CN109061567A - Voice accurate positioning method under multi-source environment - Google Patents
Voice accurate positioning method under multi-source environment Download PDFInfo
- Publication number
- CN109061567A CN109061567A CN201810926869.7A CN201810926869A CN109061567A CN 109061567 A CN109061567 A CN 109061567A CN 201810926869 A CN201810926869 A CN 201810926869A CN 109061567 A CN109061567 A CN 109061567A
- Authority
- CN
- China
- Prior art keywords
- sound source
- voice
- microphone
- source
- environment
- 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
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S5/00—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations
- G01S5/18—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations using ultrasonic, sonic, or infrasonic waves
- G01S5/22—Position of source determined by co-ordinating a plurality of position lines defined by path-difference measurements
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Abstract
The present invention relates to the voice accurate positioning methods under a kind of multi-source environment, belong to speech processes field.This method first tests the voice delivery function of environment, then a precision is given, need to export a pinpoint condition according to actual environment, and using the frequency response function vector design of minimax, and adjustment or increases the method for microphone position and number and meet condition, then the optimization problem of voice positioning is established, and develops real time algorithm and optimization problem is solved, to realize the accurate positioning within the scope of given positioning accuracy.The present invention can verify that position error meets required precision, realize and be accurately positioned, and being capable of real-time implementation for arbitrarily choosing source position and noise position.
Description
Technical field
The invention belongs to field of voice signal, the voice accurate positioning method that is related under multi-source environment.
Background technique
Multi-source environment is a kind of very universal voice environment, and single source environment is its special case, in multi-source
Voice positioning under environment is extremely important.In some conferencing environments, it is often necessary to be positioned to main sound source.Microphone array
It has a very wide range of applications in speech signal processing, includes monitoring, video conference, hands-free devices, man-machine interactive system
Deng.Voice positioning is a practical function of microphone array.
Voice localization method is generally divided into two classes.One kind is the method based on TDOA, i.e., by estimating that different microphones connect
The time difference of the signal received obtains the incident direction of voice signal.Such methods are very more, such as common GCC method.It is this kind of
It is not very strong far field model that method, which is suitable for reflection, and for reflecting very strong environment, the influence of echo is very big, and positioning can go out
Existing severe deviations.Second class is not consider the method for TDOA, general using the method and base that form (SPR) based on pilot beam
Corresponding Wave beam forming problem is established in some position or orientation, thus the mapping function guiding relevant to energy of output one
The position or orientation.The use of such methods is than wide.But it is this kind of due to the wideband of the complexity of environment and voice signal
The stability of method is also excessively poor.
Generally speaking, in terms of the difficult point of current voice localization method is stability or robustness.It often can only be at certain
A little situations, which can be realized, to be properly positioned, and severe deviations occurs in the positioning of other situations, this is the bottleneck of voice localization method
Problem.Main cause is the complexity of voice environment.On the one hand be voice be a kind of broadband signal, have the characteristics that it is multifarious,
The locating effect of different voice signals also can be variant, therefore considers that the orientation problem of all possible voice signal is simultaneously
It is very difficult.On the other hand it is the complexity of environment, there is areflexia, the intensity of reflection is different, has clear etc. that can all influence
To the communication process of voice.These reasons result in the very high unstability of current method.
Summary of the invention
In view of this, the purpose of the present invention is to provide the voice accurate positioning methods under a kind of multi-source environment.Firstly, logical
It crosses the test of voice environment and determines the voice signal transmission function under actual environment with optimization method, so as to find out true
Transmission function.Then, the condition that voice positions under multi-source environment is provided, thus to realize that the accurate positioning under given accuracy is established
Theoretical basis, and frequency response function vector is solved by a specific Wave beam forming design method.Then, by adjusting or
The method for increasing the position and number of microphone, finds out and is just able to satisfy condition under what situation, to realize accurate positioning.Most
Afterwards, voice positioning is created as an optimization problem, and provides a fast method and realizes real-time navigation capability.
In order to achieve the above objectives, the invention provides the following technical scheme:
Voice accurate positioning method under multi-source environment, method includes the following steps:
S1: voice positioning is realized using microphone array system；
S2: the voice delivery function of environment is tested；
S3: need to export a pinpoint condition according to actual environment, and using the frequency response letter of minimax
The method of number vector design and adjustment and microphone meets condition；
Voice positioning: being established as an optimization problem, and develop real time algorithm and solve to optimization problem by S4, thus
Realize the accurate positioning under given positioning accuracy.
