CN116017280A - Rapid indoor path tracking method of target portable-free equipment - Google Patents
Rapid indoor path tracking method of target portable-free equipment Download PDFInfo
- Publication number
- CN116017280A CN116017280A CN202211720021.1A CN202211720021A CN116017280A CN 116017280 A CN116017280 A CN 116017280A CN 202211720021 A CN202211720021 A CN 202211720021A CN 116017280 A CN116017280 A CN 116017280A
- Authority
- CN
- China
- Prior art keywords
- sparse
- target
- data
- positioning
- formula
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 34
- 239000013598 vector Substances 0.000 claims abstract description 26
- 230000033001 locomotion Effects 0.000 claims abstract description 13
- 239000011159 matrix material Substances 0.000 claims description 21
- 238000012360 testing method Methods 0.000 claims description 19
- 238000005516 engineering process Methods 0.000 claims description 14
- 230000006870 function Effects 0.000 claims description 11
- 238000012545 processing Methods 0.000 claims description 7
- 239000000654 additive Substances 0.000 claims description 6
- 230000000996 additive effect Effects 0.000 claims description 6
- 238000012549 training Methods 0.000 claims description 6
- 238000002474 experimental method Methods 0.000 claims description 2
- 238000004364 calculation method Methods 0.000 abstract description 3
- 230000008447 perception Effects 0.000 description 9
- 230000008901 benefit Effects 0.000 description 3
- 230000001427 coherent effect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Classifications
-
- 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
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Abstract
The invention discloses a rapid indoor path tracking method of target portable-free equipment, which is characterized by comprising the following steps: 1) Deploying an indoor positioning area to acquire positioning data and dividing a data set; 2) Obtaining a sparse vector according to the sparse representation model; 3) And positioning and obtaining a motion trail according to the sparse coefficient vector. The method is based on a sparse coding model, has low calculation complexity and high positioning efficiency, and can realize rapid and accurate target positioning.
Description
Technical Field
The invention relates to an indoor target positioning and track tracking technology, in particular to a rapid indoor path tracking technology of target carryover-free equipment based on minimum and maximum non-convex punishment MCP (Minimax Concave Penalty, MCP for short) sparse constraint and correlation regularization of a perception matrix column, and particularly relates to a rapid indoor path tracking method of target carryover-free equipment.
Background
The production and life of people are related to the position, and the ancient times are due to the limitation of science and technology and production level, and the description of the position is mostly based on experience, so that the phenomenon of 'south Yuan North rut' can be caused, and the accurate positioning is possible along with the continuous maturity of the artificial satellite technology. The wireless sensing technology provides a new intelligent positioning method, namely a device-free positioning DFL, which realizes the perception of the target position and other information by utilizing the change of wireless propagation signals in the network. The main advantage of this technique is that the target can be located and tracked without carrying the electronic device. This convenient approach has resulted in several mature technologies and applications in various fields, such as optical sensors based on depth imaging and pyroelectric infrared. Almost all DFL techniques currently use a mode based on received signal strength RSS (Receiving Signal Strength, abbreviated as RSS) with ideal positioning accuracy. However, due to multipath interference, the RSS variation of the link tends to become less and more predictable. Therefore, how to obtain accurate location information in a complex changing environment is a serious problem. With the widespread use of OFDM technology, we choose to use the more environmentally adapted channel state information CSI (Channel State Information, CSI for short).
In the research technology of DFL, the existing work generally adopts to improve the performance of algorithm and model to enhance the positioning effect, for example, adopts different modeling modes or adopts different sparse constraints to improve the algorithm performance. Common sparse coding algorithms are classified into three categories according to the difference of the sparse constraint: based on l 0 Algorithm of norm based on convex relaxation l 1 Algorithms for norms and algorithms based on non-convex constraints. Comparison l 0 Norms, l 1 The norm is l 0 Convex relaxation approximation of norms is beneficial to reducing computational complexity, but has some drawbacks, such as causing problems of weak sparsity, punishment and the like, so that the estimated value deviation is large. Recently, the problem of sparse representation based on non-convex constraints is becoming more and more focused by researchers, and the estimated value obtained by adopting the non-convex sparse constraints has the following advantagesBetter sparsity. However, the common sparse regularization may not be robust under high coherence conditions, and the general constraint selection of interrelated variables results in worsening of their generalization errors, and redundant representations tend to result in overfitting, which also results in worsening of interpretability. Empirically, this interpretability is quite important in many practical decision modeling.
