CN110225460A - A kind of indoor orientation method and device based on deep neural network - Google Patents
A kind of indoor orientation method and device based on deep neural network Download PDFInfo
- Publication number
- CN110225460A CN110225460A CN201910488845.2A CN201910488845A CN110225460A CN 110225460 A CN110225460 A CN 110225460A CN 201910488845 A CN201910488845 A CN 201910488845A CN 110225460 A CN110225460 A CN 110225460A
- Authority
- CN
- China
- Prior art keywords
- neural network
- deep neural
- rss
- point
- obtains
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/02—Services making use of location information
- H04W4/021—Services related to particular areas, e.g. point of interest [POI] services, venue services or geofences
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/30—Services specially adapted for particular environments, situations or purposes
- H04W4/33—Services specially adapted for particular environments, situations or purposes for indoor environments, e.g. buildings
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Position Fixing By Use Of Radio Waves (AREA)
Abstract
The invention discloses a kind of indoor orientation methods based on deep neural network, the described method includes: obtaining Wi-Fi access point RSS reference data, the RSS reference data of acquisition is filtered, Wi-Fi access point RSS reference database is constructed based on the RSS reference data obtained after processing;Deep neural network is constructed, using the RSS reference database as training sample, training obtains deep neural network model;The RRS data of point to be determined are pre-processed, in the deep neural network model that input training obtains, obtain the position estimation coordinate of point to be determined.The present invention is without carrying special sensor, and for the indoor orientation method for much needing people's carry sensors, the present invention does not increase additional equipment, it is only necessary to which indoor positioning can be realized in smart phone, there is great convenience, is also convenient for promoting.
Description
Technical field
The present invention relates to indoor wireless positioning fields, and in particular to a kind of indoor orientation method based on deep neural network
And device.
Background technique
With the fast development of internet and wireless technology, the intelligent economic development neomorph based on internet changes
People's lives mode is become.Intelligent terminal has almost penetrated into each corner of people's life, incident various layers
Not poor application is also come into being out, wherein just including location technology.Currently, people can by global positioning system,
The satellite positioning tech such as Beidou realize precision up to one meter of accurate positioning outdoors.In highly developed Information Age
Under, demand of the mobile subscriber to location information has been not limited solely to outdoor coarse localization, accurate location requirement in real time with
Day all increasings.But blocking due to construction materials such as house sidings buildings, indoor satellite-signal can substantially decay, and lead to it
Serious misalignment is positioned, precision reduces.Also lack widely used, more mature, the higher indoor positioning skill of precision at present
Art.Therefore, the research of indoor positioning technologies increasingly becomes current research hotspot.
Existing indoor positioning technologies mainly include bluetooth (Bluetooth) technology, infrared ray (Infrared) technology,
RFID technique and Wi-Fi technology etc..In above-mentioned indoor positioning technologies, equipment volume needed for Bluetooth technology is small, but it has
Have that transmission range is short, in the disadvantage of complex environment stability inferior difference;Infrared technology needs between detector and target to be positioned
Between it is visual, be difficult to realize in complicated indoor environment;RFID technique positioning accuracy is high, but its anti-interference ability is poor.This
Outside, other than Wi-Fi technology, remaining indoor positioning technologies does not have perfect suitable basis to set in existing public place
It applies, and establishes these infrastructure and be also required to a large amount of time and economic input.As Wi-Fi signal covers in people's life
Each place, such as market, airport, office space etc. so that Wi-Fi signal be applied to indoor positioning have it is inborn excellent
Gesture becomes an important research field in indoor positioning technologies.
Indoor orientation method based on Wi-Fi, which is broadly divided into, to be based on receiving signal strength (Received Signal
Strength, RSS) and based on ranging model two major classes, wherein the location algorithm based on ranging model needs in advance to indoor letter
Road environment is estimated, channel model is established, due to the characteristic that indoor environment is complicated and changeable, the indoor positioning based on ranging model
Method can not be applicable in new environment, and based on the indoor positioning algorithms of RSS because it adapts to more complex indoor environment have it is prominent
Advantage out.
