CN113686335A - Method for performing accurate indoor positioning through IMU data by one-dimensional convolutional neural network - Google Patents
Method for performing accurate indoor positioning through IMU data by one-dimensional convolutional neural network Download PDFInfo
- Publication number
- CN113686335A CN113686335A CN202110647858.7A CN202110647858A CN113686335A CN 113686335 A CN113686335 A CN 113686335A CN 202110647858 A CN202110647858 A CN 202110647858A CN 113686335 A CN113686335 A CN 113686335A
- Authority
- CN
- China
- Prior art keywords
- neural network
- imu
- inertial navigation
- navigation data
- convolutional neural
- 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
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 61
- 238000000034 method Methods 0.000 title claims abstract description 50
- 238000006073 displacement reaction Methods 0.000 claims abstract description 42
- 238000013528 artificial neural network Methods 0.000 claims abstract description 41
- 238000007781 pre-processing Methods 0.000 claims abstract description 7
- 230000001133 acceleration Effects 0.000 claims description 16
- 230000008569 process Effects 0.000 claims description 12
- 210000002569 neuron Anatomy 0.000 claims description 9
- 238000004422 calculation algorithm Methods 0.000 claims description 8
- 230000004044 response Effects 0.000 claims description 7
- 230000003068 static effect Effects 0.000 claims description 3
- 230000003542 behavioural effect Effects 0.000 claims description 2
- 230000001902 propagating effect Effects 0.000 claims description 2
- 230000004807 localization Effects 0.000 claims 2
- 230000006870 function Effects 0.000 description 16
- 230000006399 behavior Effects 0.000 description 5
- 238000011478 gradient descent method Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 239000002131 composite material Substances 0.000 description 3
- 238000001914 filtration Methods 0.000 description 3
- 230000010354 integration Effects 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 238000009795 derivation Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000007477 logistic regression Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 230000005236 sound signal Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/20—Instruments for performing navigational calculations
- G01C21/206—Instruments for performing navigational calculations specially adapted for indoor navigation
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/10—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration
- G01C21/12—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration executed aboard the object being navigated; Dead reckoning
- G01C21/16—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration executed aboard the object being navigated; Dead reckoning by integrating acceleration or speed, i.e. inertial navigation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
- G06F3/03—Arrangements for converting the position or the displacement of a member into a coded form
- G06F3/041—Digitisers, e.g. for touch screens or touch pads, characterised by the transducing means
- G06F3/045—Digitisers, e.g. for touch screens or touch pads, characterised by the transducing means using resistive elements, e.g. a single continuous surface or two parallel surfaces put in contact
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/06—Digital input from, or digital output to, record carriers, e.g. RAID, emulated record carriers or networked record carriers
- G06F3/08—Digital input from, or digital output to, record carriers, e.g. RAID, emulated record carriers or networked record carriers from or to individual record carriers, e.g. punched card, memory card, integrated circuit [IC] card or smart card
Landscapes
- Engineering & Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Human Computer Interaction (AREA)
- Automation & Control Theory (AREA)
- Microelectronics & Electronic Packaging (AREA)
- Navigation (AREA)
Abstract
The invention discloses a method for performing accurate indoor positioning by using IMU data through a one-dimensional convolution neural network, which comprises the following steps: acquiring inertial navigation data acquired by an IMU, preprocessing the acquired IMU inertial navigation data, and constructing a reference database based on the preprocessed IMU inertial navigation data; constructing a neural network CNN model, and training by using the reference database as a training sample to obtain the neural network CNN model; IMU inertial navigation data of an object to be positioned are input into a neural network CNN model obtained through training after being preprocessed, displacement information of the object to be positioned is obtained through output, and position coordinates of the object to be positioned are estimated according to the output displacement information. The method for performing accurate indoor positioning by using IMU data through the one-dimensional convolutional neural network can accurately calculate displacement and perform positioning.
Description
Technical Field
The invention belongs to the technical field of indoor, and particularly relates to a method for accurately positioning indoors by using IMU data through a one-dimensional convolutional neural network.
