CN108828524A - Particle filter audio source tracking localization method based on Delaunay Triangulation - Google Patents
Particle filter audio source tracking localization method based on Delaunay Triangulation Download PDFInfo
- Publication number
- CN108828524A CN108828524A CN201810560142.1A CN201810560142A CN108828524A CN 108828524 A CN108828524 A CN 108828524A CN 201810560142 A CN201810560142 A CN 201810560142A CN 108828524 A CN108828524 A CN 108828524A
- Authority
- CN
- China
- Prior art keywords
- particle
- delaunay
- point
- triangle
- sound
- 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
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Circuit For Audible Band Transducer (AREA)
Abstract
The invention proposes disclose a kind of particle filter audio source tracking localization method based on Delaunay Triangulation, this method can be under conditions of oneself knows interior space size, merely with the smart phone worn with speaker, by carrying out Delaunay Triangulation to space, utilize Delaunay Triangulation result, particle random in indoor physical space is carried out to orderly association, theory analysis based on sound failure model, determine the particle search range near moving sound, and search range set valuation is updated in particle filter target following frame, the positioning of indoor scene speaker tracking is carried out using enhancing particle filter algorithm.The localization method has certain noise immunity, anti-reverberation and robustness.The method of the present invention is applicable not only to the room sound field environment of regular shape, is also applied for the speaker's positioning and tracking of the room sound field environment of non-regular shape.
Description
Technical field
It is specifically a kind of to be based on voice sound signal failure model and Delaunay tri- the present invention relates to indoor field of sound source location
For angle subdivision as a result, determining sound source scope of activities, utilizing, which enhances particle filter algorithm, carries out the positioning of indoor scene speaker tracking
Method has important application value in the scenes such as human-computer interaction, teleconference, location-based service.
Background technique
It is 300Hz-3400Hz that international communication standards, which provide the normal audible frequency range of people,.Indoors in sound field, distance
The sound pressure level (SPL) of normal voice sound source 20cm is 72dB, and the SPL that 1 meter of distance is 58dB.The energy attenuation of voice signal with connect
The positive correlation of distance is received, and sound frequency is higher, sound energy attenuation is faster.
Oneself has some measuring and calculating based on acoustic energy to realize the scheme of auditory localization at present, but requires to build wireless network greatly
Network anchor point is carried out by the way of internode collaboration, and relies on the priori conditions of room sound field, such as：Sound reception end is placed in
In the specific region of wireless sensor network, space traversal complexity is reduced, improves system availability；Or under special scenes
Minimum reception and transmission range sets SPL minimum threshold, by comparing the SPL size between cooperative node, solves spatial data relevant issues,
Improve positioning system accuracy etc..However, such work and underusing the statistics of room sound field sound energy probability density distribution
The theory advantage of rule and sound failure model, and in the environment for having reverberation and noise, acoustic source location accuracy cannot be guaranteed.
By the sound energy probability density distribution modeling to the interior space, indoor acoustic density distribution and acoustic propagation distance are established
Mathematical relationship, both using sound failure model formula calculate acoustic receiver position sound energy, also can be applied to solve indoor sound
Field distribution and identification problem.Therefore, the modeling of sound energy probability density distribution is to calculate the indoor speaker of realization based on sound to position
Key link, the superiority and inferiority of modeling method then directly influence the accuracy and robustness of location algorithm.In order to reduce location algorithm
Complexity improves the real-time of locating scheme, and indoor physical space modeling problem is converted complex-curved subdivision problem by we,
It is multiple simple polyhedrons by interior space subdivision, the inspection of sound active regions is then carried out using these simple polyhedrons
It surveys, for the tracking iteration of each sound source position, need to only adjust the number of nodes of parallel processing, sound wave active region can be made
Domain detection is completed in tolerable time range, can be reduced detection search range, be reduced algorithm complexity, it is real-time to improve tracking
Property, and this subdivision process only need to be pre-processed once, in audio source tracking process, not need Repeated.