Further, in the step S1, the position of microphone and the position of sound source be not be overlapped, is usually laid at the upper of sound source
Side.
Further, the step S2 specifically:
Environment locating for sound source is near field model and far field model, and what far field model considered is the directional information of sound source, closely
What field model considered is the location information of sound source；What this method considered is location information, it is not limited near field model or far field
Model；
When the environment of sound source has given, the transmission function of voice determines that, it is expressed as
T(r,q,f) (1)
Wherein r is the position of sound source, and q is the position of microphone, and it is a complex function, table that f, which is the frequency of the signal function,
Show variation about frequency f of the sound from source position r to target position q, then, the transmission function formula of voice signal is
Y (q, f)=X (r, f) T (r, q, f), i=1 ..., N (2)
Wherein, X (r, f) is the Fourier transform of source signal, and Y (q, f) is the Fourier transform for receiving signal；
The determination of transmission function T (r, q, f), can be obtained by real-time testing；Played on the r of position many group voices and
Noise signal Xk(r, f), k=1 ..., m simultaneously enrolls the signal Y received on the q of positionk(r, f), k=1 ..., m, then multiple
Function T (r, q, f) is decomposed into u (f)+jv (f)；According to the specific voice signal feature of specific voice, some frequencies are selected
Set；For each frequency f, following optimization problem is established
The problem is a quadratic programming problem, solves and is very easy to.In general, it does not need to seek each frequency
Solve this optimization problem, it is only necessary to which this optimization problem is solved to interested frequency；To the possible position of each sound source and
After test solution is done in each microphone reception position, transmission function can be found out.
It is further, in step s3, described to need to export a pinpoint condition according to actual environment specifically:
Voice is positioned, will not include that noise positions to institute's sound source, it is only necessary to it is positioned against main sound source,
The intensity of main sound source needs the intensity than other sound sources big, just can determine that it is main sound source, otherwise the environment is considered as noise circumstance；
Give P sound source Xk(rk, f), k=1 ..., P, wherein rkIndicate the position of k-th of sound source；Main sound source is denoted as
X1, other P-1 sound sources Xk, k=2 ..., P are interference sound sources, are considered as noise；According to the requirement of main sound source, there is following agreement
Wherein M >=1, frequency are only needed in some frequency range where voice；
For the value of M depending on the requirement of practical application, M is bigger, and the requirement to environment is stringenter, and realization is more held
Easily；M is smaller, and the range of application is wider, and corresponding cost of implementation is also higher.
Further, in step s3, described using the frequency response function vector design of minimax and adjustment or increasing
The method of microphone is added to meet condition specifically:
For the microphone array of N number of element, the position of microphone is denoted as qi, i=1 ..., N；Then i-th microphone
Receiving signal is
Wherein YikFor the part about k-th of sound source that i-th of microphone receives, i.e.,
Yik(rk, f) and=Xk(rk,f)T(rk,qi, f), i=1 ..., N (6)
For each microphone, behind connect a filter, after adding filter, corresponding filter freguency response letter
Number is Hi(f), i=1 ..., N, by synthesizing an output signal after filter
Remember that G (r, f) is that output response function is as follows
Wherein T (r, f)=[T (r, q1,f),…,T(r,qN,f)]T, H (f)=[H1(f),…,HN(f)]T；So,
Firstly, for any given feasible locationNeed to define a pass band areasWith a stop band region Must include It must be withSeparation, there is an intermediate zone in centre；Based on examining for raising positioning accuracy
Consider,Directly it is taken asThis point,Be it is all withDistance be more than or equal to position error region, i.e.,
Then, an ideal receptance function G is definedd(r, f) is as follows:
Wherein, rcIt is a reference position point, it can arbitrarily choose, but generally when taking the center of microphone array, effect
Fruit is relatively good；The set I for taking an interested frequency establishes following optimization problem for any f ∈ I
Wherein,
It is denoted as by solving the optimal solution that the problem obtains
For any given feasible locationJust there is one to correspond toOptimization problem (13), its performance number is
Introduce the following conditions:
Condition 1: for arbitraryCorresponding performance numberAll meet
Further, the step S4 specifically:
When condition 1 is satisfied, and at least there is a frequency and meet | X1(r1, f) | when > 0, the precision of positioning must
To guarantee；
Frequency response function vector H (f) microphone system given for one just exports a function (9)；For fixed
Position information is added, and is needed according to the information of position plus corresponding frequency response function vector；Then for each position
Solve problems (13) obtain frequency response function vectorFind out following output function
It is as follows to choose objective function
Then it is as follows to establish optimization problem
The optimal solution of the problem is denoted asThe sound source position estimated value solved；
For any sound source position rk, k=1 ..., P obtain estimated value, positioning mistake by solving optimization problem (19)
Difference is denoted as
The supplementary set for defining stopband is as follows
The position of so main sound source must meet
The position r of i.e. main sound source1It certainly will not be in the optimal solution of problem (19)Stopband in；It enables
Wherein S is the region of entire feasible location,Definition be
It can so be obtained by (10) and (11)
It is found out by (22) for arbitrary r1∈ S, position error meet
The error for finding out the position estimation value and real sound source position come does not exceed ε.