Background
In recent years, with the popularization of mobile devices such as smart phones, tablet computers, and notebook computers, positioning applications based on these devices have become an important part of our daily lives. At present, most mobile devices mainly apply a Global Positioning System (GPS) or a Beidou positioning system to provide positioning service, but the two schemes can only effectively position a target outdoors, and the precision is low or even the mobile devices cannot work in a complex indoor environment. The reasons for this are mainly the following two points: (1) because a plurality of objects exist between the indoor access point AP and the mobile client, the wireless signals are reflected to a plurality of paths to be transmitted, which is called a multipath phenomenon; (2) GPS provides positioning accuracy of a few meters, more than enough for streets or city blocks in outdoor environments, but far from satisfactory for indoor environments that GPS cannot reach.
In order to effectively locate indoor objects, current indoor location technologies are mainly developed around Acoustic signals (acoustics), optical signals (Camera and Visible light), and electrical signals (UWB, Radar and RFID). The positioning technology based on the sound signal has the advantages that the precision is easily affected by environmental noise, and the positioning range is limited; the positioning technology based on optical signals excessively depends on the light intensity in the environment, cannot work in a non-line-of-sight environment, and is easy to cause the problem of target privacy disclosure; positioning technology based on electric signals mainly relies on expensive equipment and precise instruments in order to achieve high-precision positioning of targets. Therefore, the positioning method based on the above scheme cannot be generally suitable for our daily life.
Inertial navigation is a dead reckoning navigation method, which mainly measures angular velocity and linear acceleration through a gyroscope and an accelerometer which are installed on a motion carrier, and then calculates the position of the next point. The method has the advantages that the method is not influenced by external factors, the positioning precision is better in a short term, the defect is that the gyroscope has random drift errors, the noise is amplified in the 1-2 integration processes by adopting the traditional integration algorithm, the conditions of large noise interference or distortion and the like caused by filtering can occur, the positioning errors are increased along with time, and the errors of the speed and the position obtained by acceleration integration are large.
The convolutional neural network is a multi-layer supervised learning neural network, and a convolutional layer and a pool sampling layer of a hidden layer are core modules for realizing the function of extracting the characteristics of the convolutional neural network. The network model adopts a gradient descent method to minimize a loss function to reversely adjust weight parameters in the network layer by layer, and improves the accuracy of the network through frequent iterative training. The low hidden layer of the convolutional neural network is composed of convolutional layers and maximum pool sampling layers alternately, and the high layer is a hidden layer and a logistic regression classifier of a full-connection layer corresponding to the traditional multilayer perceptron. The convolutional neural network structure includes: convolutional layer, downsampling layer, full link layer. Each layer has a plurality of feature maps, each feature map extracting a feature of the input through a convolution filter, each feature map having a plurality of neurons.
Disclosure of Invention
1. Technical problem to be solved by the invention
The invention aims to solve the problems in the prior art and provides a method for performing accurate indoor positioning by using IMU data through a one-dimensional convolutional neural network.
2. Technical scheme
In order to achieve the purpose, the technical scheme provided by the invention is as follows:
a method for performing accurate indoor positioning by using IMU data through a one-dimensional convolutional neural network comprises the following steps:
s1: acquiring inertial navigation data acquired by an IMU, preprocessing the acquired IMU inertial navigation data, and constructing a reference database based on the preprocessed IMU inertial navigation data;
s2: constructing a neural network CNN model, and training by using the reference database as a training sample to obtain the neural network CNN model;
s3: IMU inertial navigation data of an object to be positioned are input into a neural network CNN model obtained through training after being preprocessed, displacement information of the object to be positioned is obtained through output, and position coordinates of the object to be positioned are estimated according to the output displacement information.
The preferable technical scheme is as follows:
in the method for performing accurate indoor positioning by using IMU data through a one-dimensional convolutional neural network as described above, in step S1, the inertial navigation data acquired by the IMU includes acceleration information, angular velocity information, and magnetic force information, and the reference database includes one or more sets of inertial navigation data acquired by the IMU.