Traditional Wi-Fi indoor positioning algorithms specifically include that K nearest neighbor method (K-Nearest Neighbor, KNN) and add
It weighs K nearest neighbor method (Weight K-Nearest Neighbor, WKNN).The algorithm is fixed relative to other interiors based on Wi-Fi
Position algorithm, the complicated rate of calculating is low, and the speed of service is fast, it is easy to accomplish.But due to popularizing for Wi-Fi technology, the Wi- in large-scale place
Fi hot spot gradually increases, and reaches hundreds of quantity sometimes, the quantity of location reference point can also increase, therefore they determine
Position efficiency will receive very big influence.Simultaneously as WKNN algorithm cannot adaptively obtain real-time anchor point in positioning stage
Effective K value, needs manual setting, not can guarantee positioning accuracy.
Therefore, urgently need a kind of be able to solve that can still shorten when the RSS data quantity of Wi-Fi becomes huge
Positioning time guarantees the solution of location efficiency.
Summary of the invention
To solve the above-mentioned problems in the prior art, the present invention provides a kind of interiors based on deep neural network
Localization method.Under the premise of ensuring locating accuracy, deep learning algorithm is applied in indoor positioning technologies by the present invention, can
So that indoor locating system when handling the RSS data of magnanimity Wi-Fi, shortens positioning time.
For achieving the above object, the present invention the following technical schemes are provided:
A kind of indoor orientation method based on deep neural network, comprising the following steps:
S1: obtaining the RSS reference data of Wi-Fi access point, pre-process to the RSS reference data of acquisition, based on pre-
The RSS reference database of the RSS reference data building Wi-Fi access point obtained after processing;
S2: building deep neural network model, using the RSS reference database as training sample, training obtains depth
Spend neural network model;
S3: the RRS data of point to be determined are input to after pretreatment in the deep neural network model that training obtains, output
Obtain the position estimation coordinate of point to be determined.
Further, the step S1 specifically comprises the following steps:
Step S11, by it is to be positioned it is indoor transversely and be longitudinally divided into L region, wherein transverse direction is most greatly enhanced along it
Degree direction is divided into L1 region, and longitudinal direction is divided into L2 region along its maximum length direction, and L=L1 × L2 takes each
Position at the center in region as the region;If the region is non-centrosymmetry region, by the lateral sideline in the region and
The intersection point of the perpendicular bisector in longitudinal sideline is as its center.
Step S12 is uniformly arranged N number of reference point in each region, and obtains the location information of N number of reference point,
In, N is positive integer;
Step S13 obtains the RSS data of the corresponding M Wi-Fi access point received in N number of reference point locations place,
In, M is positive integer;
Step S14 implements filtering processing to the RSS data obtained in previous step to remove ambient noise, and based on filter
The RSS reference data obtained after wave processing constructs the Wi-Fi access point RSS reference database of N number of reference point.
Further, the deep neural network is using back-propagation algorithm to the connection between neuron in neural network
Weight is adjusted, and refers specifically to pass to the prediction to echo signal that the last layer obtains to response according to preceding, thus into
One step obtains its error with reference signal, according to obtained error amount, constantly regulate the connection weight of neural network, reversed to pass
The learning process for broadcasting algorithm can be divided into the following four stage:
(1) propagated forward: sample set { (x(1),y(1)),...,(x(m),y(m)) layer-by-layer to output layer through hidden layer from input layer
The process of propagation;
(2) error Back-Propagation: the desired output of neural network and the difference of reality output, i.e., error signal is by output layer through hidden
It is successively propagated containing layer to input layer, corrects the process of connection weight;
(3) memory training: propagated forward and error Back-Propagation alternately and repeatedly carry out the process of memory training;
(4) study convergence: the global error of network tends to the process of minimum;
Neuron in deep neural network between each layer is full connection relationship, without even between each neuron in layer
It connecing, y is denoted as to given target value, L () is denoted as objective function, it gives any group of input pattern (x (i), y (i)), i=1,
2 ..., N, loss function may be expressed as:
It sums and is averaged in N number of data, and enable:
Wherein, w is the weighted value in neural network neuron, and b is bias, and θ is other parameters;It is weight attenuation term;
It is calculated by feedforward, obtains target function value, then according to the criterion for minimizing loss function, declined using gradient
Algorithm, back-propagation gradient value, weights learning rule are as follows:
ε represents learning rate, the amplitude updated to define every subparameter.
Further, the step S3 further includes determining point to be determined affiliated area, and obtains point to be determined APiPosition
Information.