Therefore, it is necessary to develop more efficient sparse representation algorithms for application to non-device positioning to obtain better results in processing positioning data with higher similarity.
Disclosure of Invention
The invention aims at overcoming the defects of the prior art and provides a rapid indoor path tracking method of target portable-free equipment. The method is based on a sparse coding model, has low calculation complexity and high positioning efficiency, and can realize rapid and accurate target positioning and obtain a motion trail.
The technical scheme for realizing the aim of the invention is as follows:
a rapid indoor path tracking method of a target portable-free device comprises the following steps:
1) The indoor positioning area is deployed to acquire positioning data and divide a data set: the method comprises the steps that a target is located at an assumed initial position of an indoor wireless sensor network, a receiver continuously receives CSI of different preset positions as available data aiming at target movement, grid position information is correspondingly taken as a label position, CSI of the target at all grid positions is collected, fast Fourier transform processing is carried out on all CSI amplitude data to obtain stable and strong characteristic data, and then the data are divided to obtain an observation test signal x and a training set which are taken as a dictionary D;
2) Obtaining a sparse vector according to a sparse representation model: because the number of targets is always far smaller than the total number of grid points, a sparse vector can be solved by adopting a sparse coding method, a model capable of carrying out sparse reconstruction on positioning data is designed, the measured signal strength is modeled as a combination of a plurality of atoms (columns) in a dictionary, the form is x=dz, MCP sparse regularization is adopted as sparse constraint by the model, the MCP regularization can ensure strong sparsity of a coefficient matrix, and meanwhile, a strategy for selecting uncorrelated variables is realized by considering the correlation of a perception matrix array in the constraint, so that new regularization is introduced; introducing the test signal x in the step 1) into the new sparse representation model in the step 2) to obtain a final sparse coefficient z;
3) Positioning according to the sparse coefficient vector and obtaining a motion trail: because the localization corresponds to each discrete grid solving the sparse representation problem in equation (1), we typically obtain an estimation from multiple trials under noisy conditions. And according to the sequence of the motion track position information, the obtained CSI also has the sequence, the predicted target position is the label position corresponding to the maximum value of the sparse vector, and the continuously output label position is the target track estimation.
The step 2) of bringing the test signal x in the step 1) into the new sparse representation model in the step 2) refers to bringing the received radio data in the test set in the step 1) into the sparse representation model in the step 2) as a signal, wherein a dictionary is the received radio data of a training set, and the sparse representation problem is optimized, namely, a sparse coefficient matrix meeting the MCP sparse constraint and the correlation regularization term of a perception matrix array is obtained, and a constructed sparse representation formula is shown in a formula (1):
in the above formula, x is a test signal, additive noise is expressed as x=dz+v, v is additive white gaussian noise, D is a dictionary, z is a sparse coefficient matrix, and R represents a dictionary column D j and Dk Correlation between; j (J) MCP (z) is MCP sparse constraint, wherein lambda and eta are weight coefficients, and the general value is more than 0 and less than 1, and the expression of the MCP function is shown in a formula (2):
the expression of R is as shown in formula (3):
due to the existence of non-convex MCP sparse constraint terms, a DC equation f (z), namely two convex functions g, is constructed by adopting a differential DC (Difference of Convex, DC for short) planning technology of convex functions 1(z) and g2 (z) a subtracted form, obtainable:
min z f(z)=g 1 (z)-g 2 (z) (4),
g 2 (z)=λ||z|| 1 -J MCP (z) (6),
The second step of DC algorithm is to solve: z
Then solving the above formula by using the adjacent operator technology, as shown in the formula (8) and the formula (9):
[Prox(z)] ij =(ηR+I) -1 sign(z ij )max{|z ij |-t,0},t=λ+ηR|z ij | (9),
re-optimizing { z ij And obtain a sparse matrix z, and calculate a sparse vector z= { z 1,1 ,z 1,2 ,…,z 1,τ ,z 2,1 ,z 2,2 ,…,z B,τ After each sparse vector z is summed, z p =∑ i z p,i Wherein 1.ltoreq.p.ltoreq.s, thereby obtaining z * ={z 1 ,...,z p ,…,z s The predicted single target position is z * A tag position corresponding to the maximum value of (a).
Many positioning methods in the prior art do not take into account the high coherence of the positioning data, which may mask the impact of general sparse regularization on the prediction results. The problem can be solved by combining a new regularization, the regularization adopts variable selection to select the most representative prediction subset from the input signal set, a strategy for selecting irrelevant variables is realized, the problem that CSI is not independent due to the coherence of different receiving antennas is reduced, the positioning data of each scene is more universal, and the algorithm shows very strong generalization capability. As an improvement on common sparse constraint, the technical scheme adopts MCP sparse regularization, MCP belongs to non-convex sparse regularization, can realize strong sparse constraint on coefficient matrixes, can obtain nearly unbiased estimated quantity, divides a part containing MCP into two convex functions, divides the whole problem into a group of sub-function problems related to vectors, further adopts a DC algorithm to solve a minimum optimization formula of MCP sparse representation for accurate positioning, obtains an optimal solution of the problem, and moreover, the minimum value of MCP penalty is easily obtained through a hard threshold value and is different from a soft threshold value, so that the maximum value of each sparse vector can be extracted easily due to the fact that the maximum value is not underestimated by the MCP penalty and is applied to positioning data.
According to the indoor path tracking method of the target carrying-free equipment based on the minimum and maximum non-convex punishment MCP sparse constraint and correlation regularization sparse coding, the novel correlation constraint of the sensing matrix array is adopted to effectively cope with the high correlation characteristic of the actual Channel State Information (CSI) data, the problem that received signals are not independent due to a large number of coherent signals in an indoor environment is solved, the positioning data of each scene are more universal, quick and accurate indoor target carrying-free positioning can be achieved, the proposed algorithm avoids overfitting, information variables are identified, and an interpretable result is provided. Meanwhile, MCP is used as a sparse constraint to obtain strong sparsity and accurate estimation, and a difference DC (Difference of Convex, DC for short) algorithm of a convex function and a near-end gradient method are used for solving an established non-convex target equation.
Compared with the prior art, the technical scheme has the following advantages:
1. considering that there are a large number of coherent signals in the indoor environment, the acquired CSI data has a high and stable correlation: the method has the advantages that the strategy of selecting uncorrelated variables is realized by considering the correlation of the perception matrix array in the constraint, new correlation regularization is constructed, the influence of the correlation on a model can be effectively reduced, stable and accurate positioning performance is obtained, and better universality and mobility are realized for positioning data;
2. the algorithm adopts MCP as a sparse constraint, so that sparsity is enhanced, and an estimated value with smaller deviation is obtained: aiming at a minimized model containing MCP sparse constraint, aiming at a non-convex optimization problem containing sparse constraint in a sparse coding stage, a DC programming technology is adopted to convert a non-convex MCP sparse constraint term into a form of difference between two convex functions in the sparse coding stage, a DC algorithm is used for solving, and finally, a proximity operator method is used for efficiently solving an analytic solution of a sparse coefficient for realizing rapid and accurate positioning.
The method is based on a sparse coding model, has low calculation complexity and high positioning efficiency, and can realize rapid and accurate target positioning and obtain a motion trail.
Drawings
FIG. 1 is a schematic deployment diagram of an apparatus in an embodiment, where the boxes represent grid locations, T represents transmitters, R represents receivers, the solid lines represent affected wireless links, and the dashed lines represent other potential links;
FIG. 2 is a schematic workflow diagram of a sparse coding stage in an embodiment;
FIG. 3 is a flow chart of an embodiment method.
Detailed Description
The present invention will be further described with reference to the accompanying drawings and examples, which are not intended to limit the scope of the invention.
Examples:
as shown in fig. 3, a fast indoor path tracking method of a target portable-free device includes the following steps:
1) The indoor positioning area is deployed to acquire positioning data and divide a data set: the method comprises the steps that a target is located at an assumed initial position of an indoor wireless sensor network, a receiver continuously receives CSI of different preset positions as available data and correspondingly takes grid position information as a label position aiming at target movement, in addition, CSI of the target at all grid positions is collected, fast Fourier transform processing is carried out on all CSI amplitude data to obtain stable and strong characteristic data, then data are divided to obtain an observation test signal x and a training set as a dictionary D, in the example, an indoor positioning scene is divided into N multiplied by N, namely 6 multiplied by 6 square grids with the same size, one position is selected as the initial position of the test target, the target movement is assumed to be continuous, a predefined path is adopted for testing, 4 positions including a deployment area are adopted, grid coordinate sequences are (1, 2), (3, 2), (4, 4), (5 and 4), a data acquisition scheme when the target is located at one fixed position is that a transmitting antenna is transmitted at one position, 6 receiving antennas are continuously received for 30 times, the transmitting antenna is continuously transmitted at the next position, and reception is stopped after the next position is traversed for 10 positions in sequence;
2) Obtaining a sparse vector according to a sparse representation model: because the number of targets is always far smaller than the total number of grid points, a sparse vector can be solved by adopting a sparse coding method, a model capable of carrying out sparse reconstruction on positioning data is designed, the measured signal intensity is modeled as a combination of a plurality of atoms (columns) in a dictionary, the form is x=dz, the model adopts MCP sparse regularization as sparse constraint, the MCP regularization can ensure strong sparsity of a coefficient matrix, meanwhile, the strategy of selecting uncorrelated variables is realized by considering the correlation of a perception matrix in the constraint, and new regularization is introduced; introducing the test signal x of the step 1) into the new sparse representation model of the step 2) to obtain a final sparse coefficient z; in this example, the data processing scheme is to obtain the amplitude data of the first, second and third subcarriers, calculate the average value of the three amplitude data, and obtain a set of CSI amplitudes at a fixed target position. Processing the data into a column vector, splicing all track positions into a test matrix according to columns, processing the data of 36 grid positions into a column vector, and splicing according to columns to form a perception dictionary;
3) Positioning according to the sparse coefficient vector and obtaining a motion trail: because the positioning corresponds to solving the sparse representation problem in the formula (1) for each discrete grid, the example obtains an estimation result from multiple experiments under the noise condition, and according to the sequence of the motion track position information, the obtained CSI also has the sequence, the predicted target position is the label position corresponding to the maximum value of the sparse vector, and the continuously output label position is the target track estimation.
As shown in fig. 2, the received radio data in the test set described in step 1) is brought into the sparse representation model in step 2) as a signal, the dictionary is the received radio data in the training set, the sparse representation problem is optimized, namely, a sparse coefficient matrix meeting the MCP sparse constraint and the correlation regularization term of the perception matrix array is obtained, and a constructed sparse representation formula is shown in formula (1):
in the above formula, x is a test signal, additive noise is expressed as x=dz+v, v is additive white gaussian noise, D is a dictionary, z is a sparse coefficient matrix, and R represents a dictionary column D j and Dk Correlation between; j (J) MCP (z) is MCP sparse constraint, wherein lambda and eta are weight coefficients, and the general value is more than 0 and less than 1, and the expression of the MCP function is shown in a formula (2):
the expression of R is as shown in formula (3):
due to the existence of non-convex MCP sparse constraint terms, a DC programming technology is adopted to construct a DC equation f (z), namely two convex functions g 1(z) and g2 (z) a subtracted form, obtainable:
min z f(z)=g 1 (z)-g 2 (z) (4),
g 2 (z)=λ||z|| 1 -J MCP (z) (6),
the second step of DC algorithm is to solve: z
Then solving the above formula by using the adjacent operator technology, as shown in the formula (8) and the formula (9):
[Prox(z)] ij =(ηR+I) -1 sign(z ij )max{|z ij |-t,0},t=λ+ηR|z ij | (9),
re-optimizing { z ij And obtaining a sparse matrix z, and further, step 3) obtaining a sparse vector z= { z 1,1 ,z 1,2 ,…,z 1,τ ,z 2,1 ,z 2,2 ,…,z s,τ After each sparse vector z is summed, z p =∑ i z p,i, wherein 1≤p S.ltoreq.s, thus giving z * ={z 1 ,...,z p ,...,z s The predicted single target position is z * A tag position corresponding to the maximum value of (a).
In this example, as shown in fig. 1, when an indoor scene is arranged, a receiving device is fixed on a wall, a transmitting end is not fixed, a test target moves at fixed intervals, positioning data sets of 4 positions and labels of corresponding positions are continuously received, the method of this example uses amplitude information of CSI as a basis, processes each position data, and divides the data sets, namely, a part of the data sets is used as a dictionary, and a part of the data sets is used as a test signal; solving a sparse vector based on a sparse coding method, and designing a model capable of performing sparse reconstruction on positioning data; the label position corresponding to the maximum value of each sparse vector is the predicted target position, so that sparsity of a sparse coefficient matrix and data characteristics aiming at CSI are sufficiently focused, universality of an algorithm is improved, MCP is introduced as strong sparse constraint, and correlation of a perception matrix array is introduced as regularization term.
Claims (2)
1. The rapid indoor path tracking method of the target portable-free equipment is characterized by comprising the following steps of:
1) The indoor positioning area is deployed to acquire positioning data and divide a data set: the method comprises the steps that a target is located at an assumed initial position of an indoor wireless sensor network, a receiver continuously receives CSI of different preset positions as available data aiming at target movement, grid position information is correspondingly taken as a label position, CSI of the target at all grid positions is collected, fast Fourier transform processing is carried out on all CSI amplitude data to obtain stable and strong characteristic data, and then the data are divided to obtain an observation test signal x and a training set which are taken as a dictionary D;
2) Obtaining a sparse vector according to a sparse representation model: solving a sparse vector by adopting a sparse coding method, designing a model for sparse reconstruction of positioning data, modeling the measured signal strength as a combination of a plurality of atoms, namely columns, in a dictionary in the form of x=dz, and taking the test signal x in the step 1) into a new sparse representation model in the step 2) to obtain a final sparse coefficient z;
3) Positioning according to the sparse coefficient vector and obtaining a motion trail: because the positioning corresponds to the sparse representation problem in each discrete grid solution formula (1), an estimation result is usually obtained from multiple experiments under the noise condition, the obtained CSI also has a sequence according to the sequence of the motion track position information, the predicted target position is the label position corresponding to the maximum value of the sparse vector, and the continuous output label position is the target track estimation.
2. The method for fast indoor path tracking of a target carryover-free device according to claim 1, wherein the step 2) of bringing the test signal x in step 1) into the new sparse representation model in step 2) refers to bringing the received radio data in the test set in step 1) as a signal into the sparse representation model in step 2), the dictionary is the received radio data of the training set, and the sparse representation problem is optimized, that is, a sparse coefficient matrix satisfying the MCP sparse constraint and the correlation regularization term of the perceptual moment array is obtained, and the constructed sparse representation formula is shown in formula (1):
in the above formula, x is a test signal, additive noise is expressed as x=dz+v, v is additive white gaussian noise, D is a dictionary, z is a sparse coefficient matrix, and R represents a dictionary column D j and Dk Correlation between; j (J) MCP (z) is MCP sparse constraint, wherein lambda and eta are weight coefficients, and the general value is more than 0 and less than 1, and the expression of the MCP function is shown in a formula (2):
the expression of R is as shown in formula (3):
constructing DC equation f (z) using difference DC programming technique of convex functions, i.e. two convex functions g 1(z) and g2 (z) a subtracted form, obtainable:
min z f(z)=g 1 (z)-g 2 (z) (4),
g 2 (z)=λ||z|| 1 -J MCP (z) (6),
The second step of DC algorithm is to solve: z
Then solving the above formula by using the adjacent operator technology, as shown in the formula (8) and the formula (9):
[Prox(z)] ij =(ηR+I) -1 sign(z ij )max{|z ij |-t,0},t=λ+ηR|z ij | (9),
re-optimizing { z ij And obtain a sparse matrix z, and calculate a sparse vector z= { z 1,1 ,z 1,2 ,...,z 1,τ ,z 2,1 ,z 2,2 ,...,z s,τ After each sparse vector z is summed, z p =∑ i z p,i Wherein 1.ltoreq.p.ltoreq.s, thereby obtaining z * ={z 1 ,...,z p ,...,z s The predicted single target position is z * A tag position corresponding to the maximum value of (a).