The beneficial effects of the present invention are: the present invention, can also similar verifying position error for arbitrarily choosing source position
Very small, present case is tested using any selection random site, is realized and is accurately positioned, and is realized very quick.
Detailed description of the invention
In order to keep the purpose of the present invention, technical scheme and beneficial effects clearer, the present invention provides following attached drawing and carries out
Illustrate:
Fig. 1 is microphone working principle figure；
Fig. 2 is Region Decomposition figure；
Fig. 3 is the microphone array location drawing；
Fig. 4 is target function value；
Fig. 5 is flow chart of the present invention.
Specific embodiment
Below in conjunction with attached drawing, a preferred embodiment of the present invention will be described in detail.
Voice positioning system and method
1. microphone array system
Microphone array is the system using multiple microphones, by the difference of the voice signal received between microphone
Property reaches various functions, such as signal enhancing, positioning, Signal separator, and principle is as shown in Figure 1.Microphone system passes through reception
Voice signal, and handled by subsequent filter, then synthesize an output signal.Wherein, the frequency of filter is rung
Function needs are answered to design.It designs corresponding frequency response function according to actual requirement, so that output signal reaches
Given function.
Voice positioning must use microphone array system, and the position of microphone cannot be repeated with the position of sound source, usually
It is placed in the top of sound source.
2. the test of voice delivery model
Environment locating for sound source be it is extremely complex, it is divided near field model and far field model, and what far field model considered is
The directional information of sound source, and near field model consider be sound source location information.Near field model has very more reflection echoes, because
This its transmission function is extremely complex.What this method considered is location information, it is not limited near field or far field, but be near field
It is good.
When the environment of sound source has given, the transmission function of voice determines that, it is expressed as
T(r,q,f) (1)
Wherein r is the position of sound source, and q is the position of microphone, and f is the frequency of signal.The function is a complex function, table
Shown frequency be f signal from source position r to the variation of target position q.So,
Y (q, f)=X (r, f) T (r, q, f), i=1 ..., N (2)
Wherein, X (r, f) is the Fourier transform of source signal, and Y (q, f) is the Fourier transform for receiving signal.
Determination for transmission function T (r, q, f) needs to pass through real-time testing.We play many group languages on the r of position
Sound signal Xk(r, f), k=1 ..., m simultaneously enrolls the signal Y received in position qk(r, f), k=1 ..., m, then complex function
T (r, q, f) is decomposed into u (f)+jv (f).Then, for each frequency f, consider following optimization problem
Wherein X (r, f)=[X1(r1, f) ..., Xm(r,f)]T, Y (r, f)=[Y1(r1, f) ..., Ym(r,f)]T。
This is a very common quadratic programming problem, after doing some necessary abbreviations, there is the optimal solution of the problem, can be with
Real-time solution.
Do not need to solve each frequency this optimization problem, it is only necessary to some interested frequencies (signal with
Frequency where noise) solve this optimization problem.Possible position and each microphone to each sound source receive position and do
After test solves, gained optimal solution just constitutes transmission function, this will be applied to the design of filter freguency response function.