Further, in the step S2, the deep neural network uses adaptive regularized learning algorithm rate.
Further, the adaptive regularized learning algorithm rate uses Adam algorithm, according to loss function in neural network
The gradient single order moments estimation and second order moments estimation dynamic of each parameter adjust the learning rate of each parameter.
A kind of indoor positioning device based on deep neural network, comprising:
Position information acquisition module obtains the corresponding M Wi-Fi access point received in N number of reference point locations place
RSS data, wherein M is positive integer;The module is preferably intelligent terminal, which includes but is not limited to that intelligence communication is set
It is standby, specific such as smart phone, tablet computer.
RSS reference database constructs module, refers to the RSS of the Wi-Fi access point obtained in position information acquisition module
Data are pre-processed, and construct RSS reference database based on the RSS reference data obtained after pretreatment;
Deep neural network model building and position coordinates locating module, using the RSS reference database as training
Sample training obtains deep neural network model;The RRS data of point to be determined are input to what training obtained after pretreatment
In deep neural network model, output obtains the position estimation coordinate of point to be determined.
Further, the pretreatment is filtering removal ambient noise.
Further, the pretreatment is that low-pass filtering is utilized to remove ambient noise.Preferably, the low-pass filtering is adopted
It is realized with Butterworth filtering low-pass filter.
Compared with prior art, beneficial effects of the present invention:
1, without carrying special sensor, for the indoor orientation method for much needing people's carry sensors,
The present invention does not increase additional equipment, it is only necessary to which indoor positioning can be realized in smart phone or tablet computer, there is great convenience
Property, it is also convenient for promoting.
2, Butterworth filter is applied on RSS signal, has filtered out the ambient noise of high frequency.
3, machine learning is applied in indoor positioning technologies, establishes accurate mathematical model to predict the indoor position of people
It sets, effectively improves the precision of indoor positioning.
4, deep learning algorithm is applied in indoor positioning technologies, under conditions of ensuring location efficiency and accuracy rate,
It can make indoor locating system when with the RSS data of magnanimity Wi-Fi, shorten positioning time.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to do simply to introduce.It should be evident that the accompanying drawings in the following description is only this
Some embodiments of invention for those of ordinary skill in the art, can be with root under the premise of not making the creative labor
Other accompanying drawings are obtained according to these attached drawings.
Fig. 1 is system structure diagram of the invention;
Fig. 2 is the position distribution situation map of experimental site and Wi-Fi access point AP of the invention;
Fig. 3 is the structure chart of deep neural network of the invention;
Fig. 4 is error accumulation scatter chart of the invention.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention more comprehensible, with reference to the accompanying drawings and embodiments to this
Invention is described in further detail.It should be appreciated that the specific embodiments described herein are only used to explain the present invention,
And the scope of protection of the present invention is not limited.
Fig. 1 is the flow chart of indoor orientation method according to an embodiment of the present invention, as shown in Figure 1, in the present invention one
In embodiment, the indoor orientation method the present invention is based on deep neural network includes:
Step S1 obtains Wi-Fi access point RSS reference data, is filtered to the RSS reference data of acquisition, based on place
The RSS reference data building Wi-Fi access point RSS reference database obtained after reason.
Wherein, the Wi-Fi access point RSS reference data refers to that the Wi-Fi received at reference point locations connects
Access point RSS data.Reference point is either one or more, and Wi-Fi access point AP is also possible to one or more.At this
Invent in an embodiment, it is to be positioned it is indoor transversely and be longitudinally divided into L region, wherein transverse direction is along its maximum length
Direction is divided into L1 region, and longitudinal direction is divided into L2 region along its maximum length direction, and L=L1 × L2 takes each area
Position at the center in domain as the region;If the region is non-centrosymmetry region, by the lateral sideline in the region and
The intersection point of the perpendicular bisector in longitudinal sideline is as its center.
In the present embodiment, interior to be positioned is divided evenly as the total L region L1 × L2, each region is equably selected
The RSS data of N number of reference point acquisition Wi-Fi access point AP is taken, experimental site and its surrounding arrange M WiFi access point altogether
The coverage area of AP, the WiFi access point arranged around experimental site should cover experimental site.Experimental site and WiFi access point
Position distribution it is as shown in Figure 2.
In this embodiment, the step S1 further comprises:
L × N number of reference point is arranged altogether, and obtains the location information of each reference point by step S11;
Step S12 obtains the RSS data that the Wi-Fi access point received is corresponded at L × N number of reference point locations;
Step S13, using low-pass filter filter, preferably Butterworth filter is filtered, and removes ambient noise.Relatively
In ambient noise, the frequency for the reflection signal that people walks is more much smaller than the frequency of ambient noise, so the present embodiment uses
Butterworth filtering low-pass filter removes ambient noise.
In an embodiment of the present invention, affiliated Wi-Fi access point RSS reference database can be as shown in table 1, each Wi-Fi
The information of access point RSS reference data is indicated with RSSI:
1 Wi-Fi access point RSS reference database of table
In table, label represents zone number belonging to reference point, is denoted as l, in the present embodiment, l (l ∈ [Isosorbide-5-Nitrae 2]).lRPj
Indicate j-th of reference point in the region that zone number is l, j=1 ... N, N indicate the quantity of reference point, (xj,yj) indicate ginseng
Examination pointlRPjPosition coordinates, APiIndicate that i-th of Wi-Fi access point, i=1 ... M, M indicate the quantity of Wi-Fi access point,lRSSIijExpression zone number is in the region of l in j-th of reference pointlRPjWhat place received comes from i-th of Wi-Fi access point
APiRSS reference data, i.e. reference pointlRPjReference data information can indicate are as follows: [l, (xj,yj), RSSI1j...,
RSSIij,…,RSSIMj]。
Step S2 constructs deep neural network model, does training sample using database, training obtains deep neural network
Model.
The structure of deep neural network model in an of the invention case study on implementation as shown in figure 3, it has an input layer,
One output layer and multiple hidden layers.Deep neural network uses back-propagation algorithm (Backpropagation
Algorithm, BP) connection weight between neuron in neural network is adjusted.It specifically refers to be passed according to preceding to response
It broadcasts and passes to the prediction to echo signal that the last layer obtains, to further obtain the difference of itself and reference signal, this is poor
Different is exactly the mistake for needing backpropagation to go back.According to obtained error amount, so that it may constantly regulate the connection weight of neural network
Value.The learning process of back-propagation algorithm can be divided into the following four stage:
(1) propagated forward: sample set { (x(1),y(1)),...,(x(m),y(m)) layer-by-layer to output layer through hidden layer from input layer
The process of propagation.
(2) error Back-Propagation: the desired output of neural network and the difference of reality output, i.e., error signal is by output layer through hidden
It is successively propagated containing layer to input layer, corrects the process of connection weight.
(3) memory training: propagated forward and error Back-Propagation alternately and repeatedly carry out the process of memory training.
(4) study convergence: the global error of network tends to the process of minimum.
Neuron in BP neural network between each layer is full connection relationship, connectionless between each neuron in layer.
Y is denoted as to given target value, L () is denoted as objective function, it gives any group of input pattern (x (i), y (i)), i=1,
2 ..., N, loss function may be expressed as:
It sums and is averaged in N number of data, and enable:
Wherein, w is the weighted value in neural network neuron, and b is bias, and θ is other parameters;It is weight attenuation term, the purpose is to reduce the amplitude of weight, prevents over-fitting, weight attenuation parameter is used
Two relative importances in control formula.
It is calculated by feedforward, has obtained target function value, then according to the criterion for minimizing loss function, using under gradient
Algorithm, back-propagation gradient value, weights learning rule drop are as follows:
ε represents learning rate, the amplitude updated to define every subparameter, is learned in an example of the invention using adaptive adjustment
The method of habit rate.The algorithm specifically used is Adam (Adaptive Moment Estimation) algorithm, according to loss function
Single order moments estimation and second order moments estimation dynamic adjustment to the gradient of each parameter in neural network are directed to each parameter
Learning rate.The method that Adam is also based on gradient decline, but every time the Learning Step of iterative parameter have one it is determining
Range will not lead to very big Learning Step because of very big gradient, and the value of parameter is more stable, be suitable for large data sets and height
Dimension space.
Step S3 after pretreatment by the RSS data of point to be determined, is input to training when point to be determined is positioned
In good neural network model, the position estimation coordinate of point to be determined can be obtained.
