CN109803233A - CSI fingerprint indoor orientation method based on DHNN - Google Patents
CSI fingerprint indoor orientation method based on DHNN Download PDFInfo
- Publication number
- CN109803233A CN109803233A CN201910146604.XA CN201910146604A CN109803233A CN 109803233 A CN109803233 A CN 109803233A CN 201910146604 A CN201910146604 A CN 201910146604A CN 109803233 A CN109803233 A CN 109803233A
- Authority
- CN
- China
- Prior art keywords
- dhnn
- network
- csi
- data
- reference point
- 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
Landscapes
- Position Fixing By Use Of Radio Waves (AREA)
Abstract
CSI fingerprint indoor orientation method based on DHNN, acquires and extracts the CSI data of reference point, obtain primary data set；Using low-pass filter in primary data set redundant data and partial noise handle to obtain the first intermediate data set；The reference point locations information in the first intermediate data set is characterized by phase difference correction, the CSI data for not meeting another part noise of given feature in the first intermediate data set and being influenced by multipath effect is removed and obtains the second intermediate data set；Using the CSI of reference point in the second intermediate data set as location fingerprint information, fingerprint database is constructed；Attractor is set by the CSI of each reference point of fingerprint database, using attractor as DHNN network convergence judgment basis；The CSI data of test point are input in DHNN network and carry out convergence judgement with the data in attractor, obtain positioning result.It ensure that the accuracy of finger print data, positioning accuracy significantly improve.
Description
Technical field
The present embodiments relate to indoor positioning technologies fields, and in particular to fixed in a kind of CSI fingerprint room based on DHNN
Position method.
Background technique
As location based service (Location Based Service, LBS) becomes increasingly popular, to based on position in life
The business demand set also is continuously increased.In outdoor environment, GPS, GLONASS, Beidou satellite navigation system etc. are some mature to be defended
The extensive use of star positioning system obtains more accurate positioning for people and navigation Service is provided convenience.However indoor ring
Under border, since satellite-signal is weak, cannot penetrate the problems such as building, global position system can not effectively work, therefore grind
Study carefully high-precision, high reliability, low cost indoor locating system be to current indoor location technology propose new challenge.
In recent years, indoor Wi-Fi's is widely available, so that the indoor positioning technologies based on Wi-Fi constantly develop.Compare
Typical technology mainly have based on received signal strength indicator (Received Signal Strength Indication,
RSSI), it is based on the positioning of channel state information (CSI).But indoors in environment, due to the influence of barrier, RSSI can be generated
The influence of certain deviation and the interference and indoor multipath effect that are highly prone to other signals, so enough essences can not be provided
Exactness and reliability.For having used the Wi-Fi signal of IEEE802.11n transport protocol, it can be by modifying wireless network card
Driving is to obtain orthogonal frequency division multiplexing (Orthogonal Frequency Division Multiplexing, OFDM) subcarrier
In CSI, the CSI parsed from physical layer can describe channel characteristics and state between signal transmitting terminal and receiving end.
Compared with RSSI, CSI has certain multi-path resolved ability, can perceive the faint fluctuation of signal on propagation path, compare
There are higher sensitivity, bigger sensing range and stronger sensing reliability in RSSI, CSI.
In general, it distinguishes according to positioning principle, can be divided into based on the indoor positioning technologies under Wi-Fi based on propagating mode
The positioning of type and positioning based on fingerprint.However physical model differs larger and some disturbing factors with actual environment is not considered
Inside, often locating effect is more undesirable compared with fingerprint location method.Traditional fingerprint indoor orientation method based on CSI
In, fingerprint accuracy is poor, classification and matching effect is not high, easily it is affected by environment the problems such as presence it can not be provided preferably
Locating effect and higher positioning accuracy.
Summary of the invention
For this purpose, the embodiment of the present invention provides a kind of CSI fingerprint indoor orientation method based on DHNN, finger print data ensure that
Accuracy, position error is greatly lowered, positioning accuracy is significantly increased.
To achieve the goals above, the embodiment of the present invention provides the following technical solutions: fixed in the CSI fingerprint room based on DHNN
Position method, including off-line phase and on-line stage；
The off-line phase includes:
It acquires reference Point C SI data: continuous data packet being sent to computer end by wireless access point, acquires and extracts ginseng
The CSI data of examination point, obtain primary data set；
Low-pass filter processing: using low-pass filter to the redundant data and partial noise in the primary data set
It is handled, the redundant data in primary data set is removed to obtain the first intermediate data set with partial noise；
Phase difference correction: characterizing the reference point locations information in the first intermediate data set by phase difference correction,
Remove the CSI number for not meeting another part noise of given feature in the first intermediate data set and being influenced by multipath effect
According to obtaining the second intermediate data set；
Building location information fingerprint database: the CSI of reference point in the second intermediate data set is referred to as position
Line information constructs fingerprint database；
The on-line stage includes:
Attractor is set: setting attractor for the CSI of each reference point of the fingerprint database of off-line phase,
Using the attractor as DHNN network convergence judgment basis；
Collecting test point CSI data: continuous data packet is sent to computer end by wireless access point, acquires and extracts survey
The CSI data of pilot；
The judgement of DHNN network convergence: the CSI data of the test point are input in the DHNN network and the attraction
Data in son carry out convergence judgement, obtain positioning result.
As the preferred embodiment of the CSI fingerprint indoor orientation method based on DHNN, the low-pass filter is fertile using Bart
This filter, according to the working frequency of the wireless access point, pedestrian walking speed to the cutoff frequency of the low-pass filter
It is configured.
As the preferred embodiment of the CSI fingerprint indoor orientation method based on DHNN, when acquiring reference Point C SI data, same
The CSI data of four different directions are selected to be acquired under one reference point, when the position information process for same reference point
After four direction gap exceeds to threshold value afterwards, reference Point C SI data are resurveyed.
As the preferred embodiment of the CSI fingerprint indoor orientation method based on DHNN, when phase difference correction, for reference point
Phase is made the difference.
As the preferred embodiment of the CSI fingerprint indoor orientation method based on DHNN, in the DHNN network, any one mind
Output x through memberiPass through connection weight ωijFeed back to all neuron xjAs input, one is arranged to each neuron
Threshold kjFor reflecting the control to input noise；
DHNN network is denoted as N=(W, K), the state for the feedback network that the set of entire neuron state is constituted is denoted as
X=[x1,x2,…,xn]T, the state input of feedback network is denoted as X (0)=[x as the state initial value of network1(0),x2
(0),…,xn(0)]T；
The neuron xjState change follow xj=f (netj) j=1,2 ..., n
Wherein, f (netj) be transfer function, the transfer function symbolization function of DHNN network:
Net input netjAre as follows:
When DHNN network stabilization, stable state is indicated as the output of whole network are as follows:
As the preferred embodiment of the CSI fingerprint indoor orientation method based on DHNN, when the state X of the DHNN network meets
When X=f (WX-K), using X as the attractor of DHNN network, wherein W is connection weight matrix, and K is the control threshold of input noise
Value.
As the preferred embodiment of the CSI fingerprint indoor orientation method based on DHNN, when DHNN network convergence judges, definition
The energy function of DHNN network are as follows:
Wherein, X is the attractor of DHNN network, and W is connection weight matrix, and K is the control threshold of input noise.
As the preferred embodiment of the CSI fingerprint indoor orientation method based on DHNN, when DHNN network is according to asynchronous system work
Make and connection weight matrix W be symmetrical matrix when, DHNN network convergence a to attractor；
When DHNN network adjusts network state according to the method for synchronization, and connection weight matrix W is nonnegative definite symmetrical matrix, DHNN
Network convergence is to a constant value.
The embodiment of the present invention has the advantages that firstly, pre-processing to collected CSI data, raw data packets
Contained some redundant datas and noise and some data influenced by multipath effect, after pretreatment can by redundant data with
Partial noise removal, obtains relatively accurate CSI.Secondly, by the method for phase difference correction, it can be by each reference point
The character representation of location information comes out, and the data for not meeting feature can be removed by this method, to make each reference point
Finger print information can be relatively accurate.Data in off-line phase fingerprint base can be arranged to attractor, will acquired in real time by on-line stage
The data arrived carry out convergence judgement using DHNN as input, finally determine positioning result.And by experiment to the technology of the present invention
The localization method of scheme has carried out verifying and comparative analysis, and Comprehensive Experiment result can illustrate, compared to current some fingers
Line indoor orientation method, the method for the present invention positioning accuracy is higher, while lower to the quantitative requirement of data.
Detailed description of the invention
It, below will be to embodiment party in order to illustrate more clearly of embodiments of the present invention or technical solution in the prior art
Formula or attached drawing needed to be used in the description of the prior art are briefly described.It should be evident that the accompanying drawings in the following description is only
It is merely exemplary, it for those of ordinary skill in the art, without creative efforts, can also basis
The attached drawing of offer, which is extended, obtains other implementation attached drawings.
Fig. 1 is the CSI fingerprint indoor orientation method technical framework diagram based on DHNN provided in the embodiment of the present invention；
Fig. 2 is the CSI fingerprint indoor orientation method implementation process diagram based on DHNN provided in the embodiment of the present invention；
Fig. 3 is the topological structure schematic diagram of the DHNN provided in the embodiment of the present invention；
Fig. 4 is CSI fingerprint indoor orientation method technical effect of the verifying based on DHNN provided in the embodiment of the present invention
Environment place arrangement schematic diagram；
Fig. 5 is the CSI fingerprint indoor orientation method and conventional solution Contrast on effect under stable environment based on DHNN；
Fig. 6 is CSI fingerprint indoor orientation method and the traditional technology side acquired in the lesser situation of data volume based on DHNN
The comparison of case accuracy rate.
Specific embodiment
Embodiments of the present invention are illustrated by particular specific embodiment below, those skilled in the art can be by this explanation
Content disclosed by book is understood other advantages and efficacy of the present invention easily, it is clear that described embodiment is the present invention one
Section Example, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art are not doing
Every other embodiment obtained under the premise of creative work out, shall fall within the protection scope of the present invention.
Referring to figure 1, figure 2 and figure 3, a kind of CSI fingerprint indoor orientation method based on DHNN, including off-line phase L are provided
With on-line stage Z；
The off-line phase L includes:
L01: acquisition reference Point C SI data: continuous data packet is sent to computer end by wireless access point, acquires and mentions
The CSI data for taking reference point, obtain primary data set；
L02: low-pass filter processing: using low-pass filter in the primary data set redundant data and part
Noise is handled, and the redundant data in primary data set is removed to obtain the first intermediate data set with partial noise；
L03: the letter of the reference point locations in the first intermediate data set phase difference correction: is characterized by phase difference correction
Breath removes and in the first intermediate data set does not meet another part noise of given feature and influenced by multipath effect
CSI data obtain the second intermediate data set；
L04: building location information fingerprint database: using the CSI of reference point in the second intermediate data set as position
Finger print information is set, fingerprint database is constructed；
The on-line stage Z includes:
Z01: setting attractor: the CSI of each reference point of the fingerprint database of off-line phase is set as inhaling
Introduction, using the attractor as DHNN network convergence judgment basis；
Z02: collecting test point CSI data: continuous data packet is sent to computer end by wireless access point, acquires and mentions
Take the CSI data of test point；
Z03:DHNN network convergence judgement: by the CSI data of the test point be input in the DHNN network with it is described
Data in attractor carry out convergence judgement, obtain positioning result.
