passage may also be exported to preserve or virtual generation stereo image.As the skilled person will be apparent, the output of Echo Canceller 506 can use in a number of different ways in equipment 102.Such as, the output of Echo Canceller 506 can be used as a part for the communication event that user 202 is just using equipment 102 to participate in.
With reference to figure 6, now the method for the audio signal according to preferred embodiment is described.In step S602, at the microphone (302 of microphone array 106
1, 302
2with 302
3) place's received audio signal.This audio signal such as receives from user 202, user 204 and fan 206, and these signals constitute near-end audio signal as shown in Figure 2 and Figure 5.Microphone in microphone array 106 is also from loud speaker 210 reception of echoes signal as shown in Figure 2 and Figure 5.Other interference tones signal of such as background noise and so on also can at the microphone (302 of microphone array 106
1, 302
2with 302
3) place receives, and these other interference tones signals will form further near-end audio signal.By each microphone (302 of microphone array 106
1, 302
2with 302
3) institute received audio signal y
1(t), y
2(t) and y
3t () is sent to beam-shaper 504 and processing block 508.Echo signal is that audio frequency output processing device 208 sends the result of loudspeaker audio signal x (t) and loud speaker output loudspeaker audio signal x (t) to loud speaker 210.Loudspeaker audio signal x (t) is undertaken propagating by (described by H (t, f)) echo path and is appeared at received audio signal y
1(t), y
2(t) and y
3in (t).
In step S604, beam-shaper 504 is to received audio signal (y
1(t), y
2(t) and y
3(t)) apply its beam-shaper filtration coefficient to generate beam-shaper output thus.As described above, beam-shaper 504 concentrates at microphone array 106 from Main way (d
1) audio signal that receives to strengthen the audio signal desired by receiving from user 202 thus, and suppresses other audio signal application received from other direction at microphone array 106.Beam-shaper 504 is for from main positions (namely, direction and distance) audio signal that receives should have fixed gain, wherein beam-shaper gain should preferably be passed in time is constant (but may not be constant with frequency), to prevent the distortion of desired audio signal thus.Exemplarily, beam-shaper gain can be set to numerical value 1 for main positions (that is, Main way and/or distance).The task of beam-shaper 504 is relative to the signal of other interference source lifting from main source.Beam-shaper exports and is sent to Echo Canceller 506.
In step S606, processing block 508 is to received audio signal y
1(t), y
2(t) and y
3t () carries out analyzing to determine which is institute received audio signal y
1(t), y
2(t) and y
3the strongest in (t).In other embodiments, in step S606, select the signal being different from the strongest audio signal in institute's received audio signal, but select the strongest audio signal in a preferred embodiment.The strongest institute received audio signal y
maxt () can be sent to echo cancellation module 506 as side information from processing block 508.In this sense, before to audio signal application of beam former 504, the strongest audio signal received is determined.Following equation is met to determine the strongest institute's received audio signal by finding out which institute's received audio signal:
In other embodiments, the strongest institute's received audio signal other equation can be determined according to certain.Such as, " the strongest " audio signal can be pass institute's received audio signal with most high-average power in time.Alternatively, " the strongest " audio signal can be average pass institute's received audio signal with the highest absolute value in time, or instantaneously has the most superpower audio signal.
In step S608, measure the echo path gain of the strongest audio signal
.This comprises the estimation finding out echo path gain according to following equation
:
Wherein, as described above, suppose that X (t, f) is uncorrelated with N (t, f) in the estimation.
In step 610, determine the echo suppression level will applied by Echo Canceller 506.This determines the estimation based on echo power in institute's received audio signal
.Echo power is by the echo path gain by the strongest measured in step S608 audio signal
be multiplied by the loudspeaker signal exported from loud speaker 210
and estimate, namely:
By this way, in a preferred embodiment, echo power is based on n institute received audio signal y
nt the echo path gain of the peak signal in () is estimated.Step S608 and S610 can be utilized by processing block 508 and be sent to Echo Canceller 506 and perform as the result of side information.Alternatively, step S608 and S610 one or both of can be performed based at the side information of Echo Canceller 506 from the strongest institute received audio signal that processing block 508 receives by Echo Canceller 506 self.
In step S612, Echo Canceller 506 is applied to beam-shaper and exports, and the suppression level that wherein Echo Canceller 506 is applied is determined in step S610.In step S614, output signal to process further equipment 102 from Echo Canceller 506.Such as, the signal exported from Echo Canceller 506 can use communication event, such as, for as from user 202 to the voice signal of another user by Internet Transmission to another equipment using a part of calling out as audio or video.
Be applied to compared with the model of the Echo Canceller of single institute received audio signal with described above for (when not having beam forming), the Echo Canceller 506 of preferred embodiment is applied to beam-shaper and exports.Echo signal in the audio signal received Echo Canceller 506 place affects by beam forming process to some extent, and therefore echo canceller system model needs to carry out revising can be applicable to beam-shaper output.As described above, each of the audio signal received for n uses following equation:
The signal (Y (t, f)) received at Echo Canceller 506 is the output of beam-shaper 504, and is therefore institute received audio signal (Y
n(t, f)) combination, the wherein beam-shaper algorithm that uses according to beam-shaper of this combination.Namely:
Wherein F () is the frequency domain representation of beam-shaper function f ().
Can recognize that why the second prior art equipment shown in Fig. 4 needs to apply to require so that Echo Canceller 406 successfully eliminates echo to beam forming function from above equation.Especially, in order to system model is linear, (namely the linear combination preferably needing beam-shaper 404 to be exported by microphone formed, f () must be linear function), otherwise it is just no longer applicable to be used for the linear model of echo power of echo power that estimated echo arrester 406 uses, and may need to use nonlinear model, this can reduce the near-end transparency of Echo Canceller 406.In addition, beam-shaper performance need to pass slow change in time because otherwise above system by become during height and Echo Canceller 406 can not will be transparent near end signal due to the more inaccurate estimation of echo path.
As described above, the preferred embodiments of the present invention overcome these problems by carrying out estimated echo power based on the echo path gain of the strongest institute's received audio signal.
1. from the basic assumption after standard echo cancellor and requirement, regardless of the type of beam-shaper, y
max(t, f) and X (t, f) preferably linear correlation (because y
max(t, f) is relevant to audio signal before beam forming process) and therefore this relevant can use linear estimation methods by
be similar to well.Therefore, even if beam-shaper 504 does not use linear function f (), linear model described above also can be used effectively to determine the estimation of echo power
.This to relax the constraint of beam-shaper 504 to make beam forming function f () can be linear or nonlinear.Notice, it is non-linear with what process in echo path that Echo Canceller can comprise nonlinear model.But introducing other non-linear (wherein some may even cannot be modeled) in the echo path thus affects to some extent by the near-end transparency of Echo Canceller.
2. regardless of beam-shaper performance,
time variations should be all identical for the situation of standard echo cancellor.This is because the estimation of the echo path gain of the strongest institute received audio signal
determined before beam forming process.This relaxes can in time slowly or Rapid Variable Design and can not to the estimation of the echo power that Echo Canceller 506 uses to make beam-shaper export to the constraint of beam-shaper 504
cause adverse effect.
3. because the gain of (in a preferred embodiment) echo path is estimated based on the strongest microphone signal, and because the rate variance of the echo power in array between microphone and near-end is only less due to the finite width of microphone array 106, so echo power is estimated
be confined to:
Wherein
be the actual echo power during beam-shaper exports, and B (f) is the constant wave beam shaper gain of beam-shaper 504 to the beam-shaper 504 that the audio signal received in the Main way of beam-shaper 504 is applied.Typically, beam-shaper gain B (f) is selected as 1 to all frequency f.
value be used as beam-shaper export in the estimation of echo power.This is because we know that gain B (f) depending on frequency applied at most by beam-shaper 504 in view direction.Can carry out some constraints to B (f), such as, such as known, the beam-shaper gain for the direction of loud speaker 210 is less than B (f).To being explained as follows of the constraint that echo power is estimated.Beam-shaper 504 relative to from other direction less desirable audio signal audio signal desired by Main way is strengthened.This enhancing can mean desired signal be exaggerated or institute undesirably signal be weakened, or mean both simultaneously.In a preferred embodiment, to the requirement of beam-shaper 504 be the Main way of beam-shaper 504 (namely, " viewing " direction) in the amplification of audio signal should to pass in time be constant, because otherwise required audio signal causes distortion by beam-shaper 504.Therefore, following equation will be set up all the time:
Wherein as described above, Y (t, f) represents that the beam-shaper in frequency domain exports, and Y
n(t, f) then represents institute's received audio signal of the n-th microphone of microphone array 106 in frequency domain.Beam-shaper 504 can be calculated as the mean value of input signal, and B (f)=1 in that case.In practice, B (f) preferably should be not more than 1, because otherwise just cannot ensure the sampling microphone signal of 16 bits in beam-shaper, export to provide the beam-shaper that may be stored in 16 bit-pattern forms (hint does not use Q numerical value to carry out proportion adjustment, in any case this all can make beam-shaper output saturation at it as audio file broadcasting or when storing).But notice, for integrality, can comprise be less than 1 B (f) numerical value and this may be in some cases useful (such as beam-shaper input signal be 24 bits output signal be 16 bit time).In addition, in certain embodiments, B (f) can 1 be greater than and this may be in some cases useful (such as beam-shaper input signal be 16 bits export be 24 bit time, also have when signal demand is exaggerated desired by beam-shaper, and the reason for accuracy is performed in beam-shaper to amplify to be useful).
Due to the estimation of (in a preferred embodiment) echo power
based on the strongest institute's received audio signal, so keep
value at least export with beam-shaper in actual echo power
equally large.Constant wave beam shaper gain B (f) being applied to the beam-shaper 504 of the institute's received audio signal in the Main way of beam-shaper 504 is priori for beam-shaper 504, or can estimate simply in addition.
Therefore, the preferred embodiments of the present invention provide echo power estimation
(wherein B (f) is 1 in some preferred embodiments), its:
1. based on linear model and therefore can use linear echo eliminate function estimate;
2. there is the time variations identical with the Echo Canceller implemented when not implementing beam forming; And
3. the restriction providing echo power is estimated, it is guaranteed to be not less than the echo power in beam-shaper output
, thus ensure that Echo Canceller 506 can't apply too small suppression.
Can notice, ideally, recognize that the simultaneous echo that beam-shaper realizes suppresses to be useful, reason is that this will allow the echo of being applied small amount by Echo Canceller 506 to suppress, this so that Echo Canceller 506 will be made more transparent near end signal.But the beam-shaper array of microphone usually has the microphone of limited quantity and therefore the echo suppression that beam-shaper 504 can realize is limited, thus this not to be included in transparency that consideration causes reduces be less.
In general, in a preferred embodiment of the invention, the performance of Echo Canceller 506 be based on institute's received audio signal application of beam former before multiple institutes received audio signal in peak signal (Y
max(t, f)), Echo Canceller 506 is then applied to the output of beam-shaper 504.
The advantage that the preferred embodiment of the present invention is better than the first prior art equipment shown in Fig. 3 is:
Computation complexity: for the first prior art equipment shown in Fig. 3, each microphone needs to use additional Echo Canceller, even and if some calculating can be shared between Echo Canceller, but each microphone computation complexity in array is obviously increased.On the contrary, it is only less that the complexity required by the preferred embodiment of the present invention improves, and reason is only to need enforcement Echo Canceller (such as, 506).
Be convenient to integrate: for the first prior art equipment shown in Fig. 3, need to carry out main redesign to existing Echo Canceller.On the contrary, in a preferred embodiment of the invention, the amendment of required Echo Canceller is less.
Beam forming performance: for the first prior art equipment shown in Fig. 3, the action need of beam-shaper 304 regulates, and reason is otherwise the Echo Canceller 306 being applied to the input of each beam-shaper can be shaped and the adaptivity of beam forming process by interfering beam.In addition, by introducing echo cancellation performance before beam forming, the transparency that may suppress cannot to be used to improve echo cancellor of the loud speaker echo that beam-shaper 304 realizes.On the contrary, the preferred embodiments of the present invention not in order to use Echo Canceller 506 to perform echo cancellor to the performance of beam-shaper 504 in addition requirement.In addition, any available information that any echo that beam-shaper 504 realizes is suppressed all may be used to improve the Echo Canceller transparency near end signal.
The advantage that the preferred embodiment of the present invention is better than the second prior art equipment shown in Fig. 4 is:
Beam-shaper performance: for the second prior art equipment shown in Fig. 4, the action need of beam-shaper 404 is adjusted to and relatively changes lentamente, because otherwise Echo Canceller 406 is difficult to adjustment to adapt to the performance change of beam-shaper 404 by finding.This adjustment is needed to be because beam-shaper 404 is included in the echo path of Echo Canceller 406 perception.If Echo Canceller 406 cannot suitably regulate to adapt to the change in beam-shaper 404, just may occur that echo is by Echo Canceller 406.In addition, in order to this second prior art equipment correctly works, the change of the performance of beam-shaper 404 must perform during speaker activity, the change that beam-shaper shows can be detected and regulate to adapt to it to make Echo Canceller 406.This requirement significantly limit the performance of beam-shaper 404.On the contrary, the preferred embodiment of the present invention not in order to Echo Canceller 506 performs echo cancellor to the performance of beam-shaper 504 in addition requirement.Therefore, the preferred embodiments of the present invention do not limit the performance of beam-shaper 504 as the second prior art equipment.
Beam-shaper is general: if Echo Canceller 406 will use linear model to echo cancellor, then the beam forming function f () of the beam-shaper 404 of the second prior art equipment shown in Fig. 4 must be linear.On the contrary, in a preferred embodiment of the invention, the beam forming function f () not requiring beam-shaper 504 is linear.In fact, in an embodiment of the present invention, the beam forming function f () of beam-shaper 504 can be linear or nonlinear.
In example embodiment described above, microphone array 106 is microphones (302
1, 302
2with 302
3) 1-D array, distinguish between its audio signal allowing the different angles of beam-shaper 504 in a dimension (such as, along horizontal axis) to receive.In alternative embodiments, microphone array 106 can be 2-D or the 3-D array of microphone, distinguish between its audio signal that beam-shaper will be allowed to receive the different angles respectively in two or three dimensions (such as, along level, vertical and depth axis).In addition, in a preferred embodiment, microphone array 106 is uniform arrays of microphone 302.But in alternative embodiments, microphone array 106 can be the nonuniform noise of microphone 302.
As described above, implement in beam-shaper 504, Echo Canceller 506 and processing block 508 software that can perform on CPU 104 or the hardware in equipment 102.When beam-shaper 504, Echo Canceller 506 and processing block 508 are implemented in software, they can be provided by the computer program be recorded in non-emporary computer-readable medium, and this computer program is configured to when performing on the CPU 104 of equipment 102 for performing the function of beam-shaper 504 as described above, Echo Canceller 506 and processing block 508.Method step S604 to S614 can be performed by the functional block of beam-shaper 504, Echo Canceller 506 and processing block 508 (as software or hardware block).
Meanwhile, embodiment described above relate to microphone array 106 and has received a desired audio signal (d from unique user 202
1), will be appreciated that, microphone array 106 can from the multiple user's received audio signals such as Conference Calling, and it can all be regarded as desired audio signal.In this case, multiple sources of desired audio signal arrive microphone array 106.
Equipment 102 can be TV or computer or for implementing other suitable equipment arbitrarily of the present invention.In addition, beam-shaper 504, Echo Canceller 506 and processing block 508 can be made to can be used in using the suitable arbitrarily equipment of stereophony microphone pickup.The method described above voice quality that can be implemented as in equipment promotes the part of (VQE) module, and runs on any suitably platform.
In embodiment described above, loud speaker 210 is the mono speakers for exporting monophonic audio signal, and is individual signals from the beam-shaper output of beam-shaper 504.But this is only in order to reduced representation, and the present invention is not limited to and is only used to such system.In other words, some embodiments of the present invention can use the boombox for exporting stereo audio signal, and some embodiments of the present invention can use the beam-shaper exporting multiple signal.
The present invention is useful especially for the application for the Echo Canceller suppressed based on echo, because it only needs use Echo Canceller (all Echo Cancellers 506 as described above), and because beam-shaper is incorporated in Echo Canceller very simple subsequently.For the mixing Echo Canceller used based on both reduction echo cancellor (it can be linear or nonlinear) and echo suppression, situation is just slightly different.The echo reduction part of mixing Echo Canceller need to be applied to each beam-shaper microphone input signal (with the similar fashion of carrying out in the first prior art equipment shown in Fig. 3).Similar with the first prior art equipment shown in Fig. 3, its defect had is that computation complexity will be higher, cuts down module this is because need to implement multiple echo and needs a large amount of to be integratedly incorporated in Echo Canceller solution by beam-shaper.But it is contrary with the first prior art equipment shown in Fig. 3, the performance of beam-shaper 504 can not be subject to the significant adverse impact of echo cancellor, and reason is that the echo performed before beam forming is cut down the echo in major effect institute received audio signal and can not obviously affect other received audio signal.When according to the invention process mixing echo cancellor, the echo suppression part of Echo Canceller still performs after beam-shaper.
Although in embodiment described above, processing block 508 is illustrated as independent of beam-shaper 504 and Echo Canceller 506, but in certain embodiments, some or all functions of processing block 508 can be implemented in beam-shaper 504 and/or Echo Canceller 506.
As described above, in a preferred embodiment, echo path to estimate being based on the strongest institute's received audio signal.But in alternative embodiments, echo path is estimated can based on different (or multiple) institute received audio signal.A kind of mode done like this is by adding users parameter, regulation echo path estimate based on selected by the power of audio signal cannot be lower than the power of the strongest audio signal more than x dB.Any institute received audio signal meeting this condition uses when can be selected at estimated echo path.The value of x can be fixing or can be transformable.User can change the value of x.Alternatively or in addition, equipment 102 can change the value of x.
Select the another kind of mode being different from the audio signal of the strongest audio signal be provide between any parts in microphone array 106 before known ultimate range, this can be used to calculate the maximum power difference between the audio signal that receives at different microphones 302 place of microphone array 106.It is the microphone signal with maximum power that this maximum difference can be used to subsequently by the Power Limitation of any microphone signal.Therefore, the knowledge of the power of any microphone signal can both provide the upper limit of the power of the strongest audio signal, this is because the difference between the power of known peak signal and arbitrarily other signal is restricted to be less than specified quantitative.
The echo path that requires to include the foregoing described for the Optimum Operation of Echo Canceller slowly changes and echo path is linear in make it can carry out modeling by linear model.In addition, we notice, Echo Canceller should and all types of echoes that can be able to occur in response environment 200.Term " reply " means that Echo Canceller can remove all appreciable echoes from the signal that microphone receives here.In addition, Echo Canceller also should retain any near-end sound as far as possible.Balance between the requirement of these part conflicts is subjective, and not for its clear and definite rule.
Echo is removed by estimated echo sample (in echo is cut down) and/or echo power (in echo is suppressed) and performs.These estimations are more accurate, and echo cancellor just will be lower for the impact of any near-end sound.Accuracy that echo sample/power is estimated is the accuracy based on EPM echo path model.If echo path can be considered to linear, will more adequately be estimated compared with nonlinear situation.If echo path change quickly, then also will more inaccurately be estimated compared with its change situation slowly.
(such as, in the second prior art equipment shown in Fig. 4) has the model that nonlinear beamforming device will require for its nonlinear characteristic in the echo path thus.Even if may propose such model, it also may be coarse, and will customize for concrete beam-shaper.Therefore, in any case all there is negative effect to the ability that Echo Canceller 406 retains near-end in the existence of beam-shaper 404, if but suitably design, then the ability removing echo is not affected.
In fact Echo Canceller can comprise the non-linear function of carrying out modeling caused loud speaker, thus it may be deposited in the echo path thus perform echo cancellor in the nonlinear case completely, but according to the accuracy of nonlinear model, attainable echo cancellation performance usually declines to some extent in near-end transparency.
Similarly, in the echo path thus (such as, in the second prior art equipment shown in Fig. 4) has the more inaccurate estimation that the beam-shaper changing echo path fast will cause echo path gain/echo sample estimation.Therefore, negative effect will be there is in the ability for Echo Canceller reservation near-end sound.But the ability that Echo Canceller removes echo should be affected, this is because Echo Canceller can be designed as those changes of reply.
In addition, although illustrate and describe the present invention especially with reference to preferred embodiment, those skilled in the art will be appreciated that, can carry out various change in form and details and not deviate from scope of the present invention as defined by the appended claims.
Claims (16)
1., in a method for equipment place audio signal, this equipment comprises the audio output device for output audio signal, and the method comprises:
At multiple microphone places received audio signal of this equipment;
Measure the characteristic of at least one in the audio signal that multiple microphone receives, at least one comprising the audio signal received from multiple microphone measures echo path gain;
Use described measured echo path gain to determine that the echo power in beam-shaper output is estimated;
Beam-shaper, to received audio signal application of beam former coefficient, generates beam-shaper thus and exports; And
Application echo cancellor device is exported to this beam-shaper, the echo of suppressing the audio signal exported from this audio output device to cause from this beam-shaper exports thus,
Wherein in the audio signal that receives of multiple microphone described at least one comprise the strongest audio signal received, and the operating parameter of this echo cancellor device is estimated to control based on the echo power during described beam-shaper exports.
2. method according to claim 1, comprises further and determines which institute's received audio signal is the strongest audio signal received.
3. method according to claim 1, wherein in the audio signal that receives of multiple microphone described at least one be not more than scheduled volume and be less than those audio signals received of the power of the power of the strongest audio signal received from having and select.
4. method according to claim 3, wherein this scheduled volume is variable, and the method comprises user changes this scheduled volume or equipment changes this scheduled volume.
5. method according to claim 1, the distance between at least one the power described wherein in the audio signal that receives of multiple microphone is used to based on multiple microphones of this equipment determines the upper limit of the power of the strongest institute's received audio signal.
6. method according to claim 1, comprise the beam-shaper gain depending on frequency of determining to be applied from the audio signal that the Main way of beam-shaper receives multiple microphone by beam-shaper further, wherein said beam-shaper gain be used in beam-shaper export in describedly determining of estimating of echo power.
7. method according to claim 1, the upper limit of echo power during wherein said determined echo power estimated statement oscillography beam shaper exports.
8. method according to claim 1, wherein said operating parameter comprises the echo suppression level being applied to beam-shaper output by echo cancellor device.
9. method according to claim 1, comprises further and carries out adaptation based on to the described beam-shaper coefficient of analysis to beam-shaper of received audio signal.
10. method according to claim 9, wherein to described beam-shaper coefficient carry out adaptive with in the beam forming constraint of beam-shaper by the minimum power in beam-shaper output.
11. methods according to claim 10, wherein the described beam forming constraint of beam-shaper comprises the audio signal not distortion due to application of beam former coefficient that multiple microphone receives in the Main way of beam-shaper.
12. methods according to claim 1, wherein beam-shaper coefficient describes linear function or the nonlinear function of the audio signal received.
13. 1 kinds of equipment for the treatment of audio signal, this equipment comprises:
Audio output device, for output audio signal;
Multiple microphone, for received audio signal;
Beam-shaper, it is configured to received audio signal application of beam former coefficient, generates beam-shaper thus and exports;
Echo cancellor device, it is configured to be applied to this beam-shaper and exports, the echo of suppressing the audio signal exported from this audio output device to cause from this beam-shaper exports thus; With
Processing apparatus, for measuring in audio signal that multiple microphone receives the characteristic of at least one, at least one comprising the audio signal received from multiple microphone measures echo path gain, for using described measured echo path gain to determine that the echo power in beam-shaper output is estimated, and estimate to control the operating parameter of this echo cancellor device for the echo power in exporting based on described beam-shaper, wherein in the audio signal that receives of multiple microphone described at least one comprise the strongest audio signal received.
14. equipment according to claim 13, wherein said echo cancellor device comprises the echo suppressor part being arranged to and being applied to beam-shaper and exporting.
15. equipment according to claim 14, wherein said echo cancellor device comprises further and being arranged at beam-shaper to each audio signal being applied to multiple microphone before received audio signal application of beam former coefficient and receiving.
16. according to claim 13 to the equipment according to any one of 15, and wherein this beam-shaper is the undistorted response beam-shaper of minimum variance.