3. the requirement of multi-source voice signal
In a given environment, generally there are very more source signals, but many signals all show as noise.For
Voice positioning, will not include that noise positions to institute's sound source, it is only necessary to be positioned against main sound source, such as saying in meeting
Talk about people.The intensity of main sound source needs the intensity than other sound sources big, just can determine that it is main sound source.Otherwise, the intensity ratio of main sound source
Noise also hour, it can not can determine that its position.
Give P sound source Xk(rk, f), k=1 ..., P, wherein rkIndicate the position of k-th of sound source.We remember main sound source
For X1, other P-1 sound sources Xk, k=2 ..., P are interference sound sources, are considered as noise.According to the requirement of main sound source, there is following agreement
Wherein M >=1, frequency are only needed in some frequency range where voice.
For the value of M depending on the requirement of practical application, M is bigger, and the requirement to voice environment is stringenter, realizes more
It is easy.M is smaller, and the range of application is wider, and corresponding cost of implementation is also higher.
4. the foundation of location model
For the microphone array of N number of element, the position of microphone is denoted as qi, i=1 ..., N.Then i-th microphone
Receiving signal is
Wherein YikFor the part about k-th of sound source that i-th of microphone receives, i.e.,
Yik(rk, f) and=Xk(rk,f)T(rk,qi, f), i=1 ..., N (6)
For each microphone, behind connect a filter.After adding filter, corresponding filter freguency response letter
Number is Hi(f), i=1 ..., N, by synthesizing an output signal after filter
Remember that G (r, f) is that output response function is as follows
Wherein T (r, f)=[T (r, q1,f),…,T(r,qN,f)]T, H (f)=[H1(f),…,HN(f)]T.So,
As soon as microphone system exports a function (9) for a given frequency response function vector H (f).This is not
It can be used to position.In order to location information is added, need according to the information of position plus corresponding frequency response function to
Amount.Then, for each positionIntroduce a frequency response function vectorThe vector provides later.Then,
Find out following output function
It is as follows to choose objective function
The set for some frequencies that wherein I chooses, generally comprises the frequency of signal, does not include the frequency of noise.So build
Vertical optimization problem is as follows
The solution of the problem is denoted asThe sound source position estimated value that it is just to solve for.
For any sound source position rk, k=1 ..., P obtain estimated value by solving optimization problem (19), then fixed
Position error can be denoted as
Position error is smaller, then positioning accuracy is higher, corresponding to realize that requirement is also higher.
5. the design of frequency response function vector
It is needed it can be seen from optimization problem (19) to each feasible sound source positionDesign corresponding frequency response
Functional vector.In general, there is the method much designed, but based on the considerations of guaranteeing positioning accuracy request, it is necessary to take pole
Big minimum design.
Firstly, for any given positionNeed to define a pass band areasWith a stop band region Must include It must be withSeparation, there is an intermediate zone in centre.Based on examining for raising positioning accuracy
Consider,Directly it is taken asThis point,It is all regions for being more than or equal to position error at a distance from r, i.e.,
Then, an ideal receptance function G is definedd(r, f) is as follows:
Wherein, rcIt is a reference position point, is generally taken as the center of microphone array.Establish following optimization problem
Wherein,
By solving the problem it can be concluded that optimal solution, is then denoted as
This is a Minimax Problems, can be converted into semi-infinite programming problem solution.
6. positioning accuracy request
For any given r, remember that the performance number of optimization problem (13) is
Provide the following conditions:
Condition 1: for arbitraryCorresponding performance numberAll meet
When condition 1 is satisfied, and at least there is a frequency and meet | X1(r1, f) | when > 0, the precision of positioning must
To guarantee.The supplementary set for defining stopband is as follows
So, we obtain to draw a conclusion
The position r of i.e. main sound source1It certainly will not be in the optimal solution of problem (19)Stopband in.It enables
Wherein S is the region of entire feasible location,Definition be
It can so be obtained by (10) and (11)
It can be found out by (22) for arbitraryPosition error meets
The error for finding out the actual location of the position estimation value and main sound source come does not exceed ε.It should be noted that ε
It is the worst error under worst condition.In general, actual error can be much smaller.
