CN1851806A - Adaptive microphone array system and its voice signal processing method - Google Patents
Adaptive microphone array system and its voice signal processing method Download PDFInfo
- Publication number
- CN1851806A CN1851806A CNA2006100835251A CN200610083525A CN1851806A CN 1851806 A CN1851806 A CN 1851806A CN A2006100835251 A CNA2006100835251 A CN A2006100835251A CN 200610083525 A CN200610083525 A CN 200610083525A CN 1851806 A CN1851806 A CN 1851806A
- Authority
- CN
- China
- Prior art keywords
- signal
- sef
- adapting filter
- coefficient
- adaptive
- 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.)
- Granted
Links
- 230000003044 adaptive effect Effects 0.000 title claims abstract description 106
- 238000003672 processing method Methods 0.000 title claims description 10
- 238000000034 method Methods 0.000 claims abstract description 41
- 239000004615 ingredient Substances 0.000 claims description 28
- 238000007493 shaping process Methods 0.000 claims description 15
- 238000004422 calculation algorithm Methods 0.000 claims description 14
- 238000004088 simulation Methods 0.000 claims description 14
- 230000003111 delayed effect Effects 0.000 claims description 11
- 230000005236 sound signal Effects 0.000 claims description 9
- 230000005540 biological transmission Effects 0.000 claims description 8
- 238000002789 length control Methods 0.000 claims description 8
- 230000000694 effects Effects 0.000 abstract description 11
- 238000005516 engineering process Methods 0.000 abstract description 4
- 238000003379 elimination reaction Methods 0.000 abstract 1
- 239000000203 mixture Substances 0.000 description 13
- 238000010586 diagram Methods 0.000 description 4
- 238000001914 filtration Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 2
- 230000000052 comparative effect Effects 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 239000012467 final product Substances 0.000 description 2
- 239000000047 product Substances 0.000 description 2
- 230000003313 weakening effect Effects 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000001934 delay Effects 0.000 description 1
- 230000003292 diminished effect Effects 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Landscapes
- Circuit For Audible Band Transducer (AREA)
- Soundproofing, Sound Blocking, And Sound Damping (AREA)
Abstract
This invention discloses an self-adaptive mike array system and a method for processing its phone signals, which adds an self-adaptive filter coefficient refresh control module in the mike system to control if the filter should refresh the coefficient or the step-size value when refreshing the coefficient, so that the refreshment of the coefficient is only against the noise so as to avoid the side effect of reducing effective phone signal amplitude while conducting filter coefficient refresh using existing technology when the noise component from the back signals is comparable with or smaller so than the phone component from frontage. This invented system applies a high-order adaptive filter to reduce the noise amplitude to the utmost degree and ensures the best noise-elimination effect no matter what positions the system zero-point directs when multiple noise sources exist.
Description
Technical field
The present invention relates to voice processing technology, particularly a kind of adaptive microphone array system and audio signal processing method thereof.
Background technology
In real-life situation, be full of various noises and echo, the proper communication that these interference can have a strong impact on people exchanges, and also can reduce the usefulness of speech recognition system, and a kind of settling mode just is to use microphone array system.First order difference microphone array system commonly used adopts beam forming technique, utilizes two non-directive microphones to form two back-to-back heart-shaped directive property Mikes.Microphone array not only can overcome environmental noise and the echo influence to voice signal, restores clean voice, and can record at a distance, rejects most of environmental noise in recording, solves wear-type Mike inconvenient problem with use.And this microphone array technology is not at the specific noise environment, and it goes for obtaining gratifying effect under any noisy environment.
Referring to Fig. 1, Fig. 1 is the structured flowchart of the self-adaptation first order difference microphone array system of prior art.As shown in Figure 1, this system comprises fixed directional beam shaping module 101, single order sef-adapting filter 102 and totalizer 103.
Wherein fixed directional beam shaping module 101 comprises time delay module 106 that aphalangia delays time to Mike B105, signal b (k) that Mike B is gathered to Mike A104 and aphalangia, is used for time delay module 108 and totalizer 107 and 109 that the signal f (k) that Mike A gathers is delayed time.
Mike A104 and Mike B105 form pattern back-to-back, and the distance between two Mikes is d, and hypothesis Mike A104 be preposition Mike over against target speaker mouth, be the dead ahead with this direction, i.e. 0 degree direction; Mike B be rearmounted Mike back to target speaker mouth, be the dead astern with this direction, promptly 180 the degree angular direction. Time delay module 106 and 108 delay time are d/c, and c is the velocity of sound.Totalizer 107 be used to calculate f (k) and b (k) time delayed signal reverse and; Totalizer 109 be used to calculate b (k) and f (k) time delayed signal reverse and.
Utilize system shown in Figure 1 to carry out the process flow diagram that voice signal is handled, as shown in Figure 2, this flow process may further comprise the steps:
Step 201, fixed directional beam shaping module 101 adopts two back-to-back non-directive Mike A104 and Mike B105, utilizes beam forming technique to constitute heart-shaped directive property Mike.Concrete steps comprise: the signal f (k) that collects with Mike A104 deducts the time delayed signal b (k-d/c) of the signal b (k) that Mike B105 collects, obtains the output signal x that principal ingredient is the voice signal of the place ahead incident (k).Simultaneously, the time delayed signal f (k-d/c) with b (k) deducts f (k) obtains the output signal n that principal ingredient is the noise signal of rear incident (k), thereby forms heart-shaped directive property Mike.
Heart-shaped directive property Mike is a kind of microphone of fixed directional, the zero point of its polar mode (polarpattern) is all the time towards the dead astern, the noise that the noise source that promptly main weakening is positioned at the dead astern is sent, the direction that needs to weaken signal is pointed to 180 degree directions all the time, and in the actual noisy environment, therefore the position of noise source might not need to adjust by adaptive filter coefficient w (k) the polar mode zero point of microphone array system in the dead astern.