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211720021.1A CN116017280A (en) | 2022-12-30 | 2022-12-30 | Rapid indoor path tracking method of target portable-free equipment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211720021.1A CN116017280A (en) | 2022-12-30 | 2022-12-30 | Rapid indoor path tracking method of target portable-free equipment |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116017280A true CN116017280A (en) | 2023-04-25 |
Family
ID=86033238
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211720021.1A Pending CN116017280A (en) | 2022-12-30 | 2022-12-30 | Rapid indoor path tracking method of target portable-free equipment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116017280A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117454154A (en) * | 2023-12-22 | 2024-01-26 | 江西农业大学 | Robust feature selection method for bias marker data |
-
2022
- 2022-12-30 CN CN202211720021.1A patent/CN116017280A/en active Pending
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117454154A (en) * | 2023-12-22 | 2024-01-26 | 江西农业大学 | Robust feature selection method for bias marker data |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Hsieh et al. | Deep learning-based indoor localization using received signal strength and channel state information | |
CN109993280B (en) | Underwater sound source positioning method based on deep learning | |
Brodeski et al. | Deep radar detector | |
Njima et al. | Indoor localization using data augmentation via selective generative adversarial networks | |
CN111160176B (en) | Fusion feature-based ground radar target classification method for one-dimensional convolutional neural network | |
CN105069752B (en) | The optics ocean clutter cancellation method of chaos during based on sky | |
Zhang et al. | An efficient machine learning approach for indoor localization | |
CN104299248A (en) | Method for utilizing foresight sonar image for predicting motion of multiple underwater dynamic targets | |
CN116017280A (en) | Rapid indoor path tracking method of target portable-free equipment | |
CN111757250B (en) | Channel state information positioning fingerprint construction method based on tensor decomposition | |
Klus et al. | Transfer learning for convolutional indoor positioning systems | |
Wu et al. | Designing multi-task convolutional variational autoencoder for radio tomographic imaging | |
Pegoraro et al. | Deep learning for accurate indoor human tracking with a mm-wave radar | |
Kim et al. | A study on 3D human pose estimation using through-wall IR-UWB radar and transformer | |
CN115469303A (en) | Cognitive biological radar method and device for detecting human body posture and vital signs | |
CN117119377A (en) | Indoor fingerprint positioning method based on filtering transducer | |
CN108834043B (en) | Priori knowledge-based compressed sensing multi-target passive positioning method | |
CN113740802B (en) | Signal source positioning method and system for performing matrix completion by using adaptive noise estimation | |
CN113514796B (en) | Passive positioning method, system and medium | |
CN111239682B (en) | Electromagnetic emission source positioning system and method | |
CN114724245A (en) | CSI-based incremental learning human body action identification method | |
CN113421281A (en) | Pedestrian micromotion part separation method based on segmentation theory | |
Rinchi et al. | Deep-learning-based accurate beamforming prediction using LiDAR-assisted network | |
Thanakulketsarat et al. | Ionograms scaling by using the convolutional neural network | |
CN107766607A (en) | A kind of transmitting for extension target detection receives robust design method |
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 |