7. the specific implementation and requirement of method
The selection of 7.1 frequencies
In objective function (18), need to choose some frequencies.The selection of frequency needs to meet X1(r1, f) > 0, i.e., main sound
Source is not static on that frequency.If such frequency exists, it is only necessary to which a frequency can meet condition.But due to
The wideband of voice signal, it would be desirable to provide a series of frequency range it can be selected that making within this range, at least
One frequency can satisfy this condition.This is readily satisfied very much, it is only necessary to be selected at random in the frequency range of voice
Take a series of frequency.
The selection of 7.2M
The value of M is determined by the requirement of practical application.If main sound source position has loudspeaker, then M can be obtained very greatly,
It is corresponding to realize that difficulty reduce.If, then M cannot be obtained very greatly, the difficulty of realization can increase accordingly without loudspeaker.
The specific design of 7.3 filter freguency response functional vectors
It is sliding-model control first.Design for filter freguency response functional vector, problem (13) correspond to appoint
Anticipate a feasible location point r's, it is therefore desirable to solve to unlimited problem (13).This is impossible.Based on this,
It only needs to carry out sufficient sliding-model control to area of feasible solutions S, obtains discrete point set Sd.Discretization is denser, precision
Higher, the frequency response function vector of solution is also more.
Since filter freguency response functional vector needs to obtain by solving optimization problem, runing time is not generally low,
This is unfavorable for positioning in real time.Therefore, all filter freguency response functional vectors can be first found out offline, then saved
Come.It recalls out when there are voice location tasks.
7.4 receive the interpolation of the Fourier transform of signal and objective function
For calculating target function (18), for the signal that each microphone receives, need to do it Fourier change
It changes, the Fourier transform of such as i-th of microphone received signal is Yi(r1,…,rp,f)。
After discretization, problem (13) is only limited, needs to find out corresponding frequency using optimization method to each problem
Rate receptance function vector and performance number.For other points of area of feasible solutions S, frequency response function vector is not defined, accordingly
Target function value also do not define.But have target function value in discrete point, then take interpolating function
Wherein, J is interpolating function.The selection of interpolating function do not need it is too complicated, it is linear to insert after discretization is sufficiently dense
The precision of value function is sufficiently high, and is easy to calculate.
The selection of 7.5 positioning accuracies and microphone number
There is no specific requirements for the number of microphone, but must satisfy condition 1.Positioning accuracy is higher, the choosing of position error
Take just smaller, this is needed depending on the demand of practical application, but condition 1 is just more difficult to meet.The positioning given for one
The number of error, microphone is more, and condition 1 more readily satisfies.If microphone is increased monotonically in the constant situation in position, that
Performance number is with regard to monotonic decreasing.Therefore, when condition 1 is unsatisfactory for, the number of increase microphone, constantly reduction performance can be passed through
Value can satisfy until condition 1.Because the ideal minimum value of performance number is zero, by being continuously increased the number of microphone,
Condition 1 is certain to meet.But the number of microphone is more, cost is higher, thus only need to increase microphone number until
Condition 1 is just met for.
Therefore, positioning accuracy is higher, and number needed for microphone is more, and the difficulty of realization is bigger.The adjustment of this step
It is extremely important, it is to realize pinpoint guarantee, concrete condition needs depending on actual requirement.
7.6 location algorithm
It when realizing positioning in real time, needs to solve problem (19), therefore the solution of the problem is needed very
Fastly.Due to taking sliding-model control, two stages can be divided into solve.As shown in Fig. 2, the area of feasible solutions S of position from
Become S after dispersiond, then it is divided to following two step.
Step 1: in discrete point set SdUpper solution, i.e.,
The problem can find out optimal solution r by maximizing direct solution*d。
Step 2: in r*dNear region N (r*d) on solve, i.e.,
Due to N (r*d) very little, which can be solved by further discretization.
The optimization calculating of this two step is very simple, can be realized Real-time solution and positioning, as shown in Figure 2.
Case:
It is in an inner feasible point region S of room [- 2m, 2m] × [- 2m, 2m] × [- 3m, 3m]
S={ (x, y, z): x ∈ [- 2m, 2m], y ∈ [- 2m, 2m], z=1.2m }
Discrete point set SdIt is chosen for
(x, y, z) x=-1.98+0.04*k, y=-1.98+0.04*l, k, l=0 ..., 99, z=1.2m }
Therefore it needs to establish 10000 problems and solves filter freguency response functional vector.It chooses the position of microphone
For
(x, y, z) x=-1.8+0.3*k, y=-1.8+0.3*l, z=2.7m, k=0,1 ..., 12 }
Considering five kinds of microphone array situations, number is N=9 respectively, 13,17,21,25, as shown in figure 3, z=2.7m.