In this embodiment, the step S3 further comprises:
Step S31 determines point to be determined APiAffiliated area, and obtain point to be determined APiLocation information;
In an embodiment of the present invention, point to be determined APiCoordinate can be expressed as (xi,yi), affiliated area is denoted as l.
Step S32, in point to be determined APiRSS of place's measurement from Wi-Fi access point;
In an embodiment of the present invention, the quantity of Wi-Fi access point AP is M, in point to be determined APiPlace's measurement comes from M
The RSS of a access point can be expressed as [RSSI '1i, RSSI '2i... ..., RSSI 'Mi]。
Step S33, by point to be determined APiThe Wi-Fi access point RSS at place is filtered, and obtains the RSS number of point to be determined
According to;
Wherein, described to filter the filter used, Wi-Fi access point RSS reference data is carried out with mentioned above
Filtering processing it is identical, to remove ambient noise.
Step S34, by deep neural network output to point to be determined PiPrediction coordinate be denoted as (xi’,yi’)。
Accuracy refers in all test datas that the correct ratio of labeling can indicate are as follows:
Average localization error indicates the distance between predicted position and physical location, can indicate are as follows:
It can be seen that positioning performance by the CDF position error cumulative distribution function of Fig. 4, there is 60% or more probabilistic localization
Error can be less than 2m, and 80% or so probabilistic localization error is less than 3m.
Claims (10)
1. a kind of indoor orientation method based on deep neural network, comprising the following steps:
S1: obtaining the RSS reference data of Wi-Fi access point, pre-process to the RSS reference data of acquisition, based on pretreatment
The RSS reference database of the RSS reference data building Wi-Fi access point obtained afterwards;
S2: building deep neural network model, using the RSS reference database as training sample, training obtains depth mind
Through network model;
S3: the RRS data of point to be determined are input to after pretreatment in the deep neural network model that training obtains, and output obtains
The position estimation coordinate of point to be determined.
2. as described in claim 1 based on the indoor orientation method of deep neural network, it is characterised in that: the step S1 tool
Body includes the following steps:
Step S11, by it is to be positioned it is indoor transversely and be longitudinally divided into L region, wherein transverse direction is along its maximum length side
To L1 region is divided into, longitudinal direction is divided into L2 region along its maximum length direction, and L=L1 × L2 takes each region
Center at position as the region;If the region is non-centrosymmetry region, by the lateral sideline in the region and longitudinal direction
The intersection point of the perpendicular bisector in sideline is as its center;
Step S12 is uniformly arranged N number of reference point in each region, and obtains the location information of N number of reference point, wherein N
For positive integer;
Step S13 obtains the RSS data of the corresponding M Wi-Fi access point received in N number of reference point locations place, wherein M
For positive integer;
Step S14 implements filtering processing to the RSS data obtained in previous step to remove ambient noise, and based at filtering
The RSS reference data obtained after reason constructs the Wi-Fi access point RSS reference database of N number of reference point.
3. as described in claim 1 based on the indoor orientation method of deep neural network, it is characterised in that: the depth nerve
Network is adjusted the connection weight between neuron in neural network using back-propagation algorithm, refers specifically to according to preceding Xiang Xiang
It should propagate and pass to the prediction to echo signal that the last layer obtains, to further obtain the error of itself and reference signal, root
According to obtained error amount, the connection weight of neural network is constantly regulate, the learning process of back-propagation algorithm can be divided into following
Four-stage:
(1) propagated forward: sample set { (x(1),y(1)),...,(x(m),y(m)) successively propagated through hidden layer to output layer from input layer
Process;
(2) error Back-Propagation: the desired output of neural network and the difference of reality output, i.e., error signal is by output layer through hidden layer
It is successively propagated to input layer, corrects the process of connection weight;
(3) memory training: propagated forward and error Back-Propagation alternately and repeatedly carry out the process of memory training;
(4) study convergence: the global error of network tends to the process of minimum;
Neuron in deep neural network between each layer is full connection relationship, connectionless between each neuron in layer, by y
Be denoted as given target value, L () is denoted as objective function, give any group of input pattern (x (i), y (i)), i=1,2 ...,
N, loss function may be expressed as:
It sums and is averaged in N number of data, and enable:
Wherein, w is the weighted value in neural network neuron, and b is bias, and θ is other parameters;It is power
Weight attenuation term;
It is calculated by feedforward, obtains target function value, then according to the criterion for minimizing loss function, declined using gradient and calculated
Method, back-propagation gradient value, weights learning rule are as follows:
ε represents learning rate, the amplitude updated to define every subparameter.