In the method for performing accurate indoor positioning by using IMU data through the one-dimensional convolutional neural network as described above, in step S2, the training process of the neural network CNN model includes: the neural network CNN model adopts a forward propagation algorithm to adjust the connection weight between adjacent neurons in the neural network. The process of adjusting the connection weight between adjacent neurons in the neural network comprises the following steps: and predicting the displacement information transmitted to the last layer according to the forward response to obtain an error value between the displacement information and the actual displacement information of the object to be positioned, and continuously adjusting the connection weight of the neural network according to the obtained error value.
In the method for performing accurate indoor positioning by using IMU data through the one-dimensional convolutional neural network, the error value between the displacement information prediction obtained by transmitting the current response to the last layer and the actual displacement information of the object to be positioned reaches e-7And stage, stopping regulation.
The method for performing accurate indoor positioning by using IMU data through the one-dimensional convolutional neural network as described above, wherein the displacement information comprises the direction and the magnitude of the displacement.
In the method for performing accurate indoor positioning by using IMU data through the one-dimensional convolutional neural network as described above, before acquiring inertial navigation data acquired by the IMU in step S1, the method further includes: and acquiring the position coordinate information of the initial point of the object to be positioned.
The method for performing accurate indoor positioning by using IMU data through a one-dimensional convolutional neural network as described above, after preprocessing the obtained IMU inertial navigation data in step S1, further includes: and performing behavior state classification and gender classification on the preprocessed IMU inertial navigation data.
The method for performing accurate indoor positioning by using IMU data through the one-dimensional convolutional neural network comprises the following steps: obtaining an acceleration frequency f according to acceleration information a acquired by the IMU, and judging the walking state when the frequency is more than 0Hz and less than 2 Hz; when f is 0Hz, judging the state to be static; when f is more than or equal to 2Hz, the running state is judged; the method for gender classification comprises the following steps: according to the acceleration information a acquired by the IMU, when the a is less than 5.1m/s and 4.2m/s, the male is judged; when 3.4m/s < a <4.6m/s, it is judged as female.
The method for performing accurate indoor positioning by using IMU data through a one-dimensional convolutional neural network as described above, when estimating the position coordinate of an object to be positioned according to the output obtained displacement information, further includes: the behavioral state and gender of the subject to be located are estimated.
3. Advantageous effects
Compared with the prior art, the technical scheme provided by the invention has the following beneficial effects:
compared with the traditional integral calculation method, the method for accurately positioning indoors by using IMU data through the one-dimensional convolutional neural network can amplify noise in the 1-2 integral processes, and has the conditions of large noise interference or distortion and the like caused by filtering.
Drawings
FIG. 1 is a flow chart of a method of using IMU data for accurate indoor positioning via a one-dimensional convolutional neural network in accordance with the present invention;
FIG. 2 is a schematic diagram of the principle of the neural network CNN model outputting displacement information in the present invention;
FIG. 3 is a composite function of sample point correspondences in machine learning;
fig. 4 is a composite relationship of (a + b) × (b + 1);
FIG. 5 is a derivation relationship of adjacent nodes between different layers;
FIG. 6 is a corresponding network schema for back propagation;
Detailed Description
In order to facilitate an understanding of the invention, the invention will now be described more fully hereinafter with reference to the accompanying drawings, in which several embodiments of the invention are shown, but which may be embodied in many different forms and are not limited to the embodiments described herein, but rather are provided for the purpose of providing a more thorough disclosure of the invention.
It will be understood that when an element is referred to as being "secured to" another element, it can be directly on the other element or intervening elements may also be present; when an element is referred to as being "connected" to another element, it can be directly connected to the other element or intervening elements may also be present; the terms "vertical," "horizontal," "left," "right," and the like as used herein are for illustrative purposes only.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs; the terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention; as used herein, the term "and/or" includes any and all combinations of one or more of the associated listed items.