Parameter value M is set as M=7,11,15,19,23,25 these types situation.Frequency is chosen to be 500Hz.Error with
0.01m is unit, finds out position error such as the following table 1 (unit m) of the condition of satisfaction 1.It can be seen that M increases or microphone number
When increase, the position error for meeting condition 1 is smaller, and the precision of realization is higher.
Table 1 meets the position error of condition 1
When needing to position, consider that M >=9 and microphone number are 25 situation.Selection precision is 0.6m, by verifying, it
Meet condition 1, therefore the accurate positioning within the scope of 0.6m may be implemented.Such as source position be selected as at random (0.6606m ,-
1.3456m), can to find out objective function as shown in Figure 4.And it is by the estimated value that optimization algorithm finds out source position
(0.6383m, -1.2903m), closely, position error only has 0.0596m.
It for source position, is arbitrarily chosen in feasible location, can find out position error and checks whether to meet precision
It is required that.Present case is tested using 100 random sites, the accurate positioning within the scope of realization 0.6m, and is Real-time
It is existing.Fig. 5 is main flow of the present invention.
Finally, it is stated that preferred embodiment above is only used to illustrate the technical scheme of the present invention and not to limit it, although logical
It crosses above preferred embodiment the present invention is described in detail, however, those skilled in the art should understand that, can be
Various changes are made to it in form and in details, without departing from claims of the present invention limited range.
Claims (6)
1. the voice accurate positioning method under multi-source environment, it is characterised in that: method includes the following steps:
S1: voice positioning is realized using microphone array system；
S2: the voice delivery function of environment is tested；
S3: a pinpoint condition is exported according to actual environment, and is set using the frequency response function vector of minimax
Meter and adjustment and the method for increasing microphone meet condition；
S4: establishing the optimization problem of voice positioning, and develop real time algorithm and solve to optimization problem, realizes in given positioning
Accurate positioning under precision.
2. the voice accurate positioning method under multi-source environment according to claim 1, it is characterised in that: in the step S1
In, the position of microphone and the position of sound source be not be overlapped, is placed in the top of sound source.
3. the voice accurate positioning method under multi-source environment according to claim 1, it is characterised in that: the step S2 tool
Body are as follows:
Environment locating for sound source is near field model and far field model, and what far field model considered is the directional information of sound source, near field mould
What type considered is the location information of sound source；What this method considered is location information, is not limited near field model or far field model；
When the environment of sound source has given, the transmission function of voice determines that, it is expressed as
T(r,q,f) (1)
Wherein r is the position of sound source, and q is the position of microphone, and it is a complex function, expression sound that f, which is the frequency of the signal function,
Variation about frequency f of the sound from source position r to target position q, then have
Y (q, f)=X (r, f) T (r, q, f), i=1 ..., N (2)
Wherein, X (r, f) is the Fourier transform of source signal, and Y (q, f) is the Fourier transform for receiving signal；
Determination for transmission function T (r, q, f), is obtained by real-time testing；Many group voice signal X are played on the r of positionk
(r, f), k=1 ..., m, and the signal Y received is enrolled in position qk(r, f), k=1 ..., m, then complex function T (r, q,
F) it is decomposed into u (f)+jv (f)；According to signal characteristic, some frequency sets are selected, for each frequency f, consider following optimization
Problem
Optimal solution is found out in real time to the problem；It does not need to solve each frequency this optimization problem, it is only necessary to emerging to feeling
The frequency of interest solves this optimization problem；Possible position and each microphone to each sound source receive position and do test solution
Afterwards, gained optimal solution forms transmission function.
4. the voice accurate positioning method under multi-source environment according to claim 1, it is characterised in that: in step s3,
It is described that a pinpoint condition is exported according to actual environment specifically:
Voice is positioned, will not include that noise positions to institute's sound source, it is only necessary to be positioned against main sound source, main sound
The intensity in source needs the intensity than other sound sources big, just can determine that it is main sound source, otherwise the environment is considered as noise circumstance；
Give P sound source Xk(rk, f), k=1 ..., P, wherein rkIndicate the position of k-th of sound source；Main sound source is denoted as X1, other
P-1 sound source Xk, k=2 ..., P are interference sound sources, are considered as noise；According to the requirement of main sound source, there is following agreement
Wherein M >=1, frequency only need inside some frequency range where voice；
For the value of M depending on the requirement of practical application, M is bigger, and the requirement to environment is stringenter, and it is easier to realize；M is got over
Small, the range of application is wider, and corresponding cost of implementation is also higher.