Step 202, single order sef-adapting filter 102 adopts the adaptive filter coefficient update algorithm to calculate adaptive filter coefficient w (k), and utilize result of calculation to adjust the polar mode zero point of microphone array system, utilize formula n ' (k)=w (k) n (k), calculate (k), and the result is sent to totalizer 103 after oppositely with the estimated value n ' of the residual noise signal among the x (k) of n (k) simulation.Wherein, the polar mode θ at zero point of the value of adaptive filter coefficient w (k) and microphone array system
NullRelation be approximately:
(k) reverse of the estimated value n ' that step 203, totalizer 103 are calculated the residual noise signal that obtains in x (k) and the step 202 and, draw residual signals s (k), and with s (k) as k output signal constantly.Simultaneously, this output signal s (k) is sent into sef-adapting filter as feedback signal,, improve the quality of voice signal by the polar mode zero point of further adjustment microphone array system.
Owing to can not realize the coefficient update of wave filter is controlled in the said method, so have only when from the noise signal at rear during much larger than voice signal, utilize said system and method just the noise composition among the x (k) can be filtered out, can not cause negative effect simultaneously the voice signal among the x (k).And when the noise composition is very little in from the signal at rear, there be not sounding sometime such as noise source, because the imperfection of beam shaping, the amplitude of the efficient voice signal content among the n (k) may be suitable with noise signal, even much larger than the amplitude of noise signal, carrying out filter coefficient update according to said system and method this moment can make the polar mode of system adjust to the dead ahead zero point, it is source speech signal place direction, the spinoff of efficient voice amplitude appears reducing, as speech volume is diminished, voice fog, and echo etc. occurs.
In addition, said method uses the single order sef-adapting filter as postfilter, though can realize the self-adaptation adjustment at polar mode zero point of microphone array system in theory,, can obtain denoising effect preferably at zero point when certain noise source is in this during the orientation really.But when having a plurality of noise source simultaneously, this method with system's 0:00 direction sensing noise source is infeasible in theory, because zero point is corresponding a plurality of noise sources simultaneously, and noise source generally is not unique in the actual environment, and simple multiple relation between noise composition among the output signal x (k) and the n (k), therefore consider from the angle of sef-adapting filter performance, use the single order sef-adapting filter not reach the purpose of the noise contribution among the filtering x (k) substantially.
Summary of the invention
In view of this, fundamental purpose of the present invention is to provide a kind of adaptive microphone array system, and this system can control adaptive filter coefficient updates, solves the problem that reduces the efficient voice signal amplitude.
In view of this, another object of the present invention is to provide a kind of audio signal processing method of adaptive microphone array, this method can be controlled adaptive filter coefficient updates, solves the problem that reduces the efficient voice signal amplitude.
First aspect to achieve these goals, the invention provides a kind of adaptive microphone array system, this system comprises that the fixed directional wave beam forms module, sef-adapting filter, arithmetical unit, described fixed directional wave beam forms module and is used for two non-directive Mikes are formed heart-shaped directive property Mike, this system comprises that further adaptive filter coefficient upgrades control module
Adaptive filter coefficient upgrades control module, forming module with the fixed directional wave beam links to each other, receive the two paths of signals that the fixed directional wave beam forms module output, and send adaptive filter coefficient according to the relative size of two-way output signal energy to sef-adapting filter and upgrade control signal;
Described sef-adapting filter, according to the coefficient update control signal that receives, the usage factor update algorithm is determined the coefficient of sef-adapting filter, the principal ingredient that forms module output with the fixed directional wave beam is the signal of the noise signal of rear incident, simulates the residual noise signal in the signal that the principal ingredient of exporting is the place ahead incident voice signal;
Described arithmetical unit is used for more described principal ingredient and is the residual noise signal of this signal of the signal of the place ahead incident voice signal and simulation, obtains residual signals.
Described sef-adapting filter is the sub-band adaptive wave filter; It is one or more that described adaptive filter coefficient upgrades control module, and the number that adaptive filter coefficient upgrades control module is identical with sub band number.
It is energy comparer or sef-adapting filter step-length control module that described adaptive filter coefficient upgrades control module.
Described sef-adapting filter is higher-Order Time-Domain sef-adapting filter or high-order adaptive frequency domain filter.
This adaptive microphone array system further comprises time delay module, this module forms module with described fixed directional wave beam and links to each other with arithmetical unit, sends to described arithmetical unit after being used for the signal that described voice signal with the place ahead incident is a principal ingredient delayed time.
When described sef-adapting filter was the linear phase sef-adapting filter, the delay time of described time delay module was half of described sef-adapting filter exponent number.
Described arithmetical unit is totalizer or subtracter.
Second aspect to achieve these goals the invention provides a kind of audio signal processing method of adaptive microphone array, and this method may further comprise the steps:
A, utilize beam forming technique to constitute heart-shaped directive property Mike two back-to-back non-directive Mikes;
B, according to the speech energy of the place ahead incident of fixed directional beam shaping module output and the relative size of the noise energy of rear incident, send adaptive filter coefficient renewal control signal to sef-adapting filter;
C, sef-adapting filter are according to the control signal that receives, when determining to carry out the step value of coefficient update or coefficient update correspondence, the usage factor update algorithm is determined the adaptive filter coefficient vector of current time, then in conjunction with the adaptive filter coefficient vector of determining, with fixed directional beam shaping module output be the signal of principal ingredient with rear incident noise signal, simulation is the residual noise signal in the signal of principal ingredient with the place ahead incident voice signal;
D, the output of contrast fixed directional beam shaping module be the residual noise signal of simulating among the signal of principal ingredient and the step C with the place ahead incident voice signal, obtain residual signals.
The described transmission adaptive filter coefficient of step B upgrades control signal: when being mainly the noise signal of rear incident in the signal that described Mike collects, send renewal adaptive filter coefficient updates control signal to sef-adapting filter; Otherwise transmission stops to upgrade the adaptive filter coefficient updates control signal.
The described noise signal that is mainly rear incident by the ratio of the noise energy of the speech energy of the place ahead incident and rear incident is determined upgrade the adaptive filter coefficient updates control signal if the ratio of the two greater than preset threshold value, then sends; Otherwise transmission stops to upgrade the adaptive filter coefficient updates control signal.