Referring to fig. 1, the present embodiment provides a method for performing accurate indoor positioning by using IMU data through a one-dimensional convolutional neural network, including the following steps:
s1: acquiring initial point position coordinate information of an object to be positioned;
s2: acquiring inertial navigation data acquired by an IMU, preprocessing the acquired IMU inertial navigation data, performing behavior state classification and gender classification on the preprocessed IMU inertial navigation data, and constructing a reference database based on the IMU inertial navigation data acquired after classification;
s3: constructing a neural network CNN model, and training by using the reference database as a training sample to obtain the neural network CNN modelThe process of training through the network CNN model comprises the following steps: the neural network CNN model adopts a forward propagation algorithm to adjust the connection weight between adjacent neurons in the neural network. The process of adjusting the connection weight between adjacent neurons in the neural network comprises the following steps: predicting displacement information obtained by transmitting forward response to the last layer to obtain an error value between the displacement information and actual displacement information of an object to be positioned, and continuously adjusting the connection weight of the neural network according to the obtained error value; the error value of the displacement information prediction obtained by transmitting the forward response to the last layer and the actual displacement information of the object to be positioned reaches e-7Stage, stopping regulation;
s4: the IMU inertial navigation data of the object to be positioned is input into a neural network CNN model obtained through training after being preprocessed, displacement information of the object to be positioned is obtained through output, and the position coordinate of the object to be positioned is estimated according to the displacement information obtained through output and in combination with the initial point position coordinate information of the object to be positioned, wherein the displacement information comprises the direction and the size of displacement.
In step S1, the behavior state classification method includes: obtaining an acceleration frequency f according to acceleration information a acquired by the IMU, and judging the walking state when the frequency is more than 0Hz and less than 2 Hz; when f is 0Hz, judging the state to be static; when f is more than or equal to 2Hz, the running state is judged;
the method for gender classification comprises the following steps: according to the acceleration information a acquired by the IMU, when the a is less than 5.1m/s and 4.2m/s, the male is judged; when 3.4m/s < a <4.6m/s, it is judged as female.
The steps are described in more detail below.
For step S2, the inertial navigation data collected by the IMU includes acceleration information, angular velocity information, and magnetic force information, and the reference database includes one or more sets of inertial navigation data collected by the IMU. For an IMU, the IMU typically includes a plurality of three-axis gyroscopes, three-axis accelerometers, and three-axis magnetometers to measure acceleration, angular velocity, and magnetomechanical information of an object to be positioned in indoor space. The IMU may be carried on an object to be positioned, or may be disposed in an indoor environment in which the object to be positioned is located.
For step S3, the neural network CNN model includes a convolutional layer, a downsampling layer, and a full link layer, each layer has a plurality of feature maps, each feature map extracts a feature of an input through a convolutional filter, and each feature map has a plurality of neurons. The goal of the training is to minimize the cost function (cost function) by adjusting each weight Wij. In this embodiment, the minimized cost function is effectively solved by a gradient descent method. The gradient descent method specifically includes: and giving an iteration point, solving a gradient vector of the point, searching in a certain step length by taking the negative gradient direction as a searching direction so as to determine the next iteration point, calculating a new gradient direction, and repeating until cost converges. The gradient is calculated specifically as follows:
suppose we represent the cost function as H (W)11，W12，...，Wij，...，Wmn) Then its gradient vector is equal to
Wherein eijRepresenting an orthogonal unit vector. Obtaining cost function H for each weight W by using Back propagationijPartial derivatives of (a).
Referring to fig. 3, fig. 3 shows a composite function corresponding to a sample point in machine learning.
The corresponding expression is as follows:
w in the above formulaijThe weights between two adjacent neurons are parameters to be learned, and are equivalent to parameters k and b to be solved in a straight line fitting y ═ k × x + b, but in machine learning, a cost function (cost function) is generally adopted, and the training goal is to adjust each weight WijTo minimize cost. The Cost function can also be regarded as the sum of all weights W to be calculatedijBeing a complex function of the arguments and being essentially non-convex, i.e. containing many local minima, the problem of minimizing the cost function can be solved efficiently using the gradient descent method.