5. the voice accurate positioning method under multi-source environment according to claim 1, it is characterised in that: in step s3,
It is specific that the method using the frequency response function vector design of minimax and adjustment or increase microphone meets condition
Are as follows:
For the microphone array of N number of element, the position of microphone is denoted as qi, i=1 ..., N；The then reception letter of i-th of microphone
Number it is
Wherein YikFor the part about k-th of sound source that i-th of microphone receives, i.e.,
Yik(rk, f) and=Xk(rk,f)T(rk,qi, f), i=1 ..., N (6)
For each microphone, behind connect a filter, after adding filter, corresponding filter freguency response function is
Hi(f), i=1 ..., N, by synthesizing an output signal after filter
Remember that G (r, f) is that output response function is as follows
Wherein T (r, f)=[T (r, q1,f),…,T(r,qN,f)]T, H (f)=[H1(f),…,HN(f)]T；So,
Firstly, for any given feasible locationNeed to define a pass band areasWith a stop band region Must include It must be withSeparation, there is an intermediate zone in centre；Based on the considerations of improving positioning accuracy,Directly it is taken asThis point,Be it is all withDistance be more than or equal to position error region, i.e.,
Then, an ideal receptance function G is definedd(r, f) is as follows:
Wherein, rcIt is a reference position point, takes the center of microphone array；Establish following optimization problem
Wherein,
It is denoted as by solving the optimal solution that the problem obtains
For any given positionJust there is one to correspond toOptimization problem (13), remember that the performance number of this problem is
Provide the following conditions:
Condition 1: for arbitraryCorresponding performance numberAll meet
6. the voice accurate positioning method under multi-source environment according to claim 5, it is characterised in that: the step S4 tool
Body are as follows:
When condition 1 is satisfied, and at least there is a frequency and meet | X1(r1, f) | when > 0, the precision of positioning is just protected
Card；
As soon as microphone system exports a function (9) for a given frequency response function vector H (f)；For positioning is believed
Breath is added, and is needed according to the information of position plus corresponding frequency response function vector；Then for each positionIt introduces
One frequency response function vectorFind out following output function
It is as follows to choose objective function
The set for some frequencies that wherein I chooses；It is as follows to establish optimization problem
The solution of the problem is denoted as r1, that is, the sound source position estimated value that solves；
For any sound source position rk, k=1 ..., P obtain estimated value, position error note by solving optimization problem (2.9)
For
The supplementary set for defining stopband is as follows
So, just there is conclusion
The position r of i.e. main sound source1It certainly will not be in the optimal solution of problem (19)Stopband in；It enables
Wherein S is the region of entire feasible location,Definition be
So obtained by (10) and (11)
It is found out by (22) for arbitrary r1∈ S, position error meet
The error for finding out the position estimation value and real sound source position come does not exceed ε.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810926869.7A CN109061567B (en) | 2018-08-15 | 2018-08-15 | Voice accurate positioning method under multi-source environment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810926869.7A CN109061567B (en) | 2018-08-15 | 2018-08-15 | Voice accurate positioning method under multi-source environment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109061567A true CN109061567A (en) | 2018-12-21 |
| CN109061567B CN109061567B (en) | 2022-08-09 |
Family
ID=64678419
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810926869.7A Active CN109061567B (en) | 2018-08-15 | 2018-08-15 | Voice accurate positioning method under multi-source environment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109061567B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114720943A (en) * | 2022-06-06 | 2022-07-08 | 深圳市景创科技电子股份有限公司 | Multi-channel sound source positioning method and system |
| WO2023000206A1 (en) * | 2021-07-21 | 2023-01-26 | 华为技术有限公司 | Speech sound source location method, apparatus and system |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104715758A (en) * | 2015-02-06 | 2015-06-17 | 哈尔滨工业大学深圳研究生院 | Branched processing array type speech positioning and enhancement method |