4. as described in claim 1 based on the indoor orientation method of deep neural network, it is characterised in that: the step S3 is also
Including determining point to be determined affiliated area, and obtain point to be determined APiLocation information.
5. as described in claim 1 based on the indoor orientation method of deep neural network, it is characterised in that: the step S2
In, the deep neural network uses adaptive regularized learning algorithm rate.
6. as claimed in claim 5 based on the indoor orientation method of deep neural network, it is characterised in that: the adaptive tune
Whole learning rate uses Adam algorithm, according to loss function to the gradient single order moments estimation and second order of each parameter in neural network
Moments estimation dynamic adjusts the learning rate of each parameter.
7. a kind of indoor positioning device based on deep neural network, comprising:
Position information acquisition module obtains the RSS number of the corresponding M Wi-Fi access point received in N number of reference point locations place
According to, wherein N, M are positive integer;
RSS reference database constructs module, the RSS reference data to the Wi-Fi access point obtained in position information acquisition module
It is pre-processed, RSS reference database is constructed based on the RSS reference data obtained after pretreatment;
Deep neural network model building and position coordinates locating module, using the RSS reference database as training sample
Training obtains deep neural network model;The RRS data of point to be determined are input to the depth that training obtains after pretreatment
In neural network model, output obtains the position estimation coordinate of point to be determined.
8. the indoor positioning device based on deep neural network as claimed in claim 7, it is characterised in that: the location information
Obtaining module is preferably intelligent terminal, including but not limited to intelligent communication device, specific such as smart phone, tablet computer.
9. the indoor positioning device based on deep neural network as claimed in claim 7, it is characterised in that: the pretreatment is
Filtering removal ambient noise.
10. the indoor positioning device based on deep neural network as claimed in claim 7, it is characterised in that: the pretreatment
To remove ambient noise using low-pass filtering, preferably, the low-pass filtering is real using Butterworth filtering low-pass filter
It is existing.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910488845.2A CN110225460B (en) | 2019-06-05 | 2019-06-05 | Indoor positioning method and device based on deep neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910488845.2A CN110225460B (en) | 2019-06-05 | 2019-06-05 | Indoor positioning method and device based on deep neural network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110225460A true CN110225460A (en) | 2019-09-10 |
CN110225460B CN110225460B (en) | 2021-03-23 |
Family
ID=67819493
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910488845.2A Active CN110225460B (en) | 2019-06-05 | 2019-06-05 | Indoor positioning method and device based on deep neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110225460B (en) |
Cited By (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110837806A (en) * | 2019-11-11 | 2020-02-25 | 辽宁科技学院 | Indoor positioning method based on deep convolutional memory network |
CN110856100A (en) * | 2019-10-21 | 2020-02-28 | 深圳数位传媒科技有限公司 | Method and device for terminal positioning and positioning model construction based on 5G signal |
CN110972056A (en) * | 2019-11-08 | 2020-04-07 | 宁波大学 | UWB indoor positioning method based on machine learning |
CN111031477A (en) * | 2019-12-25 | 2020-04-17 | 中山大学 | Millimeter wave indoor positioning method based on deep learning |
CN111123340A (en) * | 2020-02-10 | 2020-05-08 | 腾讯科技(深圳)有限公司 | Logistics distribution navigation method and system, near field positioning navigation device and storage medium |
CN111523667A (en) * | 2020-04-30 | 2020-08-11 | 天津大学 | Neural network-based RFID (radio frequency identification) positioning method |
CN111654910A (en) * | 2020-04-15 | 2020-09-11 | 深圳新贝奥科技有限公司 | Indoor positioning calculation method based on neural network |