The gradient descent method needs to give an iteration point, solve the gradient vector of the point, then search in a certain step length by taking the negative gradient direction as the search direction, so as to determine the next iteration point, then calculate a new gradient direction, and repeat until cost converges. The gradient is calculated specifically as follows:
suppose we represent the cost function as H (W)11，W12，...，Wij，...，Wmn) Then its gradient vector is equal to
Wherein eijRepresenting an orthogonal unit vector. The cost function H is obtained for each weight W by using a Back propagation algorithm (hereinafter, represented by BP algorithm)ijPartial derivatives of (a). The specific process comprises the following steps:
taking the partial derivative of (a + b) × (b +1) as an example, the complex relationship is shown in fig. 4.
In the figure, intermediate variables c and d are introduced, and in order to obtain the gradient of e when a is 2 and b is 1, the partial derivative relation of adjacent nodes between different layers needs to be obtained first, as shown in fig. 5 of the attached drawings.
Taking the formula shown in fig. 5 as an example, the processing is performed in units of layers with an initial value of 1 from the uppermost node e. For all children nodes of the next level of e, multiply 1 by the partial derivative value on the path of e to a node and "pile" the result in that child node. After the layer where e is located is propagated, each node of the second layer "stacks" values, and then for each node, all the "stacked" values in the node are summed to obtain the partial derivative of the vertex e to the node. Then, the nodes of the second layer are respectively used as initial vertexes, the initial values are set as partial derivatives of the vertexes e to the nodes, and the propagation process is repeated by taking the layer as a unit, so that the partial derivatives of the vertexes e to the nodes of each layer can be obtained. And the node c receives 1 x 2 sent by the e and stacks the received data, the node d receives 1 x 3 sent by the e and stacks the received data, and the total stacking amount of each node is calculated and continuously sent to the next layer until the second layer is finished. Node c sends 2 x 1 to a and stacks, node c sends 2 x 1 to b and stacks, node d sends 3 x 1 to b and stacks, and until the third layer is finished, node a stacks 2, node b stacks 2 x 1+3 x 1 to 5, namely, the partial derivative of vertex e to b is 5.
Back propagation specific calculation derivation process:
the corresponding network is shown in figure 6.
FIG. 6 shows a matrix corresponding to the network type as
Yout＝0.5
W1＝[w53 w54]＝[0.3 0.9]
First, with forward propagation computation:
then:
the same can get:
y2＝f(z2)＝f(w1*y1)＝f([0.801])＝[0.69]
the final loss can then be calculated as:
by further reducing this value by a directional propagation algorithm, according to the formula, one can obtain:
and (3) solving the partial derivative of the C to the w according to a chain rule:
the same can be obtained:
the above is the parameter partial derivative of the last layer, now also requiring w31,w32,w41,w42
The formulas are as follows:
then
The final result is:
continuously carrying out forward propagation algorithm according to the weight parameter, and continuing training until the error value of the displacement information prediction obtained by propagating forward response to the last layer and the actual displacement information of the object to be positioned reaches e-7And step two, stopping adjustment to obtain the trained neural network CNN model.
For step S4, when the position coordinate of the object to be positioned is estimated according to the output obtained displacement information, since the preprocessing is performed on the IMU inertial navigation data obtained in step S1, and the behavior state classification and the gender classification are also performed on the preprocessed IMU inertial navigation data, in this embodiment, after one preprocessed real-time IMU inertial navigation data is input to the neural network CNN model, it is not only able to estimate the position coordinate of the object to be positioned according to the output obtained displacement information, but also able to estimate the behavior state and the gender of the object to be positioned.
For the neural network CNN model in this embodiment, the input data is inertial navigation data acquired by the IMU, specifically including acceleration information, angular velocity information, and magnetic force information, and the displacement information of the object to be positioned is output through the neural network CNN model, and the estimation of the position coordinate of the object to be positioned is performed by knowing a real-time position change through an initial position coordinate of the object to be positioned and adding the displacement, thereby finally completing the estimation of the position coordinate of the object to be positioned. In the present application, the reason why the displacement information of the object to be positioned is output through the neural network CNN model rather than the estimated coordinate information of the object to be positioned is that: as shown in fig. 2, assuming that the initial position coordinate of the object to be positioned is point a, when the final coordinate of point E of the object to be positioned is estimated by using the neural network CNN model in this embodiment, the neural network CNN model only needs to obtain the displacement from point a to point E, so as to calculate the position of point E; when the estimated coordinate information of the object to be positioned is directly output by the neural network CNN model, the neural network CNN model must calculate B, C, D point coordinates between the points a and E, and then estimate the E point coordinates by the coordinates of the point D. In contrast, in the present application, by outputting the displacement information of the object to be positioned through the neural network CNN model, when the object to be positioned moves for a long time, the neural network CNN model can save a large amount of computation, thereby reducing the system load.
In summary, compared with the conventional integral calculation method which amplifies noise in 1-2 integral processes, the method for performing accurate indoor positioning by using IMU data through the one-dimensional convolutional neural network has the problems of large noise interference or distortion and the like caused by filtering.
The above-mentioned embodiments only express a certain implementation mode of the present invention, and the description thereof is specific and detailed, but not construed as limiting the scope of the present invention; it should be noted that, for those skilled in the art, without departing from the concept of the present invention, several variations and modifications can be made, which are within the protection scope of the present invention; therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (9)
1. The method for performing accurate indoor positioning by using IMU data through the one-dimensional convolutional neural network is characterized by comprising the following steps of:
s1: acquiring inertial navigation data acquired by an IMU, preprocessing the acquired IMU inertial navigation data, and constructing a reference database based on the preprocessed IMU inertial navigation data;
s2: constructing a neural network CNN model, and training by using the reference database as a training sample to obtain the neural network CNN model;
s3: IMU inertial navigation data of an object to be positioned are input into a neural network CNN model obtained through training after being preprocessed, displacement information of the object to be positioned is obtained through output, and position coordinates of the object to be positioned are estimated according to the output displacement information.
2. The method for performing indoor accurate localization using IMU data via one-dimensional convolutional neural network as claimed in claim 1, wherein in step S1, the inertial navigation data collected by the IMU comprises acceleration information, angular velocity information and magnetic force information, and the reference database comprises one or more sets of inertial navigation data collected by the IMU.
3. The method for performing accurate indoor positioning by IMU data through one-dimensional convolutional neural network as claimed in claim 2, wherein in step S2, the training process of neural network CNN model includes: the neural network CNN model adopts a forward propagation algorithm to adjust the connection weight between adjacent neurons in the neural network. The process of adjusting the connection weight between adjacent neurons in the neural network comprises the following steps: and predicting the displacement information transmitted to the last layer according to the forward response to obtain an error value between the displacement information and the actual displacement information of the object to be positioned, and continuously adjusting the connection weight of the neural network according to the obtained error value.
4. The method of claim 3 for accurate indoor positioning using IMU data via one-dimensional convolutional neural network, wherein the error value between the displacement information prediction obtained by propagating the current response to the last layer and the actual displacement information of the object to be positioned reaches e-7And stage, stopping regulation.
5. The method for accurate indoor positioning with IMU data via one-dimensional convolutional neural network of claim 1, wherein the displacement information includes direction and magnitude of displacement.
6. The method for performing accurate indoor positioning using IMU data via one-dimensional convolutional neural network as claimed in claim 1, wherein before acquiring inertial navigation data acquired by IMU in step S1, further comprising: and acquiring the position coordinate information of the initial point of the object to be positioned.
7. The method for performing accurate indoor positioning by IMU data through one-dimensional convolutional neural network as claimed in claim 2, wherein after preprocessing the IMU inertial navigation data obtained in step S1, further comprising: and performing behavior state classification and gender classification on the preprocessed IMU inertial navigation data.