| CN105068048A (en) * | 2015-08-14 | 2015-11-18 | 南京信息工程大学 | Distributed microphone array sound source positioning method based on space sparsity |
| CN106646376A (en) * | 2016-12-05 | 2017-05-10 | 哈尔滨理工大学 | P-norm noise source positioning identification method based on weight correction parameter |
| CN107271963A (en) * | 2017-06-22 | 2017-10-20 | 广东美的制冷设备有限公司 | The method and apparatus and air conditioner of auditory localization |
| CN108318862A (en) * | 2017-12-26 | 2018-07-24 | 北京大学 | A kind of sound localization method based on neural network |
-
2018
- 2018-08-15 CN CN201810926869.7A patent/CN109061567B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104715758A (en) * | 2015-02-06 | 2015-06-17 | 哈尔滨工业大学深圳研究生院 | Branched processing array type speech positioning and enhancement method |
| CN105068048A (en) * | 2015-08-14 | 2015-11-18 | 南京信息工程大学 | Distributed microphone array sound source positioning method based on space sparsity |
| CN106646376A (en) * | 2016-12-05 | 2017-05-10 | 哈尔滨理工大学 | P-norm noise source positioning identification method based on weight correction parameter |
| CN107271963A (en) * | 2017-06-22 | 2017-10-20 | 广东美的制冷设备有限公司 | The method and apparatus and air conditioner of auditory localization |
| CN108318862A (en) * | 2017-12-26 | 2018-07-24 | 北京大学 | A kind of sound localization method based on neural network |
Non-Patent Citations (1)
| Title |
|---|
| ZHI GUO FENG等: "The design of multi-dimensional acoustic beamformers via window functions", 《DIGITAL SIGNAL PROCESSING》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2023000206A1 (en) * | 2021-07-21 | 2023-01-26 | 华为技术有限公司 | Speech sound source location method, apparatus and system |
| CN114720943A (en) * | 2022-06-06 | 2022-07-08 | 深圳市景创科技电子股份有限公司 | Multi-channel sound source positioning method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109061567B (en) | 2022-08-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106093864B (en) | A kind of microphone array sound source space real-time location method | |
| US20210219053A1 (en) | Multiple-source tracking and voice activity detections for planar microphone arrays | |
| US9706298B2 (en) | Method and apparatus for localization of an acoustic source and acoustic beamforming | |
| US20170365255A1 (en) | Far field automatic speech recognition pre-processing | |
| US5581620A (en) | Methods and apparatus for adaptive beamforming | |
| CN104254819A (en) | Audio user interaction recognition and context refinement | |
| CN103181190A (en) | Systems, methods, apparatus, and computer-readable media for far-field multi-source tracking and separation | |
| CN107167770B (en) | A kind of microphone array sound source locating device under the conditions of reverberation | |
| Brendel et al. | Distributed source localization in acoustic sensor networks using the coherent-to-diffuse power ratio | |
| Sivasankaran et al. | Keyword-based speaker localization: Localizing a target speaker in a multi-speaker environment | |
| CN109061567A (en) | Voice accurate positioning method under multi-source environment | |
| CN113126028A (en) | Noise source positioning method based on multiple microphone arrays | |
| CN103901400B (en) | A kind of based on delay compensation and ears conforming binaural sound source of sound localization method | |
| CN109212481A (en) | A method of auditory localization is carried out using microphone array | |
| Xu et al. | Underwater acoustic source localization method based on TDOA with particle filtering | |
| Pan et al. | On the design of target beampatterns for differential microphone arrays | |
| CN115267671A (en) | Distributed voice interaction terminal equipment and sound source positioning method and device thereof | |
| CN113314138B (en) | Sound source monitoring and separating method and device based on microphone array and storage medium | |
| Cho et al. | Sound source localization for robot auditory systems | |
| CN112363112B (en) | Sound source positioning method and device based on linear microphone array | |
| CN102438191A (en) | Method for acoustic signal tracking | |
| Nakano et al. | Automatic estimation of position and orientation of an acoustic source by a microphone array network | |
| Song et al. | Auditory scene analysis-based feature extraction for indoor subarea localization using smartphones | |
| JP3862685B2 (en) | Sound source direction estimating device, signal time delay estimating device, and computer program | |
| Matsumoto et al. | Multiple signal classification by aggregated microphones |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |