CN109922427A - Utilize the intelligent radio positioning system and method for large scale array antenna - Google Patents
Utilize the intelligent radio positioning system and method for large scale array antenna Download PDFInfo
- Publication number
- CN109922427A CN109922427A CN201910166653.XA CN201910166653A CN109922427A CN 109922427 A CN109922427 A CN 109922427A CN 201910166653 A CN201910166653 A CN 201910166653A CN 109922427 A CN109922427 A CN 109922427A
- Authority
- CN
- China
- Prior art keywords
- mobile terminal
- convolutional neural
- fingerprint information
- location fingerprint
- user side
- 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
Abstract
The invention proposes a kind of intelligent radio positioning systems and method using large scale array antenna.Base station side extracts angle time delay domain channel strength matrix as location fingerprint information from uplink channel state information.In off-line phase, base station side utilizes location fingerprint information training concatenated convolutional neural network, and the location fingerprint information of characteristic point stores in database jointly in the network structure relevant to the neuron of each convolutional neural networks for obtaining training, weighted value and activation primitive and cell coverage area；In on-line stage, the corresponding location fingerprint information input concatenated convolutional neural network of user side mobile terminal apparatus is calculated the position coordinates of user side and feeds back to corresponding user side mobile terminal by base station side.The present invention can rely on wireless telecom equipment, significantly improve positioning accuracy of the mobile terminal under the complicated scatterer environment such as city building and interior and reduce Positioning System time expense and storage overhead.
Description
Technical field
The present invention relates to a kind of intelligent radio positioning systems and method using large scale array antenna, more particularly to one kind
The wireless location system and method for the extensive mimo system angle time delay domain channel characteristics in broadband are utilized based on convolutional neural networks.
Background technique
With intelligent mobile terminal and the fast development of the application and service based on geographical location information, people are to movement
The accuracy requirement of terminal wireless positioning increasingly improves.Most of location requirements result from the complicated scatterer such as city building and interior
Under environment, wireless signal generally goes through a plurality of obstructed path and propagates, and the precision of wireless location is made to be a greater impact.Based on position
The wireless location method for setting finger print information can sufficiently excavate influence of the scatterer environment to channel characteristics, be widely studied at present
For the wireless location under complicated scatterer environment.
The extensive mimo system in broadband has significant excellent in terms of improving wireless communication system spectrum efficiency and power efficiency
Gesture is one of the key technology of the following 5G wireless communication.Base station side obtains high space by configuring large-scale antenna array
Angular resolution；By using orthogonal frequency division multiplexing (OFDM) technology, broad-band channel is decomposed into multiple parallel narrow band channels,
Obtain high time delay resolving power.The extensive mimo system in broadband combines the advantage of two kinds of technologies, special with angle time delay domain channel
Sign, which carries out wireless location as location fingerprint information, can make full use of the extensive mimo system in broadband to the angle of wireless channel
The extracted with high accuracy of time delay domain channel characteristics.
It is existing when extracting angle time delay domain channel characteristics as location fingerprint information in the extensive mimo system in broadband
Wireless location method based on location fingerprint information can generate biggish time overhead and storage overhead mostly.
Summary of the invention
Goal of the invention: the object of the present invention is to provide a kind of intelligent radio positioning system using large scale array antenna and
Method, the search range of the possible distributed areas of user side mobile terminal is reduced using convolutional neural networks step by step, and estimates user
The position coordinates of side mobile terminal.The present invention makes full use of the extensive mimo system in broadband to wireless channel angle time delay domain channel
The extracted with high accuracy of feature can significantly improve positioning of the mobile terminal under the complicated scatterer environment such as city building and interior
Precision simultaneously reduces Positioning System time expense and storage overhead.
Technical solution: for achieving the above object, the present invention adopts the following technical scheme:
A kind of intelligent radio positioning system using large scale array antenna, comprising:
Location fingerprint information extraction modules are strong for extracting angle time delay domain channel from uplink channel estimation result
Matrix is spent as location fingerprint information；
Concatenated convolutional neural network module, including return output cascade convolutional neural networks module and/or classification output stage
Join convolutional neural networks module, for according to location fingerprint information estimate user side mobile terminal apparatus position coordinates and/or
User side mobile terminal is located at the probability of different blocks in cell coverage area；
And database module, the concatenated convolutional neural network obtained for storing off-line phase training, and use and divide
Characteristic point in the cell coverage area different blocks that off-line phase sampling obtains when class output cascade convolutional neural networks module
Location fingerprint information；
The recurrence output cascade convolutional neural networks module, the position for exporting user side mobile terminal apparatus are sat
Mark, cascaded topology are connected group by one or more levels convolutional neural networks classifier and output end convolutional neural networks recurrence device
At；
The classification output cascade convolutional neural networks module is located at MPS process model for exporting user side mobile terminal
The probability for enclosing interior different blocks, cooperate estimation user side mobile terminal with location fingerprint information matches module, position estimation module
The position coordinates of device, cascaded topology is by one or more levels convolutional neural networks classifier and output end convolutional neural networks point
Class device is composed in series；
The location fingerprint information matches module is believed for choosing from database with user side mobile terminal locations fingerprint
Cease the maximum several characteristic points of joint angle time delay likeness coefficient, composition characteristic point set；
The position estimation module, for by several blocks of location fingerprint information matches block search maximum probability
The most similar several characteristic points of location fingerprint information corresponding with user side mobile terminal apparatus calculate user side mobile terminal dress
The position coordinates set.
As preferred embodiment, off-line phase acquisition data include for cell coverage area divide step by step it is every
Level-one segmentation, the offline position for sampling characteristic point in each sub-block that mobile terminal apparatus measurement obtains of base station side device cooperation refer to
Line information；The series that wherein cell coverage area block divides is according to convolutional neural networks in concatenated convolutional neural network module
Series determines；The corresponding characteristic point position finger print information of each grade of segmentation and position coordinates are used to train the convolution mind of corresponding level
Through network.
As preferred embodiment, angle time delay domain channel strength square is extracted in the location fingerprint information extraction modules
The process of battle array are as follows: angle time delay is converted by the uplink channel state information that up channel training obtains by mobile terminal
Domain channel response matrix；By each element modulus in angle time delay domain channel response matrix, and to continuously sample is averaged several times
Or weighted average, the corresponding angle time delay domain channel strength matrix of mobile terminal is obtained, and as location fingerprint information.
As preferred embodiment, joint angle time delay similitude system is calculated in the location fingerprint information matches module
Several method are as follows: definition dislocation factor of n, misplace step-length L, and n is the integer in section (- L+1, L-1), successively by first position
Each column vector and n-component column vector after respective column in second location fingerprint information take inner product in finger print information, and by gained
Inner product summation, using it is different dislocation the factors under gained with maximum value it is similar as the joint angle time delay of two location fingerprint information
Property coefficient.
As preferred embodiment, the position coordinates of user side mobile terminal apparatus are calculated in the position estimation module
Method be, by the corresponding location fingerprint information of user side mobile terminal apparatus joint angle time delay similitude system in the database
The position coordinates weighted sum of the maximum K characteristic point of number；The wherein corresponding weight coefficient calculation method of k-th of characteristic point are as follows:
With the joint angle time delay likeness coefficient of k-th characteristic point and user side mobile terminal apparatus divided by whole K characteristic points and
The sum of joint angle time delay likeness coefficient of user side mobile terminal apparatus.
As preferred embodiment, a kind of intelligent radio positioning system using large scale array antenna, packet
Include base station side device, offline sampling mobile terminal side device and customer mobile terminal side device；
The base station side device includes: uplink channel estimation module, for sending according to each mobile terminal received
Pilot signal implement channel state information acquisition；The location fingerprint information extraction modules；The concatenated convolutional neural network
Module；The database module；The location fingerprint information matches module；The position estimation module；And information exchange mould
Block, the location information sent in off-line phase, receiving offline sampling mobile terminal, and in on-line stage to user's sidesway
Dynamic terminal sends its location estimation result；
The offline sampling mobile terminal side device includes: offline map and navigation module, is used in off-line phase, in real time
The accurate location information for obtaining offline sampling mobile terminal；Driving device, for driving offline sampling mobile eventually in off-line phase
End is moved in cell coverage area, and is stopped in each characteristic point；Uplink pilot sending module, for sending uplink
Road pilot signal；Information exchange module, for sending the position of offline sampling mobile terminal to base station side device in off-line phase
Information；
Customer mobile terminal side device includes: uplink pilot sending module, for sending uplink pilot
Signal；And user side information exchange module, for sending offline sampling mobile terminal hair to base station side device in off-line phase
The location information sent.
A kind of intelligent radio localization method using large scale array antenna described in another method of the present invention, comprising:
Cell coverage area is divided into several sub-blocks by off-line phase step by step, is divided for each grade, and measurement cell is covered
The location fingerprint information of characteristic point in each sub-block of lid range utilizes the convolutional Neural net of location fingerprint information training corresponding level
Network, and will the obtained network structure relevant to the neuron of convolutional neural networks at different levels of training, weighted value and activation primitive or
It is stored in database jointly with the location fingerprint information of characteristic point in cell coverage area；
On-line stage, by the corresponding location fingerprint information input concatenated convolutional neural network of user side mobile terminal apparatus,
Estimate the position coordinates of user side mobile terminal apparatus；
The concatenated convolutional neural network, including return output cascade convolutional neural networks and/or classification output cascade volume
Product neural network；
The recurrence output cascade convolutional neural networks, for exporting the position coordinates of user side mobile terminal apparatus,
Cascaded topology returns device with output end convolutional neural networks by one or more levels convolutional neural networks classifier and is composed in series；
The classification output cascade convolutional neural networks are located in cell coverage area for exporting user side mobile terminal
The probability of different blocks, cascaded topology is by one or more levels convolutional neural networks classifier and output end convolutional neural networks point
Class device is composed in series；
When using classification output cascade convolutional neural networks, by the corresponding location fingerprint information of user side mobile terminal apparatus
Input classification output cascade convolutional neural networks, obtain user side mobile terminal and are located at the general of different blocks in cell coverage area
Rate searches for position corresponding with user side mobile terminal apparatus by location fingerprint information matches in several blocks of maximum probability
The maximum several characteristic points of finger print information joint angle time delay likeness coefficient are set, estimate the position of user side mobile terminal apparatus
Coordinate.
As preferred embodiment, the location fingerprint information is extracted from uplink channel estimation result
Angle time delay domain channel strength matrix, extraction process are the uplink for obtaining mobile terminal by up channel training
Channel state information is converted into angle time delay domain channel response matrix；Each element in angle time delay domain channel response matrix is taken
Mould, and to continuously sample is averaged or is weighted and averaged several times, obtain the corresponding angle time delay domain channel strength of the mobile terminal
Matrix, and as location fingerprint information.
As preferred embodiment, the joint angle time delay likeness coefficient, calculation method is, definition dislocation because
Sub- n, misplace step-length L, n be section (- L+1, L-1) in integer, successively by each column vector in first location fingerprint information with
N-component column vector takes inner product after respective column in second location fingerprint information, and gained inner product is summed, by different dislocation because
Joint angle time delay likeness coefficient of the maximum value of the lower gained sum of son as two location fingerprint information.
As preferred embodiment, estimate that the method for the position coordinates of user side mobile terminal apparatus is, it will be in data
The position coordinates weighted sum of the maximum K characteristic point of joint angle time delay likeness coefficient in library, wherein k-th of characteristic point pair
The weight coefficient calculation method answered is, with the joint angle time delay similitude of k-th characteristic point and user side mobile terminal apparatus
Coefficient divided by whole K characteristic points and user side mobile terminal apparatus the sum of joint angle time delay likeness coefficient.
The utility model has the advantages that the intelligent radio positioning system and method proposed by the present invention using large scale array antenna, base
This feature is that base station side extracts angle time delay domain channel strength matrix as location fingerprint from uplink channel state information
Information.Wireless location method includes off-line phase and on-line stage two parts: in off-line phase, base station side is believed using location fingerprint
Training concatenated convolutional neural network, and the network structure relevant to the neuron of each convolutional neural networks that training is obtained are ceased,
The location fingerprint information of characteristic point stores in database jointly in weighted value and activation primitive and cell coverage area；Online
The corresponding location fingerprint information input concatenated convolutional neural network of user side mobile terminal apparatus is calculated and is used by stage, base station side
The position coordinates of family side simultaneously feed back to corresponding user side mobile terminal.The present invention can rely on wireless telecom equipment, significantly mention
Positioning accuracy of the high mobile terminal under the complicated scatterer environment such as city building and interior simultaneously reduces Positioning System time expense
And storage overhead.Compared with prior art, the present invention has the advantage that
1. broadband can be made full use of extensive as location fingerprint information using angle time delay domain channel strength matrix
Extracted with high accuracy of the mimo system to wireless channel angle time delay domain channel characteristics.
2. measuring location fingerprint information similitude using multiple convolutional neural networks with cascaded topology, tradition connection is compared
Close angle degree time delay likeness coefficient reduces modeling error, sufficiently excavates the difference of the corresponding location fingerprint information of different location coordinate
It is anisotropic.
3. multiple convolutional Neurals with cascaded topology in off-line phase training that can only by being stored in database
Network implementations positioning, does not depend on location fingerprint information database, reduces wireless location system storage overhead.
4. utilizing multiple convolutional neural networks with cascaded topology to reduce user side mobile terminal step by step in on-line stage
The search range of possible distributed areas reduces wireless location system time overhead.
5. the acquisition of location fingerprint information can rely on the existing channel state information of mobile terminal, do not need additionally to occupy logical
Believe resource, is more applicable for real system.
Detailed description of the invention
It to describe the technical solutions in the embodiments of the present invention more clearly, below will be to embodiment or description of the prior art
Needed in attached drawing be briefly described, it should be apparent that, the accompanying drawings in the following description only shows of the invention one
A little embodiments for those of ordinary skill in the art without creative efforts, can also be according to these
The attached drawing of attached drawing acquisition other embodiments.
Fig. 1 is the intelligent radio localization method flow chart using large scale array antenna.
Fig. 2 is that schematic diagram is divided in off-line phase area to be targeted step by step.
Fig. 3 is off-line phase characteristic point position finger print information acquisition methods flow chart.
Fig. 4 is off-line phase concatenated convolutional neural network training method flow chart.
Fig. 5 is on-line stage user side mobile terminal locations estimation method flow chart.
Fig. 6 is to return output cascade convolutional neural networks structural block diagram.
Fig. 7 is classification output cascade convolutional neural networks structural block diagram.
Fig. 8 is the intelligent radio positioning system structure block diagram using large scale array antenna.
Fig. 9 is the extensive mimo channel model schematic in broadband.
Figure 10 is convolutional neural networks structural schematic diagram.
Specific embodiment
In order to enable those skilled in the art to better understand the solution of the present invention, below in conjunction in the embodiment of the present invention
Attached drawing, the technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is only this
Invention a part of the embodiment, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art exist
Every other embodiment obtained under the premise of creative work is not made, should fall within the scope of the present invention.
As shown in Figure 1, a kind of intelligent radio localization method using large scale array antenna disclosed by the embodiments of the present invention,
Specifically include that (1) off-line phase area to be targeted is divided step by step；(2) off-line phase characteristic point position finger print information obtains；(3)
Off-line phase concatenated convolutional neural metwork training；(4) on-line stage user side mobile terminal locations are estimated.
Communication system configures large-scale antenna array in base station side, using the orthogonal frequency division multiplexing (OFDM) with cyclic prefix
Modulation system, in up channel training, multiple sub- loads of each mobile terminal in one or more continuous OFDM symbols in cell
Uplink pilot signal is sent on wave simultaneously, base station carries out the uplink of each mobile terminal in cell according to the pilot signal received
Path channels estimation.Base station is using the space angle resolving power of extensive mimo system and the time delay resolving power of broadband system, from upper
Angle time delay domain channel strength matrix is extracted in downlink channel state information as location fingerprint information.
In off-line phase, area to be targeted is divided step by step, as shown in Fig. 2, each sub-block includes piecemeal center
And block margin.Characteristic point delimited at certain intervals in the sub-block that segmentation obtains, and the feature dot density at piecemeal center is greater than
Block margin, the offline mobile terminal that samples moves in sub-block, and stops in each characteristic point, measures the position of this feature point
Finger print information.
As shown in figure 3, a kind of intelligent radio localization method using large scale array antenna disclosed by the embodiments of the present invention
In off-line phase characteristic point position finger print information acquisition methods, specifically include that (1) is directed to when prime block to be positioned, generate
Characteristic point position coordinate；(2) sampling mobile terminal is moved to next characteristic point offline, sends uplink pilot；(3) base
Side stand according to the pilot signal acquisition uplink channel state information received；(4) base station side is by uplink channel state
Information is converted into angle time delay domain channel response matrix；(5) base station side utilizes the angle time delay domain channel obtained in a period of time
Response matrix calculates angle time delay domain channel strength matrix；(6) whether the offline sampling mobile terminal of judgement traverses all characteristic points,
If so, (7) are entered step, if it is not, return step (2)；(7) base station side using characteristic point angle time delay domain channel strength matrix with
Prime convolutional neural networks are worked as in position coordinates training.
As shown in figure 4, a kind of intelligent radio localization method using large scale array antenna disclosed by the embodiments of the present invention
In off-line phase concatenated convolutional neural network training method, specifically include that (1) area to be targeted is divided step by step；(2) base station side
Obtain characteristic point position finger print information；(3) concatenated convolutional neural network selects, if selection returns output, enters step (4), if
Selection sort output, enters step (6)；(4) base station side training returns output cascade convolutional neural networks；(5) base station side will instruct
The concatenated convolutional neural network got is stored into database；(6) base station side training classification output cascade convolutional neural networks；
(7) the concatenated convolutional neural network and characteristic point position finger print information that base station side obtains training are stored into database.
In on-line stage, base station side in the location fingerprint information input data library by user side mobile terminal by storing
Concatenated convolutional neural network calculates the position coordinates of user side mobile terminal and feeds back to corresponding mobile terminal.
As shown in figure 5, a kind of intelligent radio localization method using large scale array antenna disclosed by the embodiments of the present invention
In on-line stage user side mobile terminal locations estimation method, specifically include that (1) base station side obtain user side mobile terminal position
Set finger print information；(2) concatenated convolutional neural network selects, if selection returns output, enters step (3), if selection sort exports,
Enter step (5)；(3) location fingerprint information input is returned output cascade convolutional neural networks by base station side；(4) user side is exported
Mobile terminal locations coordinate；(5) base station side by location fingerprint information input classify output cascade convolutional neural networks；(6) it exports
User side mobile terminal is located at the probability of different blocks；(7) base station side is searched in probability highest block determines that user side is mobile
Terminal character pair point set；(8) base station side calculates user side mobile terminal locations coordinate using K nearest neighbor algorithm；(9) base station
Lateral user side sends position coordinates estimated result.
Concatenated convolutional neural network, including return output cascade convolutional neural networks and classification output cascade convolutional Neural net
The positioning of any of them concatenated convolutional neural fusion can be used in network, wireless location method disclosed by the embodiments of the present invention.
As shown in fig. 6, a kind of recurrence output cascade convolutional neural networks disclosed by the embodiments of the present invention, for exporting user
The position coordinates of side mobile terminal apparatus specifically include that (1) input position finger print information；(2) one or more levels convolutional Neural net
Network classifier；(3) output end convolutional neural networks return device；(4) output mobile terminal location coordinate.
As shown in fig. 7, a kind of classification output cascade convolutional neural networks disclosed by the embodiments of the present invention, for exporting user
Side mobile terminal is located at the probability of different blocks in cell coverage area, specifically includes that (1) input position finger print information；(2) one
Grade or multistage convolutional neural networks classifier；(3) output end convolutional neural networks classifier；(4) output mobile terminal is located at not
With the probability of block.
Wherein convolutional neural networks return device, for the location fingerprint information based on user side mobile terminal, calculate user
The position coordinates of side mobile terminal.Convolutional neural networks classifier, for the location fingerprint information based on user side mobile terminal,
User side mobile terminal is calculated in the probability that may be distributed each sub-block in block.Every level-one convolutional neural networks classifier output
Under current block divides, terminal to be positioned is located at the probability of different blocks, and next stage convolutional neural networks classifier will be upper one
The probability that terminal to be positioned is located at different sub-blocks is further provided within the scope of the block that grade classifier reduces.
As shown in figure 8, a kind of intelligent radio positioning system using large scale array antenna disclosed by the embodiments of the present invention,
It include: base station side device, it is offline to sample mobile terminal side device and customer mobile terminal side device.
Base station side device, comprising: (1) uplink channel estimation module, for being sent according to each mobile terminal received
Pilot signal implement channel state information acquisition；(2) location fingerprint information extraction modules are used for from uplink channel state
Location fingerprint information is extracted in information；(3) concatenated convolutional neural network module, for exporting the position of user side mobile terminal apparatus
It sets coordinate or output mobile terminal is located at the probability of different blocks；(4) database module is obtained for storing off-line phase training
Concatenated convolutional neural network and by sample offline mobile terminal apparatus acquisition location fingerprint information；(5) location fingerprint is believed
Matching module is ceased, for choosing from database and the maximum feature of user side mobile terminal locations finger print information in on-line stage
Point, composition characteristic point set；(6) position estimation module, for calculating the position of user side mobile terminal in on-line stage；(7)
Information exchange module is used to receive the location information that offline sampling mobile terminal is sent in off-line phase, and in on-line stage
Its location estimation result is sent to user side mobile terminal.
Offline sampling mobile terminal side device, comprising: (1) offline map and navigation module are used in off-line phase, in real time
The accurate location information for obtaining offline sampling mobile terminal；(2) driving device, for driving offline sampling to move in off-line phase
Dynamic terminal moves in cell coverage area, and stops in each characteristic point；(3) uplink pilot sending module, for sending out
Send uplink pilot；(4) information exchange module, for sending offline sampling to base station side device and moving in off-line phase
The location information of dynamic terminal.
Customer mobile terminal side device, comprising: (1) uplink pilot sending module, for sending uplink pilot
Signal；(2) information exchange module, for sending the position that offline sampling mobile terminal is sent to base station side device in off-line phase
Confidence breath.
The method of the present invention be primarily adapted for use in base station side be equipped with large-scale antenna array with and meanwhile service the big rule of multiple users
Mould MIMO-OFDM system.Below with reference to specific communication system example to the extensive mimo system in broadband of the present invention without
The specific implementation process of line localization method elaborates, it should be noted that the method for the present invention is applicable not only to following example institute
The specific system model lifted, is applied equally to the system model of other configurations.
One, system configuration
In this embodiment, consider extensive MIMO-OFDM system, base station side configuration includes the above antenna element of dozens of
Large-scale antenna array, large-scale antenna array can be used a variety of array structures such as linear array, circular array or plate array it
One.Assuming that the antenna element number that base station side is equipped with is Nt, omnidirectional antenna or fan antenna can be used in each antenna element, when each
When antenna element is using omnidirectional antenna, 120 degree of fan antennas and 60 degree of fan antennas, the spacing between each antenna element is configurable
For 1/2 wavelength,Wavelength and 1 wavelength.Single polarization or multi-polarization antenna can be used in each antenna element.Assuming that existing in cell
The user of K outfit single antenna.Using orthogonal frequency division multiplexing (OFDM) modulation-transmission technology with cyclic prefix, subcarrier number
For Nc, circulating prefix-length Ng。
Two, location fingerprint acquisition of information with compare
1, channel model
Fig. 9 is the extensive mimo channel model schematic of single cell, does not lose its generality, it is assumed that base station side is equipped with NtA day
The even linear array of line unit composition, the spacing between each antenna element are d.Assuming that k-th of user side mobile terminal transmission is upper
Row pilot signal is through 1 scatterer propagated of P > > to base station side aerial array.And dp,kRespectively indicate kth
A user is long via the angle of arrival and physical pathway of the signal of the 1st receiving antenna of pth article scatterer propagated to base station side
Degree.Then k-th of user is represented by via the channel impulse response in pth scatterer path
WhereinIndicate the complex channel gain in pth scatterer path.According to Fig. 9, angle of arrivalIt is right
The array response vector answered is represented by
Define phase shift Fourier transform matrixWherein each element definition is
Then k-th of user is represented by via the angle domain channel impulse response in pth scatterer pathAngle domain
Channel impulse responseIt can regard traditional antenna domain channel impulse response q asp,kMapping in angle domain, each of which element
Characterize the channel gain in corresponding angle direction.
Due to using orthogonal frequency division multiplexing (OFDM) modulation-transmission technology, the frequency selective fading that Multipath Transmission generates
Channel can be changed to frequency domain flat fading channel, then the channel frequency domain response of k-th of user, first of subcarrier is represented by
Wherein np,kIndicate k-th of user via the distinguishable propagation delay time in pth scatterer path.It then include all sons
The total channel frequency domain response of carrier wave is represented by
Hk=[hk,0,hk,1,...,hk,L-1] (4)
2, location fingerprint acquisition of information
Based on above-mentioned channel model, the frequency domain response of the extensive mimo channel in broadband can be by angle time delay domain channel response table
Show.Conversely, defining angle time delay domain channel response matrix is Gk, then GkCalculation formula it is as follows
WhereinIndicate NcTie up the preceding N of tenth of the twelve Earthly Branches Discrete Fourier transformgColumn, i.e.,We define angle time delay domain channel energy matrix
Wherein ⊙ indicates Hadamard product,And then define angle time delay domain channel strength
MatrixEach of which element is the square root of angle time delay domain channel energy matrix corresponding position element, by following formula
It provides
In this embodiment, the uplink channel state information that base station obtains mobile terminal by up channel training
It is converted into angle time delay domain channel response matrix, then by each element in angle time delay domain channel response matrix and its own coupling
Close, and take statistical average whithin a period of time, obtain the corresponding angle time delay domain channel strength matrix of the mobile terminal, and by its
As location fingerprint information.
3, joint angle time delay likeness coefficient
It is fixed using the angle time delay domain distribution character of location fingerprint information in order to describe the similitude of location fingerprint information
Adopted joint angle time delay likeness coefficient is
Wherein [Ξ]tThe t of representing matrix Ξ is arranged, and n indicates the dislocation factor, and L indicates dislocation step-length, and n is section (- L+1, L-
1) integer in.Above-mentioned calculation formula indicates successively to believe each column vector in i-th of location fingerprint information and j-th of location fingerprint
N-component column vector takes inner product after respective column in breath, and gained inner product is summed, by the maximum of gained sum under the different dislocation factors
It is worth the joint angle time delay likeness coefficient as two location fingerprint information.
Three, convolutional neural networks structure
The convolutional neural networks used in this embodiment, basic structure are as shown in Figure 10.Each exemplary convolution nerve
Network can be divided into two parts: feature learning module and classification/regression block.Wherein feature learning module utilizes a series of small rulers
Degree convolution kernel extracts feature from angle time delay domain channel strength matrix position finger print information, and by full articulamentum that these are special
Sign passes to classification/regression block.Categorization module converts these features to the probability that the mobile terminal is located at different blocks.It returns
Module is returned to convert two-dimensional position coordinate for these features.
Feature learning module is made of L grades of concatenated CALP modules, and each CALP module includes:
(1) convolution (C): by KlThe convolutional layer of a convolution kernel composition；
(2) it activates (A): nonlinear response is introduced by activation primitive；
(3) local acknowledgement's regularization (L): introducing closes on feature competition mechanism；
(4) pond (P): feature is carried out down-sampled.
The implementation of each layer in CALP module is explained in detail below.It enablesIndicate the input of first of CALP module
Feature, 1 < l < L.For the 1st CALP module, input feature vector is angle time delay domain channel strength matrix, i.e.,
It enablesIndicate m-th of convolution kernel of first of CALP module, With
Convolution kernel is respectively indicated in angle domain, the size of three dimensions of time delay domain and convolution kernel depth.First CALP module convolutional layer
Output feature can be expressed asWherein each element is calculated by following formula
WhereinComplete 1 column vector is indicated for bias term, 1.For TlThe sub- sensing region that zero padding obtains, ⊙
For Hadamard product.During feature fl transmission, each convolution kernelAlong angle domain and time delay domain both direction and Tl
Interior every sub- sensing region convolution, that is, calculate the Hadamard product of convolution kernel and sub- sensing region and all results summed.
The output feature C of convolutional layerl+1It is delivered to active coating, introduces nonlinear response.The present embodiment uses line rectification
Function (ReLU) is used as activation primitive, is given by
[Al+1]i,j,m=max (0, [Cl+1]i,j,m) (11)
WhereinIt indicates the output of active coating, and is delivered to local acknowledgement's regularization (LRN) layer.Part
The output of response regularization layer is represented by
Wherein summation operation is carried out along convolution kernel depth direction.Wherein k, α, β and mdepthFor constant coefficient, value herein
For k=2, α=10-4, β=0.75, mdepth=5, value can also be adjusted flexibly according to system-operating environment.Local acknowledgement is just
Then change and introduces competition mechanism between adjacent feature and help to detect most significant feature.
It is finally down-sampled using output feature progress of the pond layer to local acknowledgement's regularization layer, it is given by
Wherein x ∈ [1, Px] ∩ Z, y ∈ [1, Py]∩Z.Wherein Px, PyThe dimension of feature is exported for CALP module.
Convolutional neural networks are exported for returning, a recurrence is connected by full articulamentum after afterbody CALP module
Device is given by for the two-dimensional position coordinate of output mobile terminal
WhereinFor two-dimensional position coordinate,For deviation.
Loss function is given by
Wherein NtrainIndicate the location fingerprint information number of training dataset.λ is the weight factor of regularization, wpFor convolution
The vector of all weight factors and deviation composition in neural network.
Convolutional neural networks are exported for classification, one is connected by full articulamentum after afterbody CALP module and classifies
Device is located at the probability of different blocks for output mobile terminal, is given by
Y=Wvec { TL}+bL (16)
Wherein y is the probability that mobile terminal is located at different blocks, and W is full articulamentum weight, bLFor deviation.
Using L2 regularization, cross entropy loss function is given by
Wherein yiWithOutput after the true probability distribution and convolutional neural networks for respectively indicating training data are trained is general
Rate distribution.λ is the weight factor of L2 regularization, wpThe vector formed for weight factors all in convolutional neural networks and deviation.
Four, wireless location method
Wireless location method disclosed in the present embodiment includes off-line phase and on-line stage:
Off-line phase, according to the series of convolutional neural networks in concatenated convolutional neural network, step by step by cell coverage area
Several sub-blocks are divided into, as shown in Figure 2.Each sub-block includes piecemeal center and block margin, in cell coverage area
Characteristic point inside delimited at certain intervals, and the feature dot density at piecemeal center is greater than block margin.Divide for each grade, base station side
The offline sampling mobile terminal apparatus of device cooperation measures the location fingerprint information of characteristic point in each sub-block, is believed using location fingerprint
Prime convolutional neural networks, and the network knot relevant to the neuron of convolutional neural networks at different levels that training is obtained are worked as in breath training
Structure, weighted value and activation primitive storage are into database.
On-line stage exports convolutional neural networks for returning, and base station side device is corresponding by user side mobile terminal apparatus
Location fingerprint information input return output cascade convolutional neural networks convolutional neural networks at different levels, directly output user side move
The position coordinates of terminal installation.
On-line stage exports convolutional neural networks for classification, and base station side device is corresponding by user side mobile terminal apparatus
Location fingerprint information input classify output cascade convolutional neural networks convolutional neural networks at different levels, obtain user side mobile terminal
The probability of different blocks in cell coverage area is searched in several blocks of maximum probability by location fingerprint information matches
The most similar K characteristic point of rope location fingerprint information corresponding with user side mobile terminal apparatus is calculated using location estimation and is used
The position coordinates of family side mobile terminal apparatus, are represented by
Wherein Ξ and ΞiRespectively indicate the most similar ith feature point stored in user side mobile terminal and database
Location fingerprint information.piFor the position coordinates of the most similar ith feature point stored in database.w(Ξ,Ξi) it is special i-th
The corresponding weight of sign point coordinate, is represented by
Wherein J (Ξ, Ξi) it is joint angle time delay domain likeness coefficient.
In embodiment provided by the present invention, it should be understood that disclosed method is being not above essence of the invention
In mind and range, it can realize in other way.Current embodiment is a kind of exemplary example, be should not be taken as
Limitation, given particular content should in no way limit the purpose of the present invention.For example, multiple units or components can combine or can
To be integrated into another system, or some features can be ignored or not executed.
The above description is merely a specific embodiment, but scope of protection of the present invention is not limited thereto, any
Those familiar with the art in the technical scope disclosed by the present invention, can easily think of the change or the replacement, and should all contain
Lid is within protection scope of the present invention.Therefore, protection scope of the present invention should be based on the protection scope of the described claims.
Claims (10)
1. a kind of intelligent radio positioning system using large scale array antenna characterized by comprising
Location fingerprint information extraction modules, for extracting angle time delay domain channel strength square from uplink channel estimation result
Battle array is used as location fingerprint information；
Concatenated convolutional neural network module, including return output cascade convolutional neural networks module and/or classification output cascade volume
Product neural network module, for estimating position coordinates and/or the user of user side mobile terminal apparatus according to location fingerprint information
Side mobile terminal is located at the probability of different blocks in cell coverage area；
And database module, the concatenated convolutional neural network obtained for storing off-line phase training, and it is defeated using classifying
The position of characteristic point in the cell coverage area different blocks that off-line phase sampling obtains when concatenated convolutional neural network module out
Finger print information；
The recurrence output cascade convolutional neural networks module, for exporting the position coordinates of user side mobile terminal apparatus,
Cascaded topology returns device with output end convolutional neural networks by one or more levels convolutional neural networks classifier and is composed in series；
The classification output cascade convolutional neural networks module is located in cell coverage area for exporting user side mobile terminal
The probability of different blocks, cooperate estimation user side mobile terminal apparatus with location fingerprint information matches module, position estimation module
Position coordinates, cascaded topology is by one or more levels convolutional neural networks classifier and output end convolutional neural networks classifier
It is composed in series；
The location fingerprint information matches module joins for choosing from database with user side mobile terminal locations finger print information
The maximum several characteristic points of close angle degree time delay likeness coefficient, composition characteristic point set；
The position estimation module, for by several blocks of location fingerprint information matches block search maximum probability with use
The most similar several characteristic points of the corresponding location fingerprint information of family side mobile terminal apparatus calculate user side mobile terminal apparatus
Position coordinates.
2. a kind of intelligent radio positioning system using large scale array antenna according to claim 1, which is characterized in that
The data of off-line phase acquisition include each grade of segmentation divided step by step for cell coverage area, and the cooperation of base station side device is offline
The location fingerprint information of characteristic point in each sub-block that sampling mobile terminal apparatus measurement obtains；Wherein cell coverage area block
The series of division is determined according to the series of convolutional neural networks in concatenated convolutional neural network module；The corresponding spy of each grade of segmentation
Sign point position finger print information and position coordinates are used to train the convolutional neural networks of corresponding level.
3. a kind of intelligent radio positioning system using large scale array antenna according to claim 1, which is characterized in that
The process of angle time delay domain channel strength matrix is extracted in the location fingerprint information extraction modules are as follows: pass through mobile terminal upper
The uplink channel state information that row channel training obtains is converted into angle time delay domain channel response matrix；By angle time delay domain
Each element modulus in channel response matrix, and to continuously sample is averaged or is weighted and averaged several times, it is corresponding to obtain mobile terminal
Angle time delay domain channel strength matrix, and as location fingerprint information.
4. a kind of intelligent radio positioning system using large scale array antenna according to claim 1, which is characterized in that
The method of joint angle time delay likeness coefficient is calculated in the location fingerprint information matches module are as follows: definition dislocation factor of n, it is wrong
Position step-length L, n is the integer in section (- L+1, L-1), successively by each column vector in first location fingerprint information and second
N-component column vector takes inner product after respective column in location fingerprint information, and gained inner product is summed, by institute under the different dislocation factors
Obtain joint angle time delay likeness coefficient of the maximum value of sum as two location fingerprint information.
5. a kind of intelligent radio positioning system using large scale array antenna according to claim 1, which is characterized in that
It is to fill user side mobile terminal that the method for the position coordinates of user side mobile terminal apparatus is calculated in the position estimation module
Set the position coordinates of the corresponding location fingerprint information maximum K characteristic point of joint angle time delay likeness coefficient in the database
Weighted sum；The wherein corresponding weight coefficient calculation method of k-th of characteristic point are as follows: mobile eventually with k-th of characteristic point and user side
The joint angle time delay likeness coefficient of end device divided by whole K characteristic points and user side mobile terminal apparatus joint angle
The sum of time delay likeness coefficient.
6. a kind of intelligent radio positioning system using large scale array antenna according to claim 1, which is characterized in that
Including base station side device, offline sampling mobile terminal side device and customer mobile terminal side device；
The base station side device includes: uplink channel estimation module, is led for what is sent according to each mobile terminal received
Frequency signal implements channel state information acquisition；The location fingerprint information extraction modules；The concatenated convolutional neural network module；
The database module；The location fingerprint information matches module；The position estimation module；And information exchange module, it uses
In in off-line phase, the location information that offline sampling mobile terminal is sent is received, and mobile eventually to user side in on-line stage
End sends its location estimation result；
The offline sampling mobile terminal side device includes: offline map and navigation module, is used in off-line phase, in real time accurately
Obtain the location information of offline sampling mobile terminal；Driving device, for driving offline sampling mobile terminal to exist in off-line phase
It moves in cell coverage area, and is stopped in each characteristic point；Uplink pilot sending module is led for sending uplink
Frequency signal；Information exchange module, in off-line phase, the position for sending offline sampling mobile terminal to base station side device to be believed
Breath；
Customer mobile terminal side device includes: uplink pilot sending module, for sending uplink pilot；
And user side information exchange module, for sending what offline sampling mobile terminal was sent to base station side device in off-line phase
Location information.
7. a kind of intelligent radio localization method using large scale array antenna characterized by comprising
Cell coverage area is divided into several sub-blocks by off-line phase step by step, is divided for each grade, and MPS process model is measured
The location fingerprint information for enclosing characteristic point in each sub-block trains the convolutional neural networks of corresponding level using location fingerprint information,
And will the obtained network structure relevant to the neuron of convolutional neural networks at different levels of training, weighted value and activation primitive or with it is small
The location fingerprint information of characteristic point stores in database jointly in area's coverage area；
On-line stage, by the corresponding location fingerprint information input concatenated convolutional neural network of user side mobile terminal apparatus, estimation
The position coordinates of user side mobile terminal apparatus；
The concatenated convolutional neural network, including return output cascade convolutional neural networks and/or classification output cascade convolution mind
Through network；
The recurrence output cascade convolutional neural networks are cascaded for exporting the position coordinates of user side mobile terminal apparatus
Topology returns device with output end convolutional neural networks by one or more levels convolutional neural networks classifier and is composed in series；
The classification output cascade convolutional neural networks are located at difference in cell coverage area for exporting user side mobile terminal
The probability of block, cascaded topology is by one or more levels convolutional neural networks classifier and output end convolutional neural networks classifier
It is composed in series；
When using classification output cascade convolutional neural networks, by the corresponding location fingerprint information input of user side mobile terminal apparatus
Classification output cascade convolutional neural networks, obtain the probability that user side mobile terminal is located at different blocks in cell coverage area,
Position corresponding with user side mobile terminal apparatus is searched for by location fingerprint information matches in several blocks of maximum probability
The position of the maximum several characteristic points of finger print information joint angle time delay likeness coefficient, estimation user side mobile terminal apparatus is sat
Mark.
8. a kind of intelligent radio localization method using large scale array antenna according to claim 7, which is characterized in that
The location fingerprint information mentions for the angle time delay domain channel strength matrix extracted from uplink channel estimation result
Taking process is to convert angle time delay domain by the uplink channel state information that up channel training obtains for mobile terminal
Channel response matrix；By each element modulus in angle time delay domain channel response matrix, and to continuously several times sample be averaged or
Weighted average, obtains the corresponding angle time delay domain channel strength matrix of the mobile terminal, and as location fingerprint information.
9. a kind of intelligent radio localization method using large scale array antenna according to claim 7, which is characterized in that
The joint angle time delay likeness coefficient, calculation method be, definition dislocation factor of n, misplace step-length L, n be section (- L+1,
L-1 the integer in), successively by each column vector in first location fingerprint information and respective column in second location fingerprint information it
N-component column vector takes inner product afterwards, and gained inner product is summed, using the maximum value of gained sum under the different dislocation factors as two positions
Set the joint angle time delay likeness coefficient of finger print information.
10. a kind of intelligent radio localization method using large scale array antenna according to claim 7, feature exist
In, estimate that the method for position coordinates of user side mobile terminal apparatus is, it will joint angle time delay similitude system in the database
The position coordinates weighted sum of the maximum K characteristic point of number, wherein the corresponding weight coefficient calculation method of k-th of characteristic point is,
With the joint angle time delay likeness coefficient of k-th characteristic point and user side mobile terminal apparatus divided by whole K characteristic points and
The sum of joint angle time delay likeness coefficient of user side mobile terminal apparatus.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910166653.XA CN109922427B (en) | 2019-03-06 | 2019-03-06 | Intelligent wireless positioning system and method using large-scale array antenna |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910166653.XA CN109922427B (en) | 2019-03-06 | 2019-03-06 | Intelligent wireless positioning system and method using large-scale array antenna |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109922427A true CN109922427A (en) | 2019-06-21 |
| CN109922427B CN109922427B (en) | 2020-09-11 |
Family
ID=66963493
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910166653.XA Active CN109922427B (en) | 2019-03-06 | 2019-03-06 | Intelligent wireless positioning system and method using large-scale array antenna |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109922427B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111313943A (en) * | 2020-02-20 | 2020-06-19 | 东南大学 | Three-dimensional positioning method and device under deep learning assisted large-scale antenna array |
| CN112637950A (en) * | 2020-12-23 | 2021-04-09 | 中国人民解放军陆军工程大学 | Fingerprint positioning method based on angle similarity |
| CN112995892A (en) * | 2021-02-08 | 2021-06-18 | 东南大学 | Large-scale MIMO fingerprint positioning method based on complex neural network |
| CN113038595A (en) * | 2020-12-30 | 2021-06-25 | 东南大学 | PQ and CNN-based rapid fingerprint positioning method |
| CN113300746A (en) * | 2021-05-24 | 2021-08-24 | 内蒙古大学 | Millimeter wave MIMO antenna and hybrid beam forming optimization method and system |
| WO2021237463A1 (en) * | 2020-05-26 | 2021-12-02 | Telefonaktiebolaget Lm Ericsson (Publ) | Method and apparatus for position estimation |
| CN114828266A (en) * | 2022-05-27 | 2022-07-29 | 电子科技大学 | Intelligent resource allocation method for optical and wireless fusion access |
| CN117222005A (en) * | 2023-11-08 | 2023-12-12 | 网络通信与安全紫金山实验室 | Fingerprint positioning method, fingerprint positioning device, electronic equipment and storage medium |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104569907A (en) * | 2014-09-04 | 2015-04-29 | 深圳市金溢科技股份有限公司 | Wireless positioning method and system based on neural network and road side unit |
| CN106059972A (en) * | 2016-05-25 | 2016-10-26 | 北京邮电大学 | Modulation identification method under MIMO related channel based on machine learning algorithm |
| CN107592611A (en) * | 2017-09-11 | 2018-01-16 | 东南大学 | The extensive mimo system wireless location method in broadband and system |
| CN107765104A (en) * | 2017-09-04 | 2018-03-06 | 华为技术有限公司 | The method and school that a kind of phased array school is surveyed survey device |
| CN108037520A (en) * | 2017-12-27 | 2018-05-15 | 中国人民解放军战略支援部队信息工程大学 | Direct deviations modification method based on neutral net under the conditions of array amplitude phase error |
| CN108182474A (en) * | 2017-12-27 | 2018-06-19 | 中国人民解放军战略支援部队信息工程大学 | Based on the direct localization method of multiple target for not correcting array and neural network |
| WO2018160465A1 (en) * | 2017-03-01 | 2018-09-07 | Intel Corporation | Neural network-based systems for high speed data links |
| CN108696932A (en) * | 2018-04-09 | 2018-10-23 | 西安交通大学 | It is a kind of using CSI multipaths and the outdoor fingerprint positioning method of machine learning |
| CN109075828A (en) * | 2016-04-26 | 2018-12-21 | 三星电子株式会社 | For realizing the method and apparatus of uplink MIMO |
| CN109379752A (en) * | 2018-09-10 | 2019-02-22 | 中国移动通信集团江苏有限公司 | Optimization method, device, equipment and the medium of Massive MIMO |
-
2019
- 2019-03-06 CN CN201910166653.XA patent/CN109922427B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104569907A (en) * | 2014-09-04 | 2015-04-29 | 深圳市金溢科技股份有限公司 | Wireless positioning method and system based on neural network and road side unit |
| CN109075828A (en) * | 2016-04-26 | 2018-12-21 | 三星电子株式会社 | For realizing the method and apparatus of uplink MIMO |
| CN106059972A (en) * | 2016-05-25 | 2016-10-26 | 北京邮电大学 | Modulation identification method under MIMO related channel based on machine learning algorithm |
| WO2018160465A1 (en) * | 2017-03-01 | 2018-09-07 | Intel Corporation | Neural network-based systems for high speed data links |
| CN107765104A (en) * | 2017-09-04 | 2018-03-06 | 华为技术有限公司 | The method and school that a kind of phased array school is surveyed survey device |
| CN107592611A (en) * | 2017-09-11 | 2018-01-16 | 东南大学 | The extensive mimo system wireless location method in broadband and system |
| CN108037520A (en) * | 2017-12-27 | 2018-05-15 | 中国人民解放军战略支援部队信息工程大学 | Direct deviations modification method based on neutral net under the conditions of array amplitude phase error |
| CN108182474A (en) * | 2017-12-27 | 2018-06-19 | 中国人民解放军战略支援部队信息工程大学 | Based on the direct localization method of multiple target for not correcting array and neural network |
| CN108696932A (en) * | 2018-04-09 | 2018-10-23 | 西安交通大学 | It is a kind of using CSI multipaths and the outdoor fingerprint positioning method of machine learning |
| CN109379752A (en) * | 2018-09-10 | 2019-02-22 | 中国移动通信集团江苏有限公司 | Optimization method, device, equipment and the medium of Massive MIMO |
Non-Patent Citations (2)
| Title |
|---|
| VLADIMIR SAVIC，ERIK G. LARSSON: "Fingerprinting-Based Positioning in Distributed Massive MIMO Systems", 《 2015 IEEE 82ND VEHICULAR TECHNOLOGY CONFERENCE (VTC2015-FALL)》 * |
| 尤力，高西奇: "毫米波大规模MIMO无线传输关键技术", 《中兴通讯技术》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111313943A (en) * | 2020-02-20 | 2020-06-19 | 东南大学 | Three-dimensional positioning method and device under deep learning assisted large-scale antenna array |
| WO2021237463A1 (en) * | 2020-05-26 | 2021-12-02 | Telefonaktiebolaget Lm Ericsson (Publ) | Method and apparatus for position estimation |
| CN112637950A (en) * | 2020-12-23 | 2021-04-09 | 中国人民解放军陆军工程大学 | Fingerprint positioning method based on angle similarity |
| CN112637950B (en) * | 2020-12-23 | 2022-09-27 | 中国人民解放军陆军工程大学 | Fingerprint positioning method based on angle similarity |
| CN113038595B (en) * | 2020-12-30 | 2022-07-01 | 东南大学 | PQ and CNN-based rapid fingerprint positioning method |
| CN113038595A (en) * | 2020-12-30 | 2021-06-25 | 东南大学 | PQ and CNN-based rapid fingerprint positioning method |
| CN112995892A (en) * | 2021-02-08 | 2021-06-18 | 东南大学 | Large-scale MIMO fingerprint positioning method based on complex neural network |
| CN112995892B (en) * | 2021-02-08 | 2022-11-29 | 东南大学 | Large-scale MIMO fingerprint positioning method based on complex neural network |
| CN113300746A (en) * | 2021-05-24 | 2021-08-24 | 内蒙古大学 | Millimeter wave MIMO antenna and hybrid beam forming optimization method and system |
| CN114828266A (en) * | 2022-05-27 | 2022-07-29 | 电子科技大学 | Intelligent resource allocation method for optical and wireless fusion access |
| CN114828266B (en) * | 2022-05-27 | 2023-04-07 | 电子科技大学 | Intelligent resource allocation method for optical and wireless fusion access |
| CN117222005A (en) * | 2023-11-08 | 2023-12-12 | 网络通信与安全紫金山实验室 | Fingerprint positioning method, fingerprint positioning device, electronic equipment and storage medium |
| CN117222005B (en) * | 2023-11-08 | 2024-04-02 | 网络通信与安全紫金山实验室 | Fingerprint positioning method, fingerprint positioning device, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109922427B (en) | 2020-09-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109922427A (en) | Utilize the intelligent radio positioning system and method for large scale array antenna | |
| Sun et al. | Fingerprint-based localization for massive MIMO-OFDM system with deep convolutional neural networks | |
| Li et al. | AF-DCGAN: Amplitude feature deep convolutional GAN for fingerprint construction in indoor localization systems | |
| CN108696932B (en) | Outdoor fingerprint positioning method using CSI multipath and machine learning | |
| CN110381436B (en) | Rapid fingerprint positioning method based on large-scale MIMO single station system | |
| CN107592611A (en) | The extensive mimo system wireless location method in broadband and system | |
| CN111479231B (en) | Indoor fingerprint positioning method for millimeter wave large-scale MIMO system | |
| Dvorecki et al. | A machine learning approach for Wi-Fi RTT ranging | |
| CN107992882A (en) | A kind of occupancy statistical method based on WiFi channel condition informations and support vector machines | |
| CN107613559A (en) | A kind of DOA fingerprint base localization methods based on 5G signals | |
| CN112995892B (en) | Large-scale MIMO fingerprint positioning method based on complex neural network | |
| CN107255793A (en) | A kind of array df method and device for wideband OFDM signal of communication | |
| CN112637950B (en) | Fingerprint positioning method based on angle similarity | |
| CN101425855B (en) | Wideband channel parameter extraction method, wideband channel simulation method and apparatus thereof | |
| CN110213185A (en) | A kind of three dimensional channel method for parameter estimation based on atom norm minimum | |
| CN110222776A (en) | A kind of indoor Passive Location based on CSI finger print information | |
| CN110113088A (en) | A kind of divergence type numerical model analysis antenna system direction of arrival intelligence estimation method | |
| CN107135023A (en) | Three-dimensional training code book design method and beam alignment for millimeter-wave communication system | |
| CN108040368A (en) | A kind of unmanned plane distribution method of time frequency resources declined based on block coordinate | |
| CN107451605A (en) | A kind of simple target recognition methods based on channel condition information and SVMs | |
| Yang et al. | Model-based learning network for 3-D localization in mmWave communications | |
| CN110049549A (en) | More fusion indoor orientation methods and its system based on WiFi fingerprint | |
| CN106291453A (en) | A kind of two-dimentional ripple based on beam space conversion reaches angle assay method | |
| CN104994041B (en) | Three-dimensional extensive Bo Dabo digression combined estimation method of the antenna system based on pilot tone | |
| Kia et al. | A CNN approach for 5G mm wave positioning using beamformed CSI measurements |
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 |