8. The method for performing accurate indoor localization using IMU data via one-dimensional convolutional neural network as claimed in claim 7, wherein the method for performing behavior state classification is: obtaining an acceleration frequency f according to acceleration information a acquired by the IMU, and judging the walking state when the frequency is more than 0Hz and less than 2 Hz; when f is 0Hz, judging the state to be static; when f is more than or equal to 2Hz, the running state is judged; the method for gender classification comprises the following steps: according to the acceleration information a acquired by the IMU, when the a is less than 5.1m/s and 4.2m/s, the male is judged; when 3.4m/s < a <4.6m/s, it is judged as female.
9. The method of claim 7, wherein estimating the position coordinates of the object to be located based on the outputted displacement information, further comprises: the behavioral state and gender of the subject to be located are estimated.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110647858.7A CN113686335B (en) | 2021-06-10 | 2021-06-10 | Method for carrying out accurate indoor positioning by using IMU data through one-dimensional convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110647858.7A CN113686335B (en) | 2021-06-10 | 2021-06-10 | Method for carrying out accurate indoor positioning by using IMU data through one-dimensional convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113686335A true CN113686335A (en) | 2021-11-23 |
| CN113686335B CN113686335B (en) | 2024-05-24 |
Family
ID=78576529
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110647858.7A Active CN113686335B (en) | 2021-06-10 | 2021-06-10 | Method for carrying out accurate indoor positioning by using IMU data through one-dimensional convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113686335B (en) |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105628027A (en) * | 2016-02-19 | 2016-06-01 | 中国矿业大学 | Indoor environment precise real-time positioning method based on MEMS inertial device |
| CN105943013A (en) * | 2016-05-09 | 2016-09-21 | 安徽华米信息科技有限公司 | Heart rate detection method and device, and intelligent wearable device |
| CN106643715A (en) * | 2016-11-17 | 2017-05-10 | 天津大学 | Indoor inertial navigation method based on bp neural network improvement |
| CN107907127A (en) * | 2017-09-30 | 2018-04-13 | 天津大学 | A kind of step-size estimation method based on deep learning |
| CN107958221A (en) * | 2017-12-08 | 2018-04-24 | 北京理工大学 | A kind of human motion Approach for Gait Classification based on convolutional neural networks |
| US20190079534A1 (en) * | 2017-09-13 | 2019-03-14 | TuSimple | Neural network architecture system for deep odometry assisted by static scene optical flow |
| CN109579853A (en) * | 2019-01-24 | 2019-04-05 | 燕山大学 | Inertial navigation indoor orientation method based on BP neural network |
| WO2020008878A1 (en) * | 2018-07-02 | 2020-01-09 | ソニー株式会社 | Positioning device, positioning method, and program |
| CN110766985A (en) * | 2019-10-09 | 2020-02-07 | 天津大学 | Wearable motion sensing interactive teaching system and motion sensing method thereof |
| US20200355503A1 (en) * | 2018-01-10 | 2020-11-12 | Oxford University Innovation Limited | Determining the location of a mobile device |
-
2021
- 2021-06-10 CN CN202110647858.7A patent/CN113686335B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105628027A (en) * | 2016-02-19 | 2016-06-01 | 中国矿业大学 | Indoor environment precise real-time positioning method based on MEMS inertial device |
| CN105943013A (en) * | 2016-05-09 | 2016-09-21 | 安徽华米信息科技有限公司 | Heart rate detection method and device, and intelligent wearable device |
| CN106643715A (en) * | 2016-11-17 | 2017-05-10 | 天津大学 | Indoor inertial navigation method based on bp neural network improvement |
| US20190079534A1 (en) * | 2017-09-13 | 2019-03-14 | TuSimple | Neural network architecture system for deep odometry assisted by static scene optical flow |
| CN107907127A (en) * | 2017-09-30 | 2018-04-13 | 天津大学 | A kind of step-size estimation method based on deep learning |