Summary of the invention
For in indoor complex sound field environment, noise jamming is big, reverberation influences seriously, indoor speaker tracking to be caused to position
The still insufficient problem of precision, the particle filter audio source tracking positioning based on Delaunay Triangulation that the invention proposes a kind of
Method realizes that indoor speaker is self-positioning and tracks.This method only needs to carry out a Delaunay tri- to indoor physical space
Subdivision result is merged with each stage Search range of particle filter target following frame, solves traditional grain by angle subdivision
Because of the problem that large area resampling causes operand huge in sub- filtering algorithm, and by the smart phone worn with human body with
The short distance relationship of voice point source of sound, so that the locating scheme has certain noise immunity, anti-reverberation and robustness.
The present invention is based on the particle filter audio source tracking localization methods of Delaunay Triangulation, include the following steps：
S1. speaker uses the voice signal for the smart phone recording current location worn with it, as target sound energy；
S2. target sound energy feature is calculated based on voice sound signal failure model；
S3. Delaunay Triangulation is carried out to the interior space；
S4. using enhancing particle filter to target following, particle is initialized by filtering algorithm, in conjunction with Delaunay tri-
Angle subdivision is as a result, convert Delaunay triangular pyramids for the range select permeability in each stage in particle filter target following frame
The select permeability of node determines grain instantly by finding the adjacent Delaunay triangular apex collection in last moment target position
The search range (i.e. sound source enlivens range) of son, obtains similarity between particle and target；
S5. the similarity of each particle obtained and target, is weighted and averaged, and obtains the estimated value of coordinates of targets；
S6. target position 3D coordinate is exported, realizes the real-time tracking positioning of moving sound (speaker).
Sound can indicate that therefore, target sound energy feature is exactly the vector V of a N*1 with a vector.
Voice sound, also known as speech sound are that the mankind are issued by phonatory organ, the sound for social communication.By sounding function
It can divide, the phonatory organ of people can be divided into three driving source, sounding body and acoustic resonator parts.
S2. the calculating target sound energy feature, the specific method is as follows：
Assuming that sound source is issued by the mouth of people, if being s (n) by the voice sound that mouth issues, what mobile microphone received
Voice signal is x (n), and room noise is that the additive white noise of zero-mean is v (n), then the relationship of this three is：
X (n)=s (n)+v (n) (1)
Wherein, γ is the gain factor of mobile microphone, and a (n- τ) is by the acoustic density of the sent out voice of mouth of people, τ
It is the propagation delay time from mouth to mobile microphone, ps(n- τ) is the vector of a 3*1, indicates mouth position；pxIt is also
The vector of one 3*1 indicates mobile microphone position；dsc=| | ps(n-τ)-px| |, it is set as a constant value；It is right
(1) formula both sides calculate average energy value, obtain：
E[x2(n)]=E [s2(n)]+E[v2(n)] (3)
Wherein
In above formula, g=γ2,As it is assumed that dsxIt is a constant value, then from sounding position
Set microphone position apart from very little, then τ can be ignored, therefore, a (n- τ) ≈ a (n), ps(n-τ)≈ps(n),
Since average energy value is carried out in the time window of a length of T, then when sample frequency is fsWhen, M=T*fsIt is to be used for
The average number of sampling points of energy；
If E [x2(n)]=yx(t), E [a2(n)]=ys(t), E [v2(n)]=∈ t, then based on the target sound of sound failure model
Energy feature is represented by：
yx(t)=ys(t)+∈(t) (5)
Wherein,If M is sufficiently large, according to central-limit theorem, ∈ (t) approaches normal state point
Cloth：σ2It is the variance of noise v (n).
Delaunay subdivision is a kind of special triangulation, needs to meet empty circle property and maximizes two standards of Minimum Internal Angle
Then.There are many kinds of the algorithms for realizing Delaunay Triangulation, and present invention employs most popular Bowyer-Watson at present to calculate
Method.
S3. described that Delaunay Triangulation is carried out to the interior space, using Bowyer-Watson algorithm, the master of algorithm
Want that steps are as follows：
(1) triangular is constructed, includes all scatterplots, is put into triangle chained list；
(2) scatterplot concentrated is sequentially inserted into, the triangle that its circumscribed circle includes insertion point is found out in triangle chained list
Shape, referred to as influence triangle, delete the common edge of the influence triangle, and insertion point is connected with the whole vertex for influencing triangle
It picks up and, to complete insertion of the point in Delaunay triangle chained list；
(3) it is optimized according to the triangle that Lawson Optimality Criteria newly forms part, specific step is as follows：
(3.1) two triangles with common edge are synthesized into a polygon；
(3.2) it is conducted a survey with largest empty circle criterion, sees its 4th vertex whether in the circumscribed circle of triangle；
(3.3) if amendment diagonal line exchanges diagonal line, that is, completes the processing of local optimization procedure, will be formed
Triangle be put into Delaunay triangle chained list；
(3.4) circulation executes above-mentioned (3.2) step, until the insertion of all scatterplots finishes.
Particle filter algorithm is a kind of statistical filtering method based on Monte Carlo and recursion Bayes, relatively other to be based on
For the target tracking algorism of filtering theory, it is more suitably applied to non-linear, non-gaussian motion model.Its basic thought is：
Target is predicted by constructing target movement model using the particle sample representation dbjective state generated at random in a certain range
Real time kinematics situation is estimated and is corrected to dbjective state in conjunction with observation model, realizes target following.
S4. described using particle filter is enhanced to target following, it mainly include following four process：
(1) particle sampler initializes
Particle sampler is M particle appearance being randomly generated in observed object areas adjacent, utilizes Delaunay subdivision pair
These particles are handled, and are allowed to form orderly point set, i.e. Delaunay triangular apex set
V={ vi|vi∈[Lx, Ly, Lz], i=1 ..., M }；
(2) search phase
Firstly, the target position posrition obtained to last momentt, carry out in Delaunay triangular apex set
Search judges that it is concentrated with the presence or absence of point；IfThen by positiontAccording to Delaunay Triangulation
Algorithm is inserted into Optimization Steps, is newly added in a point-to-point collection V, will not influence original V collection distribution, and new insertion point
positiontThe judgement of Delaunay criterion only is carried out with the point near its position, finds and constitutes Delaunay triangle with it
All triangular apex do not need to traverse entire V collection, and computation complexity is lower；If positiont∈ V, then directly
To point positiontNear, find all triangular apex that Delaunay triangle is constituted with it；
Secondly, completing to point positiontRange searching after, available point positiontRange set Vk={ vk|
vk∈ V, k=1 ..., N }, wherein vkIt indicates and point positiontDelaunay vertex of a triangle is constituted, N indicates point
positiontConstitute Delaunay vertex of a triangle number；
(3) particle weight is calculated
If search range set VkInterior Arbitrary Particles vkSound energy size bePosition coordinates areDue to mouth position ps(n- τ) distance is dsxMicrophone position pxSound energy
Size is yx(t), it enablesIndicate VkMiddle Arbitrary ParticlesvK's arrives microphone position
Distance, then according to sound pressure level with range attenuation formula, it is availableExpression formula is as follows：
The range attenuation characteristic propagated in space according to sound wave calculates particle weight, i.e. particle sound energy feature and target
The normalization similitude of sound energy feature
At this point, being weighted and averaged according to the similarity of search phase obtained each particle and target, so that it may be worked as
Preceding moment t, the estimated value of coordinates of targets
(4) resampling
Resampling, be in order to solve in particle filter object tracking process, sample degeneracy phenomenon and design, need to grain
Son normalization weight is adjusted.In the present invention, the resampling stage according to previous step weight computing as a result, utilizing each particle
SimilarityJudge number of effective particles amount PeffctiveWith threshold population PthresholdSize；
Work as Peffctive< PthresholdWhen, illustrate that current particle is degenerated seriously, needs to carry out resampling；The high ground of similarity
Side expands Delaunay point set search range, that is, the Delaunay triangular apex for increasing the particle is concentrated to search range point,
Form new search range point set；And the place that similarity is low, it is searched for without Delaunay point set, to form new search
The point for being unsatisfactory for cum rights particle the power grain such as is mapped as the Particle tracking range at subsequent time t+1 moment by range point set
Son；
Work as Peffctive> PthresholdWhen, illustrate that current particle is in good condition, does not need to update search range point set, particle
State is updated and is saved.
The invention proposes in a kind of complex sound field environment indoors, quickly and effectively realize indoor speaker tracking positioning
Method.This method, merely with the smart phone worn with speaker, can pass through under conditions of oneself knows interior space size
Delaunay Triangulation is carried out to space, effectively selects effective particle, determines that sound source enlivens range, realizes and said in indoor environment
Talk about the self-positioning of people position and tracking.This method solve conventional particle filtering algorithm because each moment carries out resampling, with
And the resampling stage scans for all particles range, causes algorithm complexity big, the problem of real-time tracking effect difference is also kept away
The drawbacks of current most of indoor auditory localization schemes rely on special equipment is exempted from.Do not increasing particle filter algorithm complexity
And in the case where not depending on infrastructure, the sound failure model and principle of effective use normal speech signals environmental dissemination indoors,
Improve tracking and positioning performance of the speaker indoors under complex sound field environment.The method of the present invention is applicable not only to the room of regular shape
Interior sound field environment is also applied for the speaker's positioning and tracking of the room sound field environment of non-regular shape.
Detailed description of the invention
Fig. 1 is that the present invention is based on the particle filter audio source tracking localization method block diagrams of Delaunay Triangulation.
Fig. 2 is the X-Y scheme of embodiment interior space Delaunay Triangulation.
Fig. 3 is the subdivision graph that a point A is newly added in original triangulation result for embodiment.
Specific embodiment
The content of present invention is further described below with reference to embodiment and attached drawing, but is not limitation of the invention.
Embodiment
Referring to Fig.1, the present invention is based on the particle filter audio source tracking localization method of Delaunay Triangulation, including it is as follows
Step：
S1. speaker uses the voice signal for the smart phone recording current location worn with it, as target sound energy；
S2. target sound energy feature is calculated based on voice sound signal failure model；
S3. Delaunay Triangulation is carried out to the interior space；
S4. using enhancing particle filter to target following, particle is initialized by filtering algorithm, in conjunction with Delaunay tri-
Angle subdivision is as a result, convert Delaunay triangular pyramids for the range select permeability in each stage in particle filter target following frame
The select permeability of node determines grain instantly by finding the adjacent Delaunay triangular apex collection in last moment target position
The search range (i.e. sound source enlivens range) of son, obtains similarity between particle and target；
S5. the similarity of each particle obtained and target, is weighted and averaged, and obtains the estimated value of coordinates of targets；
S6. target position 3D coordinate is exported, realizes the real-time tracking positioning of moving sound (speaker).
S2. the calculating target sound energy feature, the specific method is as follows：
Assuming that sound source is issued by the mouth of people, if being s (n) by the voice sound that mouth issues, what mobile microphone received
Voice signal is x (n), and room noise is that the additive white noise of zero-mean is v (n), then the relationship of this three is：
X (n)=s (n)+v (n) (1)
Wherein, γ is the gain factor of mobile microphone, and a (n- τ) is by the acoustic density of the sent out voice of mouth of people, τ
It is the propagation delay time from mouth to mobile microphone, ps(n- τ) is the vector of a 3*1, indicates mouth position；pxIt is also
The vector of one 3*1 indicates mobile microphone position；dsx=| | ps(n-τ)-px| |, it is set as a constant value；It is right
(1) formula both sides calculate average energy value, obtain：
E[x2(n)]=E [s2(n)]+E[v2(n)] (3)
Wherein
In above formula, g=γ2,As it is assumed that dsxIt is a constant value, then from sounding position
Set microphone position apart from very little, then τ can be ignored, therefore, a (n- τ) ≈ a (n), ps(n-τ)≈ps(n),
Since average energy value is carried out in the time window of a length of T, then when sample frequency is fsWhen, M=T*fsIt is to be used for
The average number of sampling points of energy；
If E [x2(n)]=yx(t), E [a2(n)]=ys(t), E [v2(n)]=∈ (t), the then target based on sound failure model
Sound energy feature is represented by：
yx(t)=ys(t)+∈(t) (5)
Wherein,If M is sufficiently large, according to central-limit theorem, ∈ (t) is approached just
State distribution：σ2It is the variance of noise v (n).
S3. described that Delaunay Triangulation is carried out to the interior space, using Bowyer-Watson algorithm, the master of algorithm
Want that steps are as follows：
(1) triangular is constructed, includes all scatterplots, is put into triangle chained list；
(2) scatterplot concentrated is sequentially inserted into, the triangle that its circumscribed circle includes insertion point is found out in triangle chained list
Shape, referred to as influence triangle, delete the common edge of the influence triangle, and insertion point is connected with the whole vertex for influencing triangle
It picks up and, to complete insertion of the point in Delaunay triangle chained list；
(3) it is optimized according to the triangle that Lawson Optimality Criteria newly forms part, specific step is as follows：
(3.1) two triangles with common edge are synthesized into a polygon；
(3.2) it is conducted a survey with largest empty circle criterion, sees its 4th vertex whether in the circumscribed circle of triangle；
(3.3) if amendment diagonal line exchanges diagonal line, that is, completes the processing of local optimization procedure, will be formed
Triangle be put into Delaunay triangle chained list；
(3.4) circulation executes above-mentioned (3.2) step, until the insertion of all scatterplots finishes.
For oneself knows indoor environment, if its a length of Lx, width Ly, a height of Lz, then in Lx*Ly*LzIt is carried out in range
The setting of random 50 sampling points, the subdivision result of two-dimensional surface are as shown in Figure 2.
S4. described using particle filter is enhanced to target following, it mainly include following four process：
(1) particle sampler initializes
Particle sampler is M particle appearance being randomly generated in observed object areas adjacent, utilizes Delaunay subdivision pair
These particles are handled, and are allowed to form orderly point set, i.e. Delaunay triangular apex set
V={ vi|vi∈[Lx, Ly, Lz], i=1 ..., M }；
(2) search phase
Firstly, the target position position obtained to last momentt, searched in Delaunay triangular apex set
Rope judges that it is concentrated with the presence or absence of point；IfThen by positiontAccording to the calculation of Delaunay Triangulation
Method is inserted into Optimization Steps, as a result as shown in figure 3, being newly added in a point A to point set V, will not influence original V collection point
Cloth, and new insertion point positiontThe judgement of Delaunay criterion is only carried out with the point near its position, searching is constituted with it
All triangular apex of Delaunay triangle do not need to traverse entire V collection, and computation complexity is lower；If
positiont∈ V, then directly to point positiontNear, find all triangles that Delaunay triangle is constituted with it
Vertex；
Secondly, completing to point positiontRange searching after, available point positiontRange set Vk={ vk|
vk∈ V, k=1 ..., N }, wherein vkIt indicates and point positiontDelaunay vertex of a triangle is constituted, N indicates point
positiontConstitute Delaunay vertex of a triangle number；
(3) particle weight is calculated
If search range set VkInterior Arbitrary Particles vkSound energy size bePosition coordinates areDue to mouth position ps(n- τ) distance is dsxMicrophone position pxSound energy
Size is yx(t), it enablesIndicate VkMiddle Arbitrary Particles vkArrive microphone position
Distance, then according to sound pressure level with range attenuation formula, it is availableExpression formula is as follows：
The range attenuation characteristic propagated in space according to sound wave calculates particle weight, i.e. particle sound energy feature and target
The normalization similitude of sound energy feature
At this point, being weighted and averaged according to the similarity of search phase obtained each particle and target, so that it may be worked as
Preceding moment t, the estimated value of coordinates of targets
(4) resampling
According to previous step weight computing as a result, using each particle similarityJudge number of effective particles amount Peffctive
With threshold population PthresholdSize；
Work as Peffctive< PthresholdWhen, illustrate that current particle is degenerated seriously, needs to carry out resampling；The high ground of similarity
Side expands Delaunay point set search range, that is, the Delaunay triangular apex for increasing the particle is concentrated to search range point,
Form new search range point set；And the place that similarity is low, it is searched for without Delaunay point set, to form new search
The point for being unsatisfactory for cum rights particle the power grain such as is mapped as the Particle tracking range at subsequent time t+1 moment by range point set
Son；
Work as Peffctive> PthresholdWhen, illustrate that current particle is in good condition, does not need to update search range point set, particle
State is updated and is saved.
The present invention is based on the particle filter audio source tracking localization methods of Delaunay Triangulation, are embodied in utilization
Particle random in indoor physical space as a result, is carried out orderly association, based on sound failure model by Delaunay Triangulation
Theory analysis determines the particle search range near moving sound, and search range set valuation is updated to particle filter
In target following frame, the positioning of indoor scene speaker tracking is carried out using enhancing particle filter algorithm.The localization method has
Certain noise immunity, anti-reverberation and robustness.
Claims (5)
1. the particle filter audio source tracking localization method based on Delaunay Triangulation, which is characterized in that include the following steps：
S1. speaker uses the voice signal for the smart phone recording current location worn with it, as target sound energy；
S2. target sound energy feature is calculated based on voice sound signal failure model；
S3. Delaunay Triangulation is carried out to the interior space；
S4. using enhancing particle filter to target following, particle is initialized by filtering algorithm, is cutd open in conjunction with Delaunay triangle
Divide as a result, converting Delaunay triangular pyramids node for the range select permeability in each stage in particle filter target following frame
Select permeability determine particle instantly by finding the adjacent Delaunay triangular apex collection in last moment target position
Search range obtains similarity between particle and target；
S5. the similarity of each particle obtained and target, is weighted and averaged, and obtains the estimated value of coordinates of targets；
S6. target position 3D coordinate is exported, realizes the real-time tracking positioning of moving sound.
2. the particle filter audio source tracking localization method according to claim 1 based on Delaunay Triangulation, special
Sign is：Target sound energy feature is calculated based on voice sound signal failure model described in S2, the specific method is as follows：
Assuming that sound source is issued by the mouth of people, if being s (n), the voice that mobile microphone receives by the voice sound that mouth issues
Signal is x (n), and room noise is that the additive white noise of zero-mean is v (n), then the relationship of this three is：
X (n)=s (n)+v (n) (1)
Wherein, γ is the gain factor of mobile microphone, a (n- τ) be by the acoustic density of the sent out voice of mouth of people, τ be from
Propagation delay time of the mouth to mobile microphone, ps(n- τ) is the vector of a 3*1, indicates mouth position；pxIt is also one
The vector of 3*1 indicates mobile microphone position；dsx=| | ps(n-τ)-px| |, it is set as a constant value；To (1) formula
Both sides calculate average energy value, obtain：
E[x2(n)]=E [s2(n)]+E[v2(n)] (3)
Wherein
In above formula, g=γ2,As it is assumed that dsxA constant value, then from sounding position to
Microphone position apart from very little, then τ can be ignored, therefore, a (n- τ) ≈ a (n), ps(n-τ)≈ps(n),
Since average energy value is carried out in the time window of a length of T, then when sample frequency is fsWhen, M=T*fsIt is for energy
Average number of sampling points；
If E [x2(n)]=yx(t), E [a2(n)]=ys(t), E [v2(n)]=∈ (t), then the target sound energy based on sound failure model
Feature is represented by：
yx(t)=ys(t)+∈(t) (5)
Wherein,If M is sufficiently large, according to central-limit theorem, ∈ (t) approaches normal distribution：σ2It is the variance of noise v (n).
3. the particle filter audio source tracking localization method according to claim 1 based on Delaunay Triangulation, special
Sign is：S3. it is described to the interior space carry out Delaunay Triangulation, using Bowyer-Watson algorithm, algorithm it is main
Steps are as follows：
(1) triangular is constructed, includes all scatterplots, is put into triangle chained list；
(2) scatterplot concentrated is sequentially inserted into, the triangle that its circumscribed circle includes insertion point is found out in triangle chained list, is claimed
For influence triangle, delete the common edge of the influence triangle, by insertion point with influence triangle whole vertex connect
Come, to complete insertion of the point in Delaunay triangle chained list；
(3) it is optimized according to the triangle that Lawson Optimality Criteria newly forms part.
4. the particle filter audio source tracking localization method according to claim 3 based on Delaunay Triangulation, special
Sign is：Step (3) triangle newly formed according to Lawson Optimality Criteria to part optimizes, and specific steps are such as
Under：
(3.1) two triangles with common edge are synthesized into a polygon；
(3.2) it is conducted a survey with largest empty circle criterion, sees its 4th vertex whether in the circumscribed circle of triangle；
(3.3) if amendment diagonal line exchanges diagonal line, that is, the processing of local optimization procedure is completed, by the three of formation
It is angular to be put into Delaunay triangle chained list；
(3.4) circulation executes above-mentioned (3.2) step, until the insertion of all scatterplots finishes.
5. the particle filter audio source tracking localization method according to claim 1 based on Delaunay Triangulation, special
Sign is：It mainly include following four process using enhancing particle filter to target following described in S4：
(1) particle sampler initializes
Particle sampler is M particle appearance being randomly generated in observed object areas adjacent, using Delaunay subdivision to these
Particle is handled, and is allowed to form orderly point set, i.e. Delaunay triangular apex set
V={ vi|vi∈[Lx, Ly, Lz], i=1 ..., M }；
(2) search phase
Firstly, the target position position obtained to last momentt, scan in Delaunay triangular apex set, sentence
Breaking, it is concentrated with the presence or absence of point；IfThen by positiontAccording to Delaunay Triangulation algorithm with
Optimization Steps are inserted into, and are newly added in a point-to-point collection V, and new insertion point positiontOnly with the point near its position
The judgement of Delaunay criterion is carried out, all triangular apex for constituting Delaunay triangle with it is found, does not need to entire V
Collection is traversed；If positiont∈ V, then directly to point positiontNear, it finds and constitutes Delaunay triangle with it
All triangular apex of shape；
Secondly, completing to point positiontRange searching after, available point positiontRange set Vk={ vk|vk∈
V, k=1 ..., N }, wherein vkIt indicates and point positiontDelaunay vertex of a triangle is constituted, N indicates point
positiontConstitute Delaunay vertex of a triangle number；
(3) particle weight is calculated
If search range set VkInterior Arbitrary Particles vkSound energy size bePosition coordinates areDue to mouth position ps(n- τ) distance is dsxMicrophone position pxSound can be big
Small is yx(t), it enablesIndicate VkMiddle Arbitrary Particles vkTo microphone position away from
From, then according to sound pressure level with range attenuation formula, it is availableExpression formula is as follows：
The range attenuation characteristic propagated in space according to sound wave calculates particle weight, i.e. particle sound energy feature and target sound energy
The normalization similitude of feature
At this point, being weighted and averaged according to the similarity of search phase obtained each particle and target, so that it may when obtaining current
Carve t, the estimated value of coordinates of targets
(4) resampling
According to previous step weight computing as a result, using each particle similarityJudge number of effective particles amount PeffctiveWith threshold
It is worth particle PthresholdSize；
Work as Peffctive< PthresholdWhen, illustrate that current particle is degenerated seriously, needs to carry out resampling；The high place of similarity is expanded
Big Delaunay point set search range, that is, the Delaunay triangular apex for increasing the particle are concentrated to search range point, are formed
New search range point set；And the place that similarity is low, it is searched for without Delaunay point set, to form new search range
The point for being unsatisfactory for cum rights particle the power particle such as is mapped as the Particle tracking range at subsequent time t+1 moment by point set；
Work as Peffctive> PthresholdWhen, illustrate that current particle is in good condition, does not need to update search range point set, particle state
It updates and saves.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810560142.1A CN108828524B (en) | 2018-06-03 | 2018-06-03 | Particle filter sound source tracking and positioning method based on Delaunay triangulation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810560142.1A CN108828524B (en) | 2018-06-03 | 2018-06-03 | Particle filter sound source tracking and positioning method based on Delaunay triangulation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108828524A true CN108828524A (en) | 2018-11-16 |
| CN108828524B CN108828524B (en) | 2021-04-06 |
Family
ID=64145717
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810560142.1A Active CN108828524B (en) | 2018-06-03 | 2018-06-03 | Particle filter sound source tracking and positioning method based on Delaunay triangulation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108828524B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111239691A (en) * | 2020-03-08 | 2020-06-05 | 九江职业技术学院 | Multi-sound-source tracking method for restraining main sound source |
| WO2024007352A1 (en) * | 2022-07-07 | 2024-01-11 | 浙江德清知路导航科技有限公司 | Audio-based positioning and object finding method, system and device |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CA2505496A1 (en) * | 2005-04-27 | 2006-10-27 | Universite De Sherbrooke | Robust localization and tracking of simultaneously moving sound sources using beamforming and particle filtering |
| JP5170440B2 (en) * | 2006-05-10 | 2013-03-27 | 本田技研工業株式会社 | Sound source tracking system, method, and robot |