The step value of correspondence that the described transmission adaptive filter coefficient of step B upgrades control signal when upgrading for sending the adaptive filter coefficient vector, described step value is definite with the ratio of the noise energy of rear incident by the speech energy of the place ahead incident.
The step value μ of described coefficient update correspondence is:
μ wherein
RefThe reference step-length of be setting, the reference energy ratio of ratio_ref for setting, ratio_env (k) be the ratio of the noise energy of the speech energy of k moment the place ahead incident and rear incident.
Step D is described further to be comprised after drawing residual signals: described residual signals is sent into sef-adapting filter, and sef-adapting filter is further adjusted adaptive filter coefficient according to the coefficient update control signal that next receives constantly.
When described sef-adapting filter was the single order sef-adapting filter, the described simulation of step C was that the residual noise signal in the signal of principal ingredient is that to be estimated as with rear incident noise signal be the signal of principal ingredient and the product of single order adaptive filter coefficient to described residual noise signal with the place ahead incident voice signal.
When described sef-adapting filter is the high-order sef-adapting filter, the described simulation of step C is that the residual noise signal in the signal of principal ingredient serves as in conjunction with being the signal of principal ingredient and the coefficient vector of determined sef-adapting filter with rear incident noise signal with the place ahead incident voice signal, estimates described residual noise signal with linear convolution.
Step D is described to be that the signal of principal ingredient serves as through the signal after the time-delay with the place ahead incident voice signal.
When described sef-adapting filter was the linear phase sef-adapting filter, described delay time was half of sef-adapting filter exponent number.
Step C is described to determine to carry out the step value of coefficient update or coefficient update correspondence for determining once every Fixed Time Interval.
By technical scheme of the present invention as seen, this adaptive microphone array system of the present invention and audio signal processing method thereof, upgrade control module by in microphone array system, increasing adaptive filter coefficient, two paths of signals to the output of fixed directional beam shaping module compares, and according to comparative result output filter coefficient update control signal to sef-adapting filter, whether the control adaptive filter coefficient upgrades or adaptive filter coefficient employed step value when upgrading, when thereby the self-adaptation first order difference microphone array system of having avoided utilizing prior art carries out filter coefficient update, the noise composition that in signal, is comprised from the rear with can compare or under the situation of phonetic element from the phonetic element in the place ahead, the efficient voice signal amplitude occurs reducing, introduce spinoffs such as voice distortion.
In addition, sef-adapting filter in adaptive microphone array system of the present invention and the audio signal processing method thereof adopts the high-order sef-adapting filter, compare with the single order sef-adapting filter, the high-order sef-adapting filter has better filtering and noise reduction effect, high-order sef-adapting filter length is long more, and analog capability is strong more.High-order sef-adapting filter of the present invention is simulated noise composition among the x (k) with a linear convolution process, directly pursue the output signal-noise ratio maximization, it is maximum filter out noise composition, make when having a plurality of noise source, no matter actual sensing at system's zero point where, can the assurance system obtain best denoising effect.
With just output behind the place ahead output signal time-delay T, guaranteed the causality of sef-adapting filter in adaptive microphone array system of the present invention and the audio signal processing method thereof.Simultaneously for fear of introducing unnecessary time-delay, the present invention adopts linear phase adaptation wave filter, and half that time-delay T is made as high-order sef-adapting filter exponent number gets final product.
Description of drawings
Fig. 1 is the structured flowchart of prior art self-adaptation first order difference microphone array system;
Fig. 2 utilizes system shown in Figure 1 to carry out the process flow diagram that voice signal is handled;
Fig. 3 is the structured flowchart of a preferred embodiment of adaptive microphone array system of the present invention;
Fig. 4 utilizes system shown in Figure 3 to carry out the process flow diagram that voice signal is handled.
Embodiment
In order to make purpose of the present invention, technical scheme and beneficial effect clearer, the present invention is described in further detail below in conjunction with drawings and Examples.
Adaptive microphone array system of the present invention and audio signal processing method thereof, upgrade control module by in microphone array system, increasing adaptive filter coefficient, whether the control adaptive filter coefficient upgrades or adaptive filter coefficient employed step value when upgrading, thereby avoided prior art self-adaptation first order difference microphone array system, the noise composition that in signal, is comprised from the rear with can compare or under the situation of phonetic element from the phonetic element in the signal in the place ahead, the efficient voice signal amplitude occurs reducing, introduce the voice distortion spinoff.
Referring to Fig. 3, Fig. 3 is the structured flowchart of a preferred embodiment of adaptive microphone array system of the present invention.As shown in Figure 3, this system comprises that the fixed directional wave beam forms module 301, sef-adapting filter 303, totalizer 305, adaptive filter coefficient renewal control module 302, time delay module 304.
Described fixed directional wave beam forms module 301 and is used for two non-directive Mikes are formed heart-shaped directive property Mike.
Adaptive microphone array system structure of the present invention and prior art system structure shown in Figure 1 difference are, system of the present invention has increased adaptive filter coefficient and has upgraded control module 302, this module forms module 301 with the fixed directional wave beam and links to each other with sef-adapting filter 303, be used to receive the fixed directional wave beam form module 301 outputs based on the signal x (k) of the place ahead incident voice signal with based on the signal n (k) of the noise signal of rear incident, and send adaptive filter coefficient according to the relative size of the noise energy of the speech energy of current time the place ahead incident and rear incident to sef-adapting filter 303 and upgrade control signal, the step value of correspondence when whether control sef-adapting filter 303 carries out coefficient update or coefficient update.
Sef-adapting filter 303, determine whether adaptive filter coefficient upgrades or step value during coefficient update according to the control signal that receives, and the usage factor update algorithm is determined update coefficients, use the residual noise signal among the signal x (k) of n (k) simulation the place ahead incident then, obtain estimated value n ' (k), and this estimated value is sent to totalizer 305.
Different with prior art system shown in Figure 1, system shown in Figure 3 also after fixed directional beam shaping module 301, has increased time delay module 304 before the totalizer 305, is used for x (k) signal is delayed time, and guarantees the causality of sef-adapting filter 303.
The time delayed signal x ' that totalizer 305 is used for comparison x (k) (k) and the estimated value n ' of the residual noise signal of the x (k) of sef-adapting filter 303 simulations (k), obtain residual signals.
Adaptive filter coefficient in the system shown in Figure 3 upgrades control module 302 can be energy comparer or sef-adapting filter step-length control module.
If it is the energy comparer that adaptive filter coefficient upgrades control module 302, then the energy comparer compares the two paths of signals x (k) of fixed directional beam shaping module 301 outputs and the energy of n (k), obtain the relative size of the two, and utilize the value generation adaptive filter coefficient of this relative size to upgrade control signal, whether the control sef-adapting filter carries out coefficient update.
If it is sef-adapting filter step-length control module that adaptive filter coefficient upgrades control module 302, then sef-adapting filter step-length control module is calculated the value of the pairing step size mu of adaptive filter coefficient renewal amount, and result of calculation sent to sef-adapting filter 303, sef-adapting filter 303 is determined the coefficient of sef-adapting filter according to this step value.
Referring to Fig. 4, Fig. 4 utilizes system shown in Figure 3 to carry out the process flow diagram that voice signal is handled.This flow process may further comprise the steps:
Concrete steps comprise: the signal f (k) that collects with Mike A306 deducts the time delayed signal b (k-d/c) of the signal b (k) that Mike B307 collects, obtains the output signal x that principal ingredient is the voice signal of the place ahead incident (k).Simultaneously, the time delayed signal f (k-d/c) with b (k) deducts f (k) obtains the output signal n that principal ingredient is the noise signal of rear incident (k), thereby forms heart-shaped directive property Mike.
If it is the energy comparer that adaptive filter coefficient upgrades control module 302, then to be obtained from the method for adaptive filter coefficient update control signal as follows for the energy comparer: at first adopt the method for single order recursion to calculate the energy envelope of x (k) and n (k), shown in formula (1) and (2)
x_env(k)＝α·x_env(k-1)+(1-α)·x
2(k) (1)
n_env(k)＝α·n_env(k-1)+(1-α)·n
2(k) (2)
In the following formula, x_env (k) is the energy envelope of signal x (k), and n_env (k) is the energy envelope of signal n (k), and α is a smoothing factor, and its span is 0＜α＜1.Formula (1) and (2) substitution formula (3) just can be compared in the hope of the energy envelope of x (k) and n (k):
The value of ratio_env (k) has reflected the relative size of the noise energy of the speech energy of current time the place ahead incident and rear incident, can utilize it to generate adaptive filter coefficient and upgrade control signal Adapt_En:
In the following formula, Thresh_ratio is a preset threshold, can be taken as 1.Following formula is a simple binary decision, can get in conjunction with formula (3), when x_env (k)＞n_env (k), ratio_env (k)＞1, be ratio_env (k)＞Thresh_ratio, Adapt_En=0 then, stop coefficient update this moment, stop the coefficient update of sef-adapting filter when promptly the signal that collects as Mike is mainly noise signal from the rear, otherwise, when x_env (k)＜n_env (k), ratio_env (k)＜1, be Pratio_env (k)＜Thresh_ratio, then Adapt_En=1 carries out the coefficient update of sef-adapting filter this moment.
Compare the two paths of signals that the fixed directional wave beam forms module output by the energy comparer, whether the coefficient of control sef-adapting filter upgrades, if in the signal that Mike collects from the noise composition at rear when very little or can be with phonetic element comparable, stop to carry out coefficient update, thereby the polar mode of having avoided making microphone array system is adjusted to the dead ahead zero point, make this system the efficient voice amplitude occur reducing, introduce the spinoff of voice distortion.
Described energy comparer obtains the process of control signal Adapt_En, it is a kind of embodiment that the coefficient update of sef-adapting filter is controlled, purpose is by the control to the coefficient update of sef-adapting filter, avoids weakening the problem of input target voice signal amplitude.This embodiment also can replace with the method for controlling the sef-adapting filter step-length more flexibly, and promptly adaptive filter coefficient renewal control module 302 can be sef-adapting filter step-length control module.
Sef-adapting filter step-length control module is according to the two paths of signals x (k) and the n (k) of 301 outputs of fixed directional beam shaping module, utilize formula (3) to determine the relative size of the noise energy of the speech energy of current time the place ahead incident and rear incident, utilize the speech energy of the place ahead incident of the current time of trying to achieve to determine that with the relative size of the noise energy of rear incident the current time adaptive filter coefficient upgrades corresponding step value then, this step value is sent to sef-adapting filter as the coefficient update control signal, reach the purpose that the control adaptive filter coefficient upgrades.
Be the example explanation with NLMS adaptive filter coefficient update algorithm below.The form that the NLMS algorithm adopts in engineering practice is represented with formula (5):
Can draw sef-adapting filter by the given NLMS coefficient update formula of formula (5) in k coefficient update amount constantly is:
Obviously, adjusting step size mu can the control coefrficient renewal amount.A kind of method according to ratio_env (k) adjustment step size mu is as follows:
In the following formula, μ
RefThe reference step-length of be setting, can be in 0 to 1 scope value, the reference energy ratio of ratio_ref for setting can be taken as 1, ratio_env (k) is the x (k) that utilizes formula (3) and obtain and the energy envelope ratio of n (k).By following formula as can be known, energy envelope is bigger more than the value of ratio_env (k), and the value of step size mu is just more little, and adaptive filter coefficient updates amount σ (k) is also just more little.
The coefficient update algorithm of sef-adapting filter can be LMS, NLMS, BNLMS, RLS scheduling algorithm.In sef-adapting filter update coefficients adjustment process, if adopt different adaptive filter coefficient update algorithm, then need the data difference obtained, the process of realization is also different fully.In general, sef-adapting filter has one the tunnel to be used for simulating the input signal and one tunnel echo signal that need simulate, the i.e. reference signal of other signal.
With NLMS adaptive filter coefficient update algorithm is the example explanation, as shown in Equation (5), (k) is a k adaptive filter coefficient vector constantly, if the exponent number of sef-adapting filter is N, then this vector comprises N element, when the exponent number of sef-adapting filter was single order, this vector had only an element w (k).
From formula (5) as can be seen, (k) be tried to achieve and following data: the k-l adaptive filter coefficient vector (k-1) in the moment, step size mu, k-l residual signals e_s (k-1), the k-1 constantly reference signal vector of sef-adapting filter constantly need be obtained
And k-1 reference signal constantly
Energy P
x(k-1).
After obtaining adaptive filter coefficient vector (k) according to formula (5), in conjunction with the adaptive filter coefficient vector (k) that obtains, utilize the noise composition among n (k) the simulation x (k), the estimated value n ' that obtains the noise composition among the x (k) (k).Adopt the single order sef-adapting filter in the system of present embodiment, then n ' (k) equals the product of n (k) and single order adaptive filter coefficient w (k).
E_s (k)=x ' (k)-n ' (k) (8) residual signals e_s (k) be k output signal constantly.With x (k) time-delay T is in order to guarantee the causality of sef-adapting filter, and delay time T can rule of thumb be determined by actual conditions.
Simultaneously the residual signals e_s (k) that obtains is sent into sef-adapting filter as feedback signal, be used for calculating next adaptive filter coefficient constantly according to formula (5).Next n (k+1) signal constantly is as the input signal of sef-adapting filter, sef-adapting filter is further adjusted adaptive filter coefficient according to next adaptive filter coefficient renewal control signal constantly of energy comparer or the output of sef-adapting filter step-length control module, carry out filtering repeatedly, progressively filter out the noise composition in the output signal, improve the quality of voice signal.
The fundamental starting point of using the single order sef-adapting filter in the above-described embodiments be zero point of adjusting adaptive microphone array system towards, when when zero point, orientation adjustment was to orientation, noise source place, to obtain denoising effect preferably.But when having a plurality of noise source simultaneously, this that system's 0:00 direction is pointed to the method for noise source is infeasible in theory because zero point corresponding a plurality of noise sources simultaneously.On the other hand, because exponent number is too short, the single order sef-adapting filter can't be simulated actual acoustic enviroment, its output signal n ' (k) often with x (k) in the noise composition differ greatly so poor effect of filtering and noise reduction.
The present invention adopts the single order sef-adapting filter in high-order sef-adapting filter replacement the foregoing description, and establishing high-order sef-adapting filter exponent number is N, and then the filter coefficient number is N, and filter length is N.The high-order sef-adapting filter is compared with the single order sef-adapting filter, has better filter effect, and present embodiment adopts a linear convolution process to come the noise composition of analog x ' in (k), and filter length is long more, and analog capability is strong more.Signal source n (k) obtains output signal n ' (k) through this high-order sef-adapting filter, and as shown in Equation (9), the w of the N in this formula (l) is exactly N the element of the coefficient vector (k) of the high-order sef-adapting filter obtained according to formula (5).
The method of utilizing the high-order sef-adapting filter provided by the invention is directly pursued the signal to noise ratio (S/N ratio) maximization of output signal, makes when having a plurality of noise source, and no matter where system points to zero point, can the assurance system obtain best denoising effect.
High-order sef-adapting filter of the present invention can be higher-Order Time-Domain sef-adapting filter or high-order adaptive frequency domain filter.
In the step 404 for the causality that guarantees sef-adapting filter with x (k) time-delay T, for fear of when guaranteeing the causality of sef-adapting filter, introducing unnecessary time-delay, can adopt the linear phase sef-adapting filter, as linear phase FIR filter, in this case, time-delay T be made as sef-adapting filter exponent number N half get final product.
Adaptive filter coefficient updates is that pointwise is carried out, consider the stationarity in short-term of voice signal, in order in the robustness that guarantees algorithm, to reduce the complexity of system, the scheme of the coefficient update of above-mentioned two kinds of control sef-adapting filters can the pointwise adjustment, but adjusts once about per 10～25ms.
Sef-adapting filter among the present invention can also replace with the sub-band adaptive wave filter.When using sub-filter, if carry out coefficient update control with the energy comparer, then each subband uses independently energy comparer; If adopt the scheme of control sef-adapting filter step-length to carry out coefficient update control, then each subband independently carries out the step-length adjustment.
Adopt in the embodiments of the invention totalizer comparison output signal n ' (k) with reference signal x ' (k), in actual applications, described totalizer also can replace with other arithmetical unit, such as subtracter etc.
In a word, the above is preferred embodiment of the present invention only, is not to be used to limit protection scope of the present invention.Within the spirit and principles in the present invention all, any modification of being done, be equal to replacement, improvement etc., all should be included within protection scope of the present invention.
Claims (18)
1, a kind of adaptive microphone array system, comprise that the fixed directional wave beam forms module, sef-adapting filter, arithmetical unit, described fixed directional wave beam forms module and is used for two non-directive Mikes are formed heart-shaped directive property Mike, it is characterized in that, this system comprises that further adaptive filter coefficient upgrades control module
Adaptive filter coefficient upgrades control module, forming module with the fixed directional wave beam links to each other, receive the two paths of signals that the fixed directional wave beam forms module output, and send adaptive filter coefficient according to the relative size of two-way output signal energy to sef-adapting filter and upgrade control signal;
Described sef-adapting filter, upgrade control signal according to the adaptive filter coefficient that receives, the usage factor update algorithm is determined the coefficient of sef-adapting filter, the principal ingredient that forms module output with the fixed directional wave beam is the signal of the noise signal of rear incident, and the principal ingredient that simulation fixed directional wave beam forms module output is the residual noise signal in the signal of the place ahead incident voice signal;
Described arithmetical unit is used for more described principal ingredient and is the residual noise signal of this signal of the signal of the place ahead incident voice signal and simulation, obtains residual signals.
2, the system as claimed in claim 1 is characterized in that, described sef-adapting filter is the sub-band adaptive wave filter; It is one or more that described adaptive filter coefficient upgrades control module, and the number that adaptive filter coefficient upgrades control module is identical with sub band number.
3, the system as claimed in claim 1 is characterized in that, described sef-adapting filter is higher-Order Time-Domain sef-adapting filter or high-order adaptive frequency domain filter.
As claim 1,2 or 3 described systems, it is characterized in that 4, it is energy comparer or sef-adapting filter step-length control module that described adaptive filter coefficient upgrades control module.
5, the system as claimed in claim 1, it is characterized in that, this system further comprises time delay module, this module forms module with described fixed directional wave beam and links to each other with arithmetical unit, sends to described arithmetical unit after being used for the signal that described voice signal with the place ahead incident is a principal ingredient delayed time.
6, system as claimed in claim 5 is characterized in that, when described sef-adapting filter was the linear phase sef-adapting filter, the delay time of described time delay module was half of described sef-adapting filter exponent number.
As claim 1 or 5 described systems, it is characterized in that 7, described arithmetical unit is totalizer or subtracter.
8, a kind of audio signal processing method of adaptive microphone array is characterized in that, this method may further comprise the steps:
A, utilize beam forming technique to constitute heart-shaped directive property Mike two back-to-back non-directive Mikes;
B, according to the speech energy of the place ahead incident of fixed directional beam shaping module output and the relative size of the noise energy of rear incident, send adaptive filter coefficient renewal control signal to sef-adapting filter;
C, sef-adapting filter are according to the control signal that receives, when determining to carry out the step value of coefficient update or coefficient update correspondence, the usage factor update algorithm is determined the adaptive filter coefficient vector of current time, then in conjunction with the adaptive filter coefficient vector of determining, with fixed directional beam shaping module output be the signal of principal ingredient with rear incident noise signal, simulation is the residual noise signal in the signal of principal ingredient with the place ahead incident voice signal;
D, the output of contrast fixed directional beam shaping module be the residual noise signal of simulating among the signal of principal ingredient and the step C with the place ahead incident voice signal, obtain residual signals.
9, method as claimed in claim 8, it is characterized in that, the described transmission adaptive filter coefficient of step B upgrades control signal: when being mainly the noise signal of rear incident in the signal that described Mike collects, send renewal adaptive filter coefficient updates control signal to sef-adapting filter; Otherwise transmission stops to upgrade the adaptive filter coefficient updates control signal.
10, method as claimed in claim 9, it is characterized in that, the described noise signal that is mainly rear incident is determined by the ratio of the noise energy of the speech energy of the place ahead incident and rear incident, if greater than preset threshold value, then sending, the ratio of the two upgrades the adaptive filter coefficient updates control signal; Otherwise transmission stops to upgrade the adaptive filter coefficient updates control signal.
11, method as claimed in claim 8, it is characterized in that, the described transmission adaptive filter coefficient of step B upgrades control signal and upgrades corresponding step value for sending adaptive filter coefficient, and the step value of described coefficient update correspondence is definite with the ratio of the noise energy of rear incident by the speech energy of the place ahead incident.
12, method as claimed in claim 11 is characterized in that, the step value μ of described coefficient update correspondence is:
μ wherein
RefThe reference step-length of be setting, the reference energy ratio of ratio_ref for setting, ratio_env (k) be the ratio of the noise energy of the speech energy of k moment the place ahead incident and rear incident.
13, as claim 8,9 or 11 described methods, it is characterized in that, step D is described further to be comprised after drawing residual signals: described residual signals is sent into sef-adapting filter, and sef-adapting filter is further adjusted adaptive filter coefficient according to the coefficient update control signal that next receives constantly.
As claim 8,9 or 11 described methods, it is characterized in that 14, described sef-adapting filter is the single order sef-adapting filter,
The described simulation of step C is that the residual noise signal in the signal of principal ingredient is that to be estimated as with rear incident noise signal be the signal of principal ingredient and the product of single order adaptive filter coefficient to described residual noise signal with the place ahead incident voice signal.
As claim 8,9 or 11 described methods, it is characterized in that 15, described sef-adapting filter is the high-order sef-adapting filter,
The described simulation of step C is that the residual noise signal in the signal of principal ingredient serves as in conjunction with being the signal of principal ingredient and the coefficient vector of determined sef-adapting filter with rear incident noise signal with the place ahead incident voice signal, estimates described residual noise signal with linear convolution.
16, method as claimed in claim 8 is characterized in that, step D is described to be that the signal of principal ingredient serves as through the signal after the time-delay with the place ahead incident voice signal.
17, method as claimed in claim 16 is characterized in that, when described sef-adapting filter was the linear phase sef-adapting filter, described delay time was half of sef-adapting filter exponent number.
18, method as claimed in claim 8 is characterized in that, step C is described to determine to carry out the step value of coefficient update or coefficient update correspondence for determining once every Fixed Time Interval.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200610083525A CN100578622C (en) | 2006-05-30 | 2006-05-30 | A kind of adaptive microphone array system and audio signal processing method thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200610083525A CN100578622C (en) | 2006-05-30 | 2006-05-30 | A kind of adaptive microphone array system and audio signal processing method thereof |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1851806A true CN1851806A (en) | 2006-10-25 |
| CN100578622C CN100578622C (en) | 2010-01-06 |
Family
ID=37133284
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN200610083525A Active CN100578622C (en) | 2006-05-30 | 2006-05-30 | A kind of adaptive microphone array system and audio signal processing method thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100578622C (en) |
Cited By (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101192411B (en) * | 2007-12-27 | 2010-06-02 | 北京中星微电子有限公司 | Large distance microphone array noise cancellation method and noise cancellation system |
| CN101587712B (en) * | 2008-05-21 | 2011-09-14 | 中国科学院声学研究所 | Directional speech enhancement method based on small microphone array |
| CN102201231A (en) * | 2010-03-23 | 2011-09-28 | 创杰科技股份有限公司 | Voice sensing method |
| CN102421050A (en) * | 2010-09-17 | 2012-04-18 | 三星电子株式会社 | Apparatus and method for enhancing audio quality using non-uniform configuration of microphones |
| CN101647059B (en) * | 2007-02-26 | 2012-09-05 | 杜比实验室特许公司 | Speech enhancement in entertainment audio |
| CN102737645A (en) * | 2012-06-15 | 2012-10-17 | 武汉天喻信息产业股份有限公司 | Algorithm for estimating pitch period of voice signal |
| CN102779524A (en) * | 2008-10-10 | 2012-11-14 | 三星电子株式会社 | Apparatus and method for noise estimation, and noise reduction apparatus employing the same |
| CN101203063B (en) * | 2007-12-19 | 2012-11-28 | 北京中星微电子有限公司 | Method and apparatus for noise elimination of microphone array |
| CN101577847B (en) * | 2008-05-08 | 2013-01-02 | 索尼株式会社 | Signal processing device and signal processing method |
| WO2014085978A1 (en) * | 2012-12-04 | 2014-06-12 | Northwestern Polytechnical University | Low noise differential microphone arrays |
| CN104811860A (en) * | 2014-01-28 | 2015-07-29 | 华为技术有限公司 | Pickup signal focusing method and device and pickup device |
| CN105489224A (en) * | 2014-09-15 | 2016-04-13 | 讯飞智元信息科技有限公司 | Voice noise reduction method and system based on microphone array |
| CN105575397A (en) * | 2014-10-08 | 2016-05-11 | 展讯通信（上海）有限公司 | Voice noise reduction method and voice collection device |
| CN106331710A (en) * | 2009-07-01 | 2017-01-11 | 索尼公司 | Image processing device and method |
| CN106710603A (en) * | 2016-12-23 | 2017-05-24 | 上海语知义信息技术有限公司 | Speech recognition method and system based on linear microphone array |
| CN109714689A (en) * | 2018-12-21 | 2019-05-03 | 南京理工大学 | A kind of directionality acoustics index acquisition methods based on difference microphone linear array |
| CN106331710B (en) * | 2009-07-01 | 2019-07-16 | 维洛媒体国际有限公司 | Image processing apparatus and method |
| CN110419228A (en) * | 2017-03-16 | 2019-11-05 | 三菱电机株式会社 | Signal processing apparatus |
| CN111540372A (en) * | 2020-04-28 | 2020-08-14 | 北京声智科技有限公司 | Method and device for multi-microphone array noise reduction processing |
| EP3764360A1 (en) | 2019-07-10 | 2021-01-13 | Analog Devices International Unlimited Company | Signal processing methods and systems for beam forming with improved signal to noise ratio |
| EP3764664A1 (en) | 2019-07-10 | 2021-01-13 | Analog Devices International Unlimited Company | Signal processing methods and systems for beam forming with microphone tolerance compensation |
| EP3764358A1 (en) | 2019-07-10 | 2021-01-13 | Analog Devices International Unlimited Company | Signal processing methods and systems for beam forming with wind buffeting protection |
| EP3764359A1 (en) | 2019-07-10 | 2021-01-13 | Analog Devices International Unlimited Company | Signal processing methods and systems for multi-focus beam-forming |
| EP3764660A1 (en) | 2019-07-10 | 2021-01-13 | Analog Devices International Unlimited Company | Signal processing methods and systems for adaptive beam forming |
-
2006
- 2006-05-30 CN CN200610083525A patent/CN100578622C/en active Active
Cited By (42)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101647059B (en) * | 2007-02-26 | 2012-09-05 | 杜比实验室特许公司 | Speech enhancement in entertainment audio |
| CN101203063B (en) * | 2007-12-19 | 2012-11-28 | 北京中星微电子有限公司 | Method and apparatus for noise elimination of microphone array |
| CN101192411B (en) * | 2007-12-27 | 2010-06-02 | 北京中星微电子有限公司 | Large distance microphone array noise cancellation method and noise cancellation system |
| CN101577847B (en) * | 2008-05-08 | 2013-01-02 | 索尼株式会社 | Signal processing device and signal processing method |
| CN101587712B (en) * | 2008-05-21 | 2011-09-14 | 中国科学院声学研究所 | Directional speech enhancement method based on small microphone array |
| CN102779524A (en) * | 2008-10-10 | 2012-11-14 | 三星电子株式会社 | Apparatus and method for noise estimation, and noise reduction apparatus employing the same |
| CN102779524B (en) * | 2008-10-10 | 2015-01-07 | 三星电子株式会社 | Apparatus and method for noise estimation, and noise reduction apparatus employing the same |
| US9159335B2 (en) | 2008-10-10 | 2015-10-13 | Samsung Electronics Co., Ltd. | Apparatus and method for noise estimation, and noise reduction apparatus employing the same |
| CN106331710A (en) * | 2009-07-01 | 2017-01-11 | 索尼公司 | Image processing device and method |
| US11328452B2 (en) | 2009-07-01 | 2022-05-10 | Velos Media, Llc | Image processing device and method |
| US10614593B2 (en) | 2009-07-01 | 2020-04-07 | Velos Media, Llc | Image processing device and method |
| CN106331710B (en) * | 2009-07-01 | 2019-07-16 | 维洛媒体国际有限公司 | Image processing apparatus and method |
| CN102201231B (en) * | 2010-03-23 | 2012-10-24 | 创杰科技股份有限公司 | Voice sensing method |
| CN102201231A (en) * | 2010-03-23 | 2011-09-28 | 创杰科技股份有限公司 | Voice sensing method |
| CN102421050A (en) * | 2010-09-17 | 2012-04-18 | 三星电子株式会社 | Apparatus and method for enhancing audio quality using non-uniform configuration of microphones |
| CN102421050B (en) * | 2010-09-17 | 2017-04-12 | 三星电子株式会社 | Apparatus and method for enhancing audio quality using non-uniform configuration of microphones |
| CN102737645A (en) * | 2012-06-15 | 2012-10-17 | 武汉天喻信息产业股份有限公司 | Algorithm for estimating pitch period of voice signal |
| US9749745B2 (en) | 2012-12-04 | 2017-08-29 | Northwestern Polytechnical University | Low noise differential microphone arrays |
| US9237391B2 (en) | 2012-12-04 | 2016-01-12 | Northwestern Polytechnical University | Low noise differential microphone arrays |
| WO2014085978A1 (en) * | 2012-12-04 | 2014-06-12 | Northwestern Polytechnical University | Low noise differential microphone arrays |
| CN104811860A (en) * | 2014-01-28 | 2015-07-29 | 华为技术有限公司 | Pickup signal focusing method and device and pickup device |
| CN105489224A (en) * | 2014-09-15 | 2016-04-13 | 讯飞智元信息科技有限公司 | Voice noise reduction method and system based on microphone array |
| CN105489224B (en) * | 2014-09-15 | 2019-10-18 | 讯飞智元信息科技有限公司 | A kind of voice de-noising method and system based on microphone array |
| CN105575397A (en) * | 2014-10-08 | 2016-05-11 | 展讯通信（上海）有限公司 | Voice noise reduction method and voice collection device |
| CN105575397B (en) * | 2014-10-08 | 2020-02-21 | 展讯通信（上海）有限公司 | Voice noise reduction method and voice acquisition equipment |
| CN106710603A (en) * | 2016-12-23 | 2017-05-24 | 上海语知义信息技术有限公司 | Speech recognition method and system based on linear microphone array |
| CN106710603B (en) * | 2016-12-23 | 2019-08-06 | 云知声（上海）智能科技有限公司 | Utilize the audio recognition method and system of linear microphone array |
| CN110419228A (en) * | 2017-03-16 | 2019-11-05 | 三菱电机株式会社 | Signal processing apparatus |
| CN110419228B (en) * | 2017-03-16 | 2020-12-29 | 三菱电机株式会社 | Signal processing device |
| CN109714689A (en) * | 2018-12-21 | 2019-05-03 | 南京理工大学 | A kind of directionality acoustics index acquisition methods based on difference microphone linear array |
| EP3764360A1 (en) | 2019-07-10 | 2021-01-13 | Analog Devices International Unlimited Company | Signal processing methods and systems for beam forming with improved signal to noise ratio |
| EP3764664A1 (en) | 2019-07-10 | 2021-01-13 | Analog Devices International Unlimited Company | Signal processing methods and systems for beam forming with microphone tolerance compensation |
| EP3764358A1 (en) | 2019-07-10 | 2021-01-13 | Analog Devices International Unlimited Company | Signal processing methods and systems for beam forming with wind buffeting protection |
| EP3764359A1 (en) | 2019-07-10 | 2021-01-13 | Analog Devices International Unlimited Company | Signal processing methods and systems for multi-focus beam-forming |
| EP3764660A1 (en) | 2019-07-10 | 2021-01-13 | Analog Devices International Unlimited Company | Signal processing methods and systems for adaptive beam forming |
| WO2021005225A1 (en) | 2019-07-10 | 2021-01-14 | Ruwisch Patent Gmbh | Signal processing methods and systems for beam forming with microphone tolerance compensation |
| WO2021005219A1 (en) | 2019-07-10 | 2021-01-14 | Ruwisch Patent Gmbh | Signal processing methods and systems for beam forming with improved signal to noise ratio |
| WO2021005227A1 (en) | 2019-07-10 | 2021-01-14 | Ruwisch Patent Gmbh | Signal processing methods and systems for adaptive beam forming |
| WO2021005221A1 (en) | 2019-07-10 | 2021-01-14 | Ruwisch Patent Gmbh | Signal processing methods and systems for beam forming with wind buffeting protection |
| WO2021005217A1 (en) | 2019-07-10 | 2021-01-14 | Analog Devices International Unlimited Company | Signal processing methods and systems for multi-focus beam-forming |
| CN111540372A (en) * | 2020-04-28 | 2020-08-14 | 北京声智科技有限公司 | Method and device for multi-microphone array noise reduction processing |
| CN111540372B (en) * | 2020-04-28 | 2023-09-12 | 北京声智科技有限公司 | Method and device for noise reduction processing of multi-microphone array |
Also Published As
| Publication number | Publication date |
|---|---|
| CN100578622C (en) | 2010-01-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN1851806A (en) | Adaptive microphone array system and its voice signal processing method | |
| CN100524465C (en) | A method and device for noise elimination | |
| KR101340215B1 (en) | Systems, methods, apparatus, and computer-readable media for dereverberation of multichannel signal | |
| EP2562752A1 (en) | Sound source separator device, sound source separator method, and program | |
| CN103871421B (en) | A kind of self-adaptation noise reduction method and system based on subband noise analysis | |
| CN108604452B (en) | Sound signal enhancement device | |
| CN105575397B (en) | Voice noise reduction method and voice acquisition equipment | |
| AU2003296976A1 (en) | System and method for speech processing using independent component analysis under stability constraints | |
| CN101031956A (en) | Headset for separation of speech signals in a noisy environment | |
| WO2014181330A1 (en) | A method and apparatus for suppression of unwanted audio signals | |
| TW201030733A (en) | Systems, methods, apparatus, and computer program products for enhanced active noise cancellation | |
| CN106531145B (en) | Recurrence active noise control method based on M estimator | |
| US11373667B2 (en) | Real-time single-channel speech enhancement in noisy and time-varying environments | |
| CN102883243A (en) | Method and device for balancing frequency response of sound reproduction system through online iteration | |
| CN105913836B (en) | A kind of fixed point real-time noise-reducing method based on DSP | |
| CN105976806A (en) | Active noise control method based on maximum entropy | |
| US20150296318A1 (en) | Sound-source separation method, apparatus, and program | |
| CN112289337B (en) | Method and device for filtering residual noise after machine learning voice enhancement | |
| CN1868136A (en) | Echo canceller | |
| CN112435681B (en) | Voice enhancement method based on acoustic focusing and microphone array beam forming | |
| CN109243482A (en) | Improve the miniature array voice de-noising method of ACRANC and Wave beam forming | |
| Palla et al. | Wearable speech enhancement system based on MEMS microphone array for disabled people | |
| Prasad et al. | Two microphone technique to improve the speech intelligibility under noisy environment | |
| CN113639136A (en) | Noise reduction system and method for noise of ventilating duct | |
| Pengxing et al. | An active impulsive noise control algorithm with self-p-normalized method |
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: 20210302 Address after: No. 602, 6th floor, shining building, 35 Xueyuan Road, Haidian District, Beijing 100083 Patentee after: BEIJING ZHONGXINGTIANSHI TECHNOLOGY Co.,Ltd. Address before: 100083, Haidian District, Xueyuan Road, Beijing No. 35, Nanjing Ning building, 15 Floor Patentee before: Vimicro Corp. |