| CN107958221A (en) * | 2017-12-08 | 2018-04-24 | 北京理工大学 | A kind of human motion Approach for Gait Classification based on convolutional neural networks |
| US20200355503A1 (en) * | 2018-01-10 | 2020-11-12 | Oxford University Innovation Limited | Determining the location of a mobile device |
| WO2020008878A1 (en) * | 2018-07-02 | 2020-01-09 | ソニー株式会社 | Positioning device, positioning method, and program |
| CN109579853A (en) * | 2019-01-24 | 2019-04-05 | 燕山大学 | Inertial navigation indoor orientation method based on BP neural network |
| CN110766985A (en) * | 2019-10-09 | 2020-02-07 | 天津大学 | Wearable motion sensing interactive teaching system and motion sensing method thereof |
Non-Patent Citations (4)
| Title |
|---|
| YARONG LUO;CHI GUO;JINTENG SU;WENFEI GUO;QUAN ZHANG: "Learning-Based Complex Motion Patterns Recognition for Pedestrian Dead Reckoning", vol. 21, no. 4, 31 December 2020 (2020-12-31) * |
| 刘宇;向高林;王伊冰;陈燕苹;吕玲;黄河明;: "一种改进的行人导航算法研究", 重庆邮电大学学报(自然科学版), no. 02, 15 April 2016 (2016-04-15) * |
| 张晓雪: "基于智能手机的穿戴式无缝导航定位***研究与实现_张晓雪", 中国优秀硕士学位论文全文数据库 (信息科技辑), no. 03, 15 March 2018 (2018-03-15) * |
| 黄河泽: "智能手机多模式误差补偿和导航方法研究", 中国优秀硕士学位论文全文数据库 (信息科技辑), no. 07, 15 July 2021 (2021-07-15) * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113686335B (en) | 2024-05-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111693047B (en) | Visual navigation method for micro unmanned aerial vehicle in high-dynamic scene | |
| Tian et al. | Pedestrian dead reckoning for MARG navigation using a smartphone | |
| CN110118560B (en) | Indoor positioning method based on LSTM and multi-sensor fusion | |
| CN113074739B (en) | UWB/INS fusion positioning method based on dynamic robust volume Kalman | |
| CN106912105B (en) | Three-dimensional positioning method based on PSO _ BP neural network | |
| Lv et al. | WSN localization technology based on hybrid GA-PSO-BP algorithm for indoor three-dimensional space | |
| CN107727095B (en) | 3D indoor positioning method based on spectral clustering and weighted back propagation neural network | |
| Real Ehrlich et al. | Indoor localization for pedestrians with real-time capability using multi-sensor smartphones | |
| CN109655786B (en) | Mobile ad hoc network cooperation relative positioning method and device | |
| CN110260885B (en) | Satellite/inertia/vision combined navigation system integrity evaluation method | |
| Chiang et al. | Magnetic field-based localization in factories using neural network with robotic sampling | |
| CN112729301A (en) | Indoor positioning method based on multi-source data fusion | |
| CN105674989A (en) | Indoor target motion track estimation method based on mobile phone built-in sensors | |
| Wang et al. | Modified zeroing neurodynamics models for range-based WSN localization from AOA and TDOA measurements | |
| Hasan et al. | Smart phone based sensor fusion by using Madgwick filter for 3D indoor navigation | |
| Deng et al. | Heading estimation fusing inertial sensors and landmarks for indoor navigation using a smartphone in the pocket | |
| CN110072192B (en) | WiFi indoor positioning method for smart phone | |
| Hassan | A performance model of pedestrian dead reckoning with activity-based location updates | |
| CN117516517A (en) | Passive fusion positioning method and system in indoor environment and electronic equipment | |
| Wang et al. | Recent advances in floor positioning based on smartphone | |
| CN108981689B (en) | UWB/INS combined navigation system based on DSP TMS320C6748 | |
| CN113686335B (en) | Method for carrying out accurate indoor positioning by using IMU data through one-dimensional convolutional neural network | |
| Huang et al. | An intelligent and autonomous MEMS IMU/GPS integration scheme for low cost land navigation applications | |
| CN108668254B (en) | WiFi signal characteristic area positioning method based on improved BP neural network | |
| Kang et al. | Smartphone indoor positioning system based on geomagnetic field |
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 |