| CN103064059A (en) * | 2012-12-24 | 2013-04-24 | 天津工业大学 | Wireless sensor network sound source locating method |
| CN105044676A (en) * | 2015-09-15 | 2015-11-11 | 北京理工大学 | Energy-based sound source localization method |
| CN105513094A (en) * | 2015-12-17 | 2016-04-20 | 上海交通大学 | Stereo vision tracking method and stereo vision tracking system based on 3D Delaunay triangulation |
-
2018
- 2018-06-03 CN CN201810560142.1A patent/CN108828524B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CA2505496A1 (en) * | 2005-04-27 | 2006-10-27 | Universite De Sherbrooke | Robust localization and tracking of simultaneously moving sound sources using beamforming and particle filtering |
| JP5170440B2 (en) * | 2006-05-10 | 2013-03-27 | 本田技研工業株式会社 | Sound source tracking system, method, and robot |
| CN103064059A (en) * | 2012-12-24 | 2013-04-24 | 天津工业大学 | Wireless sensor network sound source locating method |
| CN105044676A (en) * | 2015-09-15 | 2015-11-11 | 北京理工大学 | Energy-based sound source localization method |
| CN105513094A (en) * | 2015-12-17 | 2016-04-20 | 上海交通大学 | Stereo vision tracking method and stereo vision tracking system based on 3D Delaunay triangulation |
Non-Patent Citations (1)
| Title |
|---|
| 蔡卫平 等: ""一种基于粒子滤波的鲁棒声源跟踪算法"", 《电子测量与仪器学报》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111239691A (en) * | 2020-03-08 | 2020-06-05 | 九江职业技术学院 | Multi-sound-source tracking method for restraining main sound source |
| WO2024007352A1 (en) * | 2022-07-07 | 2024-01-11 | 浙江德清知路导航科技有限公司 | Audio-based positioning and object finding method, system and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108828524B (en) | 2021-04-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Umair et al. | An enhanced K-Nearest Neighbor algorithm for indoor positioning systems in a WLAN | |
| Dorfan et al. | Tree-based recursive expectation-maximization algorithm for localization of acoustic sources | |
| CN111935644B (en) | Positioning method and device based on fusion information and terminal equipment | |
| CN109901112B (en) | Acoustic simultaneous positioning and mapping method based on multi-channel sound acquisition | |
| CN108828501B (en) | Method for real-time tracking and positioning of mobile sound source in indoor sound field environment | |
| CN108828524A (en) | Particle filter audio source tracking localization method based on Delaunay Triangulation | |
| JP2015197417A (en) | Position estimation device, method and program | |
| Alexandridis et al. | Multiple sound source location estimation and counting in a wireless acoustic sensor network | |
| WO2019105194A1 (en) | Propagation path search method and apparatus | |
| Zhang et al. | Localization in 3D sensor networks using stochastic particle swarm optimization | |
| Qingli et al. | Node placement optimization for distributed sensor network using adaptive genetic algorithm | |
| Zhong et al. | Acoustic source tracking in a reverberant environment using a pairwise synchronous microphone network | |
| CN116930963A (en) | Through-wall imaging method based on wireless communication system | |
| Spiesberger | Geometry of locating sounds from differences in travel time: Isodiachrons | |
| CN113923599B (en) | VSLAM closed-loop detection method based on wireless fusion signal | |
| CN113066161B (en) | Modeling method of urban radio wave propagation model | |
| CN108872939A (en) | Interior space geometric profile reconstructing method based on acoustics mirror image model | |
| Tai et al. | Effects of crowd density on radio propagation at 24 GHz in a pedestrian tunnel for 5G communications | |
| Chong et al. | Multiple sound source tracking and identification via degenerate unmixing estimation technique and cardinality balanced multi-target multi-bernoulli filter (DUET-CBMeMBer) with track management | |
| WO2020124091A1 (en) | Automatic fine-grained radio map construction and adaptation | |
| Dai et al. | A propagation modeling approach to source location and navigation | |
| Kawecki et al. | The use of Multiwall model for reference RSSI data preparation for indoor localization purposes | |
| Chen et al. | A new indoor positioning technique based on neural network | |
| Yen et al. | Performance evaluation of sound source localisation and tracking methods using multiple drones | |
| Chang et al. | Integrated cross-correlation and genetic algorithm to improve TDOA-based locating system in preventing radio interference |
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 | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20181116 Assignee: Wuhan xingeno Technology Co.,Ltd. Assignor: GUILIN University OF ELECTRONIC TECHNOLOGY Contract record no.: X2022450000387 Denomination of invention: Particle filter based sound source tracking and location method based on Delaunay triangulation Granted publication date: 20210406 License type: Common License Record date: 20221226 |
|
| EE01 | Entry into force of recordation of patent licensing contract |