Applications Claiming Priority (4)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| GB1116847.3A GB2495130B (en) | 2011-09-30 | 2011-09-30 | Processing audio signals |
| GB1116847.3 | 2011-09-30 | ||
| US13/307994 | 2011-11-30 | ||
| US13/307,994 US8824693B2 (en) | 2011-09-30 | 2011-11-30 | Processing audio signals |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102957819A CN102957819A (en) | 2013-03-06 |
| CN102957819B true CN102957819B (en) | 2015-01-28 |
Family
ID=47291203
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210368101.5A Active CN102957819B (en) | 2011-09-30 | 2012-09-28 | Method and apparatus for processing audio signals |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102957819B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB2521881B (en) * | 2014-04-02 | 2016-02-10 | Imagination Tech Ltd | Auto-tuning of non-linear processor threshold |
| CN106375527B (en) * | 2016-08-24 | 2019-05-17 | 维沃移动通信有限公司 | Speech quality detection method and mobile terminal |
| CN106782584B (en) * | 2016-12-28 | 2023-11-07 | 北京地平线信息技术有限公司 | Audio signal processing device, method and electronic device |
| US10056091B2 (en) * | 2017-01-06 | 2018-08-21 | Bose Corporation | Microphone array beamforming |
| US10299039B2 (en) * | 2017-06-02 | 2019-05-21 | Apple Inc. | Audio adaptation to room |
| CN108766456B (en) * | 2018-05-22 | 2020-01-07 | 出门问问信息科技有限公司 | Voice processing method and device |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1809105A (en) * | 2006-01-13 | 2006-07-26 | 北京中星微电子有限公司 | Dual-microphone speech enhancement method and system applicable to mini-type mobile communication devices |
| EP1722545A1 (en) * | 2005-05-09 | 2006-11-15 | Mitel Networks Corporation | A method to reduce training time of an acoustic echo canceller in a full-duplex beamforming-based audio conferencing system |
| CN101015001A (en) * | 2004-09-07 | 2007-08-08 | 皇家飞利浦电子股份有限公司 | Telephony device with improved noise suppression |
| CN101207663A (en) * | 2006-12-15 | 2008-06-25 | 美商富迪科技股份有限公司 | Internet communication device and method for controlling noise thereof |
| CN100407594C (en) * | 2002-07-19 | 2008-07-30 | 日本电气株式会社 | Sound echo inhibitor for hand free voice communication |
| CN100446530C (en) * | 1998-01-30 | 2008-12-24 | 艾利森电话股份有限公司 | Generating calibration signals for an adaptive beamformer |
| CN101667426A (en) * | 2009-09-23 | 2010-03-10 | 中兴通讯股份有限公司 | Device and method for eliminating environmental noise |
-
2012
- 2012-09-28 CN CN201210368101.5A patent/CN102957819B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100446530C (en) * | 1998-01-30 | 2008-12-24 | 艾利森电话股份有限公司 | Generating calibration signals for an adaptive beamformer |
| CN100407594C (en) * | 2002-07-19 | 2008-07-30 | 日本电气株式会社 | Sound echo inhibitor for hand free voice communication |
| CN101015001A (en) * | 2004-09-07 | 2007-08-08 | 皇家飞利浦电子股份有限公司 | Telephony device with improved noise suppression |
| EP1722545A1 (en) * | 2005-05-09 | 2006-11-15 | Mitel Networks Corporation | A method to reduce training time of an acoustic echo canceller in a full-duplex beamforming-based audio conferencing system |
| CN1809105A (en) * | 2006-01-13 | 2006-07-26 | 北京中星微电子有限公司 | Dual-microphone speech enhancement method and system applicable to mini-type mobile communication devices |
| CN101207663A (en) * | 2006-12-15 | 2008-06-25 | 美商富迪科技股份有限公司 | Internet communication device and method for controlling noise thereof |
| CN101667426A (en) * | 2009-09-23 | 2010-03-10 | 中兴通讯股份有限公司 | Device and method for eliminating environmental noise |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102957819A (en) | 2013-03-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP2749016B1 (en) | Processing audio signals | |
| US9210504B2 (en) | Processing audio signals | |
| CN102957819B (en) | Method and apparatus for processing audio signals | |
| CN103000185B (en) | Processing signals | |
| US8693704B2 (en) | Method and apparatus for canceling noise from mixed sound | |
| US9031257B2 (en) | Processing signals | |
| US9111543B2 (en) | Processing signals | |
| US8634547B2 (en) | Echo canceller operative in response to fluctuation on echo path | |
| JP5331201B2 (en) | Audio processing | |
| GB2495472B (en) | Processing audio signals | |
| KR20120101457A (en) | Audio zoom | |
| KR101934999B1 (en) | Apparatus for removing noise and method for performing thereof | |
| EP2700161B1 (en) | Processing audio signals | |
| CN102968999B (en) | Audio signal processing | |
| US11902758B2 (en) | Method of compensating a processed audio signal | |
| JPH09261792A (en) | Sound receiving method and its device | |
| CN102970638B (en) | Processing signals | |
| KR20220157475A (en) | Echo Residual Suppression | |
| JP4298466B2 (en) | Sound collection method, apparatus, program, and recording medium | |
| JP2002261659A (en) | Multi-channel echo cancellation method, its apparatus, its program, and its storage medium | |
| WO2022041030A1 (en) | Low complexity howling suppression for portable karaoke | |
| CN110121890B (en) | Method and apparatus for processing audio signal and computer readable medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20200514 Address after: Washington State Patentee after: MICROSOFT TECHNOLOGY LICENSING, LLC Address before: Ai Erlandubailin Patentee before: Skype |