In one embodiment of CSI fingerprint indoor orientation method based on DHNN, due to the complexity of indoor environment, and
The fine granularity of CSI, collected CSI contains noise and some generates because of multipath effect under usual indoor environment
Redundant data.Low pass rate filter allows the signal lower than cutoff frequency to pass through, and prevents the signal for being higher than cutoff frequency.It is indoor
The frequency that some noise frequencies will walk much higher than people, therefore for collected CSI, the present embodiment utilizes Butterworth low pass
Filter is made an uproar to filter.As needed, it is necessary to which different cutoff frequency f is setj, it is assumed that the speed of walking is about 1m/s, if nothing
Line access point works under the frequency of 5.32GHz, according to wavelength=velocity of wave/frequency, then
SoDue to needing to acquire the finger print information of each reference point, walked according to indoor people
Capable minimum speed goes to calculate corresponding wavelength and frequency, really waits tester to stablize in a certain reference as far as possible when measurement
Data acquisition is carried out again when point, to guarantee the quality of data.
In one embodiment of CSI fingerprint indoor orientation method based on DHNN, when acquiring reference Point C SI data, same
The CSI data of four different directions are selected to be acquired under one reference point, when the position information process for same reference point
After four direction gap exceeds to threshold value afterwards, reference Point C SI data are resurveyed.When phase difference correction, for reference point
Phase is made the difference.
Specifically, being selected under the same reference point when effect in order to guarantee phase difference correction, first selection reference point
The data in four different directions are acquired.Secondly, for the location information of same reference point, if four direction is poor after processing
Away from larger, then resurveyed, guarantee that the location information of each reference point is relatively accurate with this.It is same used in the present embodiment
Computational algorithm under reference point is as shown in algorithm 1:
Since the phase difference under different reference points is also different, provided can preferably to distinguish different reference points
More effective informations.Linear change is carried out using to all subcarriers, and antenna does phase difference to obtain effective fortune two-by-two
Dynamic information can pass through elimination phase offset as caused by clock skew and unknown constant after linear change
Relatively unordered phase information can be distributed concentration after antenna makes the difference two-by-two.Calculating under different reference points used in the present embodiment is calculated
Method is as shown in algorithm 2.
For indoor orientation method, clock skew is affected for personnel's detection of motion state, but due to
The influence of some uncertain factors, phase can shift in the selection of reference point and the collection process of data, and antenna is done two-by-two
Poor phase calibration both can carry out respective antenna with different data for same reference point and make the difference, can also be to identical data
Different antennae makes the difference.Similarly, for different reference points, correlation method can also be used to be corrected to data.
Specifically, DHNN network is a kind of single layer unity feedback network referring to Fig. 3, n neuron is shared, its weight is not
It is to be obtained by repetition learning, but calculated according to certain implementation rule, it is therefore intended that change the state of network.
In DHNN network, the output x of any one neuroniPass through connection weight ωijFeed back to all neuron xjIt is right as input
A threshold k is arranged in each neuronjFor reflecting the control to input noise.DHNN network is denoted as N=(W, K), it will
The state for the feedback network that the set of entire neuron state is constituted is denoted as X=[x1,x2,…,xn]T, by the state of feedback network
It inputs and is denoted as X (0)=[x as the state initial value of network1(0),x2(0),…,xn(0)]T；
Whole network enters dynamic evolution process from original state, during this period net under the excitation of external world's input
The state of each of network neuron is all in continuous variation.Neuron xjState change follow:
xj=f (netj) j=1,2 ..., n (1)
Wherein, f (netj) be transfer function, the transfer function symbolization function of DHNN network:
Net input netjAre as follows:
For entire DHNN, generally there is ωii=0, ωij=ωji.When network stabilization, the state of each neuron is not
Change again, stable state at this time is exactly the output of whole network, and stable state is indicated as the output of whole network are as follows:
In one embodiment of CSI fingerprint indoor orientation method based on DHNN, when the state X of the DHNN network meets
When X=f (WX-K), using X as the attractor of DHNN network, wherein W is connection weight matrix, and K is the control threshold of input noise
Value.
Specifically, DHNN network is a kind of network that can store several pre-set stable points.When operation, when to
After one initial input of the role of network, whole network just outputs it the input fed back as next time, each mind
Input through member does not include self feed back.After recycling several times, in the case where network structure meets some requirements, final network meeting
Stablize in a certain preset stable point.If the state X of network meets X=f (WX-K), X is referred to as the attractor of network, i.e.,
State X when network reaches stable, the referred to as attractor of network.For DHNN network, if being adjusted according to asynchronous system network-like
State (only one each neuron carries out state adjustment, and the state of other neurons remains unchanged), and connection weight matrix W is pair
Claim battle array, then for any initial state, whole network finally can all converge to an attractor；If being adjusted according to the method for synchronization network-like
State (all neurons adjust state simultaneously), and connection weight matrix W is nonnegative definite symmetrical matrix, then for any initial state, network is all
Finally converge to an attractor.
Specifically, defining the energy function of DHNN network when DHNN network convergence judges are as follows:
Wherein, X is the attractor of DHNN network, and W is connection weight matrix, and K is the control threshold of input noise.
The knots modification for enabling network energy is Δ E, and the state knots modification of network is Δ X, then has:
Δ E (t)=E (t+1)-E (t) (5)
Δ X (t)=X (t+1)-X (t) (6)
By (5) (6) Shi Ke get:
If asynchronously working, then adjusting state in t-th only one neuron of moment, it is assumed that the neuron is
J, then:
Δ X (t)=[0 ..., 0, Δ xj(t),0,…,0]T (8)
It brings (8) formula into (7) formula, then has:
Because self feed back is not present in each neuron, there is ωjj=0, by (2) (3) Shi Ke get: Δ E (t)=- Δ xj(t)
netj(t).Analysis can obtain, and can be respectively as follows: there are three types of situation
xj(t)=- 1, xj(t+1)=1；
xj(t)=1, xj(t+1)=- 1；
xj(t)=xj(t+1)。
Find there is Δ E (t)≤0 under any circumstance after verifying, in network dynamic differentiation, energy is always continuous
Decline is constant, and final network can all converge on a constant.I.e. network works according to asynchronous system and connection weight matrix W is
When symmetrical matrix, network will finally converge to an attractor.Similarly, if adjusting network state, and connection weight square by the method for synchronization
Battle array W is nonnegative definite symmetrical matrix, can be derived:
According to verifying before, it can be found that-Δ xj(t)netj(t)
≤ 0, and W is nonnegative definite symmetrical matrix, then Δ E (t)≤0, E (t) will finally converge to a constant value.
In one embodiment of CSI fingerprint indoor orientation method based on DHNN, when DHNN network is according to asynchronous system work
Make and connection weight matrix W be symmetrical matrix when, DHNN network convergence a to attractor.When DHNN network is adjusted according to the method for synchronization
Network state, and connection weight matrix W be nonnegative definite symmetrical matrix when, DHNN network convergence a to constant value.Specifically, this implementation
Example using real-time measurement input of the CSI as DHNN network, no matter network use either synchronously or asynchronously mode works, when energy most
As soon as be stable at constant eventually, this constant corresponds to the minimum state of network energy, also just corresponds to the attraction in network
Son.In any case, if network just will appear self-holding concussion, therefore in reality it cannot be guaranteed that W is nonnegative definite symmetrical matrix
In testing, design DHNN works according to asynchronous system, and each of fingerprint database reference point is respectively set to attract
When data that are sub and calculating its weight pilot to be measured are input in network, only one neuron carries out state tune each time
Whole, the state of other neurons remains unchanged, and after comparing one by one, obtains the information of estimated location, it has in contrast
Higher stability can generate lesser position error than synchronous working mode.
Referring to fig. 4, Fig. 5 and Fig. 6, sample plot point select indoor environment, and entire indoor environment size is 10.01 × 6.9m2,
Experimental Area is divided into 5 × 5 as shown in figure 4, being provided with 25 grid spaces by experimental situation schematic diagram, totally 25 squares,
Each square area size is 0.8m × 0.8m, is workbench on the left of Experimental Area, placed on workbench equipped with Intel
The desktop computer of 5300 network interface cards, CPU model Intel Core i3-4150, operating system Ubuntu11.04；Experimental Area
AP is placed on the workbench of right side, and ((wireless access point), the end AP are a router, model TL-WDR5300, two workbench
With high, equipment is arranged to ap mode when test, working frequency range is at 5GHz.
Test the method DHNN and the DeepFi and method SVR based on support vector regression for being proposed the embodiment of the present invention
Performance has been carried out under different scenes to compare.In a stable environment, experimenter does not carry mobile device, Experimental Area in addition to
It walks about, maintain to a certain extent spacious and stablizes without anyone except data collector；Under interference environment, people is tested
Member carries mobile device, and someone walks about by Experimental Area.The embodiment of the present invention is in test, to be in a square (0.8m)
Allowable range of error, i.e. position error position successfully in 0.8m, are otherwise denoted as positioning failure.As shown in figure 5, in stabilizing ring
Under border, 3 kinds of methods locating accuracy all with higher, because all using the building and deep learning, nerve net of fingerprint base
The algorithm of network is handled and is modeled to data, so the sample size of fingerprint base is big, accuracy relative to other methods
It is higher.But under interference environment, 3 kinds of methods all receive different degrees of influence.Wherein, SVR is the most serious by being influenced,
Locating accuracy is fallen below less than 30%, and since other signals and multipath effect bring interfere, RSS information becomes vulnerable to environment
Change and fluctuate it is larger, so accuracy rate decline it is larger.The DHNN method that DeepFi and the embodiment of the present invention propose equally all declines
To 50% hereinafter, being still better than SVR method, this is because both of which use CSI be used as finger print information, for position
For, need more fine-grained CSI to improve precision.On the other hand, although CSI information is biggish since environmental disturbances can generate
Variation, but because both of which uses tagsort, it is possible to reduce influence of the environmental change for locating accuracy.
The DHNN method that DeepFi is proposed with the embodiment of the present invention in both environments is not much different, this is because DeepFi
The information collection of big data quantity has been carried out, the cost and complexity of positioning has been improved, constructs accurate big data quantity
Location fingerprint library.And the embodiment of the present invention propose DHNN method, in information collection will more lightweight, eased reality
More accurate locating effect is showed, as shown in fig. 6, the DHNN method that the embodiment of the present invention proposes is lesser in acquisition data volume
In the case of, other two methods are compared, higher locating accuracy can be obtained.It is positioned and is missed using the intermediate value of the localization method of SVR
Difference is 2m, uses the intermediate value position error of the localization method of DeepFi for 1.71m, the DHNN method that the embodiment of the present invention proposes
Intermediate value position error is 1.6m, therefore the method that this patent proposes is better than other two methods.
Although above having used general explanation and specific embodiment, the present invention is described in detail, at this
On the basis of invention, it can be made some modifications or improvements, this will be apparent to those skilled in the art.Therefore,
These modifications or improvements without departing from theon the basis of the spirit of the present invention are fallen within the scope of the claimed invention.
Claims (8)
1. the CSI fingerprint indoor orientation method based on DHNN, which is characterized in that including off-line phase and on-line stage；
The off-line phase includes:
It acquires reference Point C SI data: continuous data packet being sent to computer end by wireless access point, acquires and extracts reference point
CSI data, obtain primary data set；
Low-pass filter processing: using low-pass filter to the redundant data and partial noise progress in the primary data set
Redundant data in primary data set is removed to obtain the first intermediate data set by processing with partial noise；
Phase difference correction: characterizing the reference point locations information in the first intermediate data set by phase difference correction, removal
The CSI data for not meeting another part noise of given feature in the first intermediate data set and being influenced by multipath effect obtain
To the second intermediate data set；
It constructs location information fingerprint database: believing the CSI of reference point in the second intermediate data set as location fingerprint
Breath constructs fingerprint database；
The on-line stage includes:
Attractor is set: attractor is set by the CSI of each reference point of the fingerprint database of off-line phase, by institute
Attractor is stated as DHNN network convergence judgment basis；
Collecting test point CSI data: continuous data packet is sent to computer end by wireless access point, acquires and extracts test point
CSI data；
DHNN network convergence judgement: by the CSI data of the test point be input in the DHNN network in the attractor
Data carry out convergence judgement, obtain positioning result.
2. the CSI fingerprint indoor orientation method according to claim 1 based on DHNN, which is characterized in that the low pass filtered
Wave device uses Butterworth filter, according to the working frequency of the wireless access point, pedestrian walking speed to the low pass filtered
The cutoff frequency of wave device is configured.
3. the CSI fingerprint indoor orientation method according to claim 1 based on DHNN, which is characterized in that acquisition reference point
When CSI data, the CSI data of four different directions are selected to be acquired under the same reference point, when for same reference point
Position information process after four direction gap exceed to threshold value after, resurvey reference Point C SI data.
4. the CSI fingerprint indoor orientation method according to claim 1 based on DHNN, which is characterized in that phase difference correction
When, the phase of reference point is made the difference.
5. the CSI fingerprint indoor orientation method according to claim 1 based on DHNN, which is characterized in that the DHNN net
In network, the output x of any one neuroniPass through connection weight ωijFeed back to all neuron xjAs input, to each
A threshold k is arranged in neuronjFor reflecting the control to input noise；
DHNN network is denoted as N=(W, K), the state for the feedback network that the set of entire neuron state is constituted is denoted as X=
[x1,x2,…,xn]T, the state input of feedback network is denoted as X (0)=[x as the state initial value of network1(0),x2
(0),…,xn(0)]T；
The neuron xjState change follow xj=f (netj) j=1,2 ..., n
Wherein, f (netj) be transfer function, the transfer function symbolization function of DHNN network:
Net input netjAre as follows:
When DHNN network stabilization, stable state is indicated as the output of whole network are as follows:
6. the CSI fingerprint indoor orientation method according to claim 1 based on DHNN, which is characterized in that as the DHNN
When the state X of network meets X=f (WX-K), using X as the attractor of DHNN network, wherein W is connection weight matrix, and K is defeated
Enter the control threshold of noise.
7. the CSI fingerprint indoor orientation method according to claim 1 based on DHNN, which is characterized in that DHNN network is received
When holding back judgement, the energy function of DHNN network is defined are as follows:
Wherein, X is the attractor of DHNN network, and W is connection weight matrix, and K is the control threshold of input noise.
8. the CSI fingerprint indoor orientation method according to claim 7 based on DHNN, which is characterized in that when DHNN network
According to asynchronous system work and when connection weight matrix W is symmetrical matrix, DHNN network convergence a to attractor；
When DHNN network adjusts network state according to the method for synchronization, and connection weight matrix W is nonnegative definite symmetrical matrix, DHNN network
Converge to a constant value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910146604.XA CN109803233A (en) | 2019-02-27 | 2019-02-27 | CSI fingerprint indoor orientation method based on DHNN |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910146604.XA CN109803233A (en) | 2019-02-27 | 2019-02-27 | CSI fingerprint indoor orientation method based on DHNN |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109803233A true CN109803233A (en) | 2019-05-24 |
Family
ID=66561360
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910146604.XA Pending CN109803233A (en) | 2019-02-27 | 2019-02-27 | CSI fingerprint indoor orientation method based on DHNN |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109803233A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113485261A (en) * | 2021-06-29 | 2021-10-08 | 西北师范大学 | CAEs-ACNN-based soft measurement modeling method |
| CN113938823A (en) * | 2021-10-14 | 2022-01-14 | 上海大学 | Wireless positioning device and method based on multi-band CSI cooperation |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105813194A (en) * | 2016-05-06 | 2016-07-27 | 西安电子科技大学昆山创新研究院 | Indoor positioning method based on fingerprint database secondary correction |
| US9445241B2 (en) * | 2014-07-17 | 2016-09-13 | Verizon Patent And Licensing Inc. | Location tracking for mobile device using Wi-Fi access point cluster |
| CN108650628A (en) * | 2018-03-24 | 2018-10-12 | 西安电子科技大学 | The indoor orientation method that ranging and fingerprint based on Wi-Fi network are combined |
| CN109151731A (en) * | 2018-06-14 | 2019-01-04 | 西北师范大学 | A kind of indoor orientation method based on CSI |
-
2019
- 2019-02-27 CN CN201910146604.XA patent/CN109803233A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9445241B2 (en) * | 2014-07-17 | 2016-09-13 | Verizon Patent And Licensing Inc. | Location tracking for mobile device using Wi-Fi access point cluster |
| CN105813194A (en) * | 2016-05-06 | 2016-07-27 | 西安电子科技大学昆山创新研究院 | Indoor positioning method based on fingerprint database secondary correction |
| CN108650628A (en) * | 2018-03-24 | 2018-10-12 | 西安电子科技大学 | The indoor orientation method that ranging and fingerprint based on Wi-Fi network are combined |
| CN109151731A (en) * | 2018-06-14 | 2019-01-04 | 西北师范大学 | A kind of indoor orientation method based on CSI |
Non-Patent Citations (1)
| Title |
|---|
| 徐耀群: "混沌神经网络研究及应用", 《中国博士学位论文全文数据库》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113485261A (en) * | 2021-06-29 | 2021-10-08 | 西北师范大学 | CAEs-ACNN-based soft measurement modeling method |
| CN113485261B (en) * | 2021-06-29 | 2022-06-28 | 西北师范大学 | CAEs-ACNN-based soft measurement modeling method |
| CN113938823A (en) * | 2021-10-14 | 2022-01-14 | 上海大学 | Wireless positioning device and method based on multi-band CSI cooperation |
| CN113938823B (en) * | 2021-10-14 | 2024-03-29 | 上海大学 | Wireless positioning device and method based on multi-band CSI cooperation |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10884112B2 (en) | Fingerprint positioning method and system in smart classroom | |
| Shangguan et al. | STPP: Spatial-temporal phase profiling-based method for relative RFID tag localization | |
| CN112073895A (en) | High-precision personnel positioning and tracking method based on CSI | |
| Moreira et al. | Wi-Fi fingerprinting in the real world-RTLS@ UM at the EvAAL competition | |
| Wang et al. | Dude, where's my card? RFID positioning that works with multipath and non-line of sight | |
| Jiang et al. | Communicating is crowdsourcing: Wi-Fi indoor localization with CSI-based speed estimation | |
| CN103561462B (en) | Indoor positioning system and method totally based on smart mobile terminal platform | |
| CN112153736A (en) | Personnel action identification and position estimation method based on channel state information | |
| CN109951807A (en) | Fusion RSS and CSI indoor orientation method based on WiFi signal | |
| CN105704652B (en) | Fingerprint base acquisition and optimization method in a kind of positioning of WLAN/ bluetooth | |
| CN107769828A (en) | A kind of the CSI MIMO indoor orientation methods and system of the extraction of feature based value | |
| CN105629196B (en) | Alignment system and correlation method based on computer vision and dynamic fingerprint | |
| CN103874200A (en) | Floor recognition method and system | |
| CN108650628A (en) | The indoor orientation method that ranging and fingerprint based on Wi-Fi network are combined | |
| CN109029429B (en) | WiFi and geomagnetic fingerprint based multi-classifier global dynamic fusion positioning method | |
| CN110475221A (en) | A kind of personnel's action recognition and location estimation method based on channel state information | |
| Majeed et al. | Indoor localization and radio map estimation using unsupervised manifold alignment with geometry perturbation | |
| CN109803233A (en) | CSI fingerprint indoor orientation method based on DHNN | |
| Chen et al. | SWIM: Speed-aware WiFi-based passive indoor localization for mobile ship environment | |
| WO2015065375A1 (en) | Signal space based navigation | |
| Xun et al. | Depthwise separable convolution based passive indoor localization using CSI fingerprint | |
| Deng et al. | WiFi positioning based on user orientation estimation and smartphone carrying position recognition | |
| Li et al. | Location-free CSI based activity recognition with angle difference of arrival | |
| Wang et al. | Through the wall detection and localization of autonomous mobile device in indoor scenario | |
| Khan et al. | Experimental testbed evaluation of cell level indoor localization algorithm using Wi-Fi and LoRa protocols |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190524 |