CN111929674A (en) * | 2020-07-10 | 2020-11-13 | 西安电子科技大学 | Intelligent amplitude comparison angle measurement method based on neural network, storage medium and equipment |
WO2020259718A1 (en) * | 2019-09-20 | 2020-12-30 | 深圳前海微众银行股份有限公司 | Federated learning-based indoor positioning method, apparatus, and terminal device, and storage medium |
CN112188388A (en) * | 2020-08-25 | 2021-01-05 | 海盐南原电力工程有限责任公司 | Hybrid indoor positioning method based on machine learning |
CN113051976A (en) * | 2019-12-27 | 2021-06-29 | 广东博智林机器人有限公司 | Fingerprint positioning method and device, electronic equipment and storage medium |
CN113543305A (en) * | 2020-04-22 | 2021-10-22 | 维沃移动通信有限公司 | Positioning method, communication equipment and network equipment |
CN113596724A (en) * | 2021-08-05 | 2021-11-02 | 广东工业大学 | Indoor positioning method, device, equipment and medium based on transfer learning |
CN114514759A (en) * | 2019-10-04 | 2022-05-17 | 三星电子株式会社 | System and method for WiFi-based indoor positioning via unsupervised domain adaptation |
WO2023097634A1 (en) * | 2021-12-03 | 2023-06-08 | Oppo广东移动通信有限公司 | Positioning method, model training method, and device |
CN111123340B (en) * | 2020-02-10 | 2024-05-28 | 腾讯科技(深圳)有限公司 | Logistics distribution navigation method and system, near-field positioning navigation device and storage medium |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105101408A (en) * | 2015-07-23 | 2015-11-25 | 常熟理工学院 | Indoor positioning method based on distributed AP selection strategy |
CN106658422A (en) * | 2017-01-09 | 2017-05-10 | 深圳市智开科技有限公司 | Network side positioning method and network side positioning system for aiming at high-sparse WiFi data |
CN108303672A (en) * | 2017-12-26 | 2018-07-20 | 武汉创驰蓝天信息科技有限公司 | WLAN indoor positionings error correcting method based on location fingerprint and system |
US20180300608A1 (en) * | 2017-04-12 | 2018-10-18 | Yodlee, Inc. | Neural Networks for Information Extraction From Transaction Data |
CN109447244A (en) * | 2018-10-11 | 2019-03-08 | 中山大学 | A kind of advertisement recommended method of combination gating cycle unit neural network |
-
2019
- 2019-06-05 CN CN201910488845.2A patent/CN110225460B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105101408A (en) * | 2015-07-23 | 2015-11-25 | 常熟理工学院 | Indoor positioning method based on distributed AP selection strategy |
CN106658422A (en) * | 2017-01-09 | 2017-05-10 | 深圳市智开科技有限公司 | Network side positioning method and network side positioning system for aiming at high-sparse WiFi data |
US20180300608A1 (en) * | 2017-04-12 | 2018-10-18 | Yodlee, Inc. | Neural Networks for Information Extraction From Transaction Data |
CN108303672A (en) * | 2017-12-26 | 2018-07-20 | 武汉创驰蓝天信息科技有限公司 | WLAN indoor positionings error correcting method based on location fingerprint and system |
CN109447244A (en) * | 2018-10-11 | 2019-03-08 | 中山大学 | A kind of advertisement recommended method of combination gating cycle unit neural network |
Cited By (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2020259718A1 (en) * | 2019-09-20 | 2020-12-30 | 深圳前海微众银行股份有限公司 | Federated learning-based indoor positioning method, apparatus, and terminal device, and storage medium |
CN114514759A (en) * | 2019-10-04 | 2022-05-17 | 三星电子株式会社 | System and method for WiFi-based indoor positioning via unsupervised domain adaptation |
CN110856100A (en) * | 2019-10-21 | 2020-02-28 | 深圳数位传媒科技有限公司 | Method and device for terminal positioning and positioning model construction based on 5G signal |
CN110856100B (en) * | 2019-10-21 | 2021-04-23 | 深圳数位传媒科技有限公司 | Method and device for terminal positioning and positioning model construction based on 5G signal |
CN110972056A (en) * | 2019-11-08 | 2020-04-07 | 宁波大学 | UWB indoor positioning method based on machine learning |
CN110837806A (en) * | 2019-11-11 | 2020-02-25 | 辽宁科技学院 | Indoor positioning method based on deep convolutional memory network |
CN110837806B (en) * | 2019-11-11 | 2020-07-28 | 辽宁科技学院 | Indoor positioning method based on deep convolutional memory network |
CN111031477A (en) * | 2019-12-25 | 2020-04-17 | 中山大学 | Millimeter wave indoor positioning method based on deep learning |
CN113051976A (en) * | 2019-12-27 | 2021-06-29 | 广东博智林机器人有限公司 | Fingerprint positioning method and device, electronic equipment and storage medium |
CN111123340A (en) * | 2020-02-10 | 2020-05-08 | 腾讯科技(深圳)有限公司 | Logistics distribution navigation method and system, near field positioning navigation device and storage medium |
CN111123340B (en) * | 2020-02-10 | 2024-05-28 | 腾讯科技(深圳)有限公司 | Logistics distribution navigation method and system, near-field positioning navigation device and storage medium |
CN111654910A (en) * | 2020-04-15 | 2020-09-11 | 深圳新贝奥科技有限公司 | Indoor positioning calculation method based on neural network |
CN113543305A (en) * | 2020-04-22 | 2021-10-22 | 维沃移动通信有限公司 | Positioning method, communication equipment and network equipment |
CN111523667A (en) * | 2020-04-30 | 2020-08-11 | 天津大学 | Neural network-based RFID (radio frequency identification) positioning method |
CN111523667B (en) * | 2020-04-30 | 2023-06-27 | 天津大学 | RFID positioning method based on neural network |
CN111929674A (en) * | 2020-07-10 | 2020-11-13 | 西安电子科技大学 | Intelligent amplitude comparison angle measurement method based on neural network, storage medium and equipment |
CN112188388A (en) * | 2020-08-25 | 2021-01-05 | 海盐南原电力工程有限责任公司 | Hybrid indoor positioning method based on machine learning |
CN113596724A (en) * | 2021-08-05 | 2021-11-02 | 广东工业大学 | Indoor positioning method, device, equipment and medium based on transfer learning |
WO2023097634A1 (en) * | 2021-12-03 | 2023-06-08 | Oppo广东移动通信有限公司 | Positioning method, model training method, and device |
Also Published As
Publication number | Publication date |
---|---|
CN110225460B (en) | 2021-03-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110225460A (en) | A kind of indoor orientation method and device based on deep neural network | |
CN108696932B (en) | Outdoor fingerprint positioning method using CSI multipath and machine learning | |
CN103561463B (en) | A kind of RBF neural indoor orientation method based on sample clustering | |
CN113064117B (en) | Radiation source positioning method and device based on deep learning | |
CN103476118B (en) | A kind of WLAN indoor location fingerprint positioning method for monitoring in real time | |
CN105430664B (en) | It is a kind of to be fitted the method and apparatus that path loss is propagated in prediction based on classification | |
CN109212476B (en) | RFID indoor positioning algorithm based on DDPG | |
CN106851573A (en) | Joint weighting k nearest neighbor indoor orientation method based on log path loss model | |
Lee et al. | Path loss exponent prediction for outdoor millimeter wave channels through deep learning | |
CN107705359A (en) | The method and apparatus for obtaining wireless channel model | |
CN113962147A (en) | Construction system and method of wide area communication-oriented backscatter low-power-consumption communication network | |
CN111294921A (en) | RSSI wireless sensor network three-dimensional cooperative positioning method | |
CN112149294B (en) | Elastic weather grid design method | |
Yang | A new localization method based on improved particle swarm optimization for wireless sensor networks | |
Cui et al. | The RFID data clustering algorithm for improving indoor network positioning based on LANDMARC technology | |
CN108462545A (en) | A kind of ionosphere foF based on single receiving station2Parameter reconstructing method | |
CN108761455A (en) | Inverse synthetic aperture radar imaging resource-adaptive dispatching method in networking | |
Chen et al. | TDOA/AOA Hybrid Localization Based on Improved Dandelion Optimization Algorithm for Mobile Location Estimation Under NLOS Simulation Environment | |
Huan et al. | Indoor location fingerprinting algorithm based on path loss parameter estimation and bayesian inference | |
Shen et al. | Node localization based on improved PSO and mobile nodes for environmental monitoring WSNs | |
CN111505566A (en) | Ultrahigh frequency radio frequency signal DOA estimation method | |
Yang et al. | AI-enabled data-driven channel modeling for future communications | |
CN109581280A (en) | The adaptive tuning on-line method, system and device of terminal | |
Ren et al. | A novel high precision and low consumption indoor positioning algorithm for internet of things | |
Pattanayak | A genetically trained neural network for prediction of path loss in outdoor microcell |
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 |