CN117295156A - Model training method, indoor positioning method and device and electronic equipment - Google Patents

Model training method, indoor positioning method and device and electronic equipment Download PDF

Info

Publication number
CN117295156A
CN117295156A CN202311267236.7A CN202311267236A CN117295156A CN 117295156 A CN117295156 A CN 117295156A CN 202311267236 A CN202311267236 A CN 202311267236A CN 117295156 A CN117295156 A CN 117295156A
Authority
CN
China
Prior art keywords
preset length
prediction model
position prediction
sample
track
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
Application number
CN202311267236.7A
Other languages
Chinese (zh)
Inventor
吴贞
胡鹏
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China Telecom Technology Innovation Center
China Telecom Corp Ltd
Original Assignee
China Telecom Technology Innovation Center
China Telecom Corp Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by China Telecom Technology Innovation Center, China Telecom Corp Ltd filed Critical China Telecom Technology Innovation Center
Priority to CN202311267236.7A priority Critical patent/CN117295156A/en
Publication of CN117295156A publication Critical patent/CN117295156A/en
Pending legal-status Critical Current

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W64/00—Locating users or terminals or network equipment for network management purposes, e.g. mobility management
    • 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
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
    • 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • 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
    • G06N3/048—Activation functions
    • 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
    • H04W4/30—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
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
    • H04W4/80—Services using short range communication, e.g. near-field communication [NFC], radio-frequency identification [RFID] or low energy communication

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Software Systems (AREA)
  • Evolutionary Computation (AREA)
  • Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Signal Processing (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Position Fixing By Use Of Radio Waves (AREA)

Abstract

The embodiment of the disclosure relates to a model training method, an indoor positioning method and device and electronic equipment, and relates to the technical field of positioning, wherein the method comprises the following steps: collecting sample signals indoors through terminal equipment, and dividing the sample signals into sample signals corresponding to tracks with preset lengths; inputting the sample signals corresponding to the preset length tracks into a position prediction model for feature extraction to obtain the predicted positions of the sample signals corresponding to the preset length tracks; and updating model parameters of the position prediction model based on the predicted position and the actual position of the sample signal corresponding to the track with the preset length to obtain a trained position prediction model for fingerprint positioning. The indoor positioning accuracy can be improved.

Description

Model training method, indoor positioning method and device and electronic equipment
Technical Field
The embodiment of the disclosure relates to the technical field of positioning, in particular to a model training method, an indoor positioning method, a model training device, an indoor positioning device and electronic equipment.
Background
With the rapid development of internet of things, location-based services gradually penetrate into various aspects of human production and life. At present, the global navigation satellite system can realize outdoor positioning more accurately, but in indoor scene positioning application, satellite signals are blocked by objects such as buildings and the like, and the satellite signals are seriously attenuated and even completely blocked.
Currently, for indoor positioning, a common technology is to use an uplink arrival time difference technology. However, this method often requires deployment of multiple base stations to work cooperatively to avoid the existence of a location hole, and thus the deployment cost is high. The fingerprint positioning technology using Bluetooth or WiFi signals has larger positioning errors due to the problems of equipment variability, spatial ambiguity and the like.
Disclosure of Invention
The present disclosure is directed to a model training method, an indoor positioning method, a model training device, an indoor positioning device, and an electronic apparatus, and further, to overcome the problem of poor positioning accuracy due to the limitations and drawbacks of the related art at least to a certain extent.
According to one aspect of the present disclosure, there is provided a model training method including: collecting sample signals indoors through terminal equipment, and dividing the sample signals into sample signals corresponding to tracks with preset lengths; inputting the sample signal corresponding to the track with the preset length into a position prediction model for feature extraction to obtain the predicted position of the sample signal corresponding to the track with the preset length at the last moment; and updating model parameters of the position prediction model based on the predicted position and the actual position of the sample signal corresponding to the track with the preset length to obtain a trained position prediction model for fingerprint positioning.
In an exemplary embodiment of the present disclosure, inputting the sample signal corresponding to the preset length track to a position prediction model for feature extraction, to obtain a predicted position of the sample signal corresponding to the preset length track includes: extracting features according to sample signals corresponding to the tracks with preset lengths to obtain output features corresponding to each moment; weighting scoring is carried out on the output characteristics at each moment to obtain the weight at each moment, and intermediate characteristics are obtained according to the weight at each moment and the output characteristics; and determining a predicted position according to the intermediate characteristic and the output characteristic.
In an exemplary embodiment of the present disclosure, the scoring the weights of the output features at each time to obtain weights at each time, and obtaining the intermediate features according to the weights at each time and the output features includes: carrying out logic processing on the weight of each moment and the feature dimension of the high-dimensional feature to obtain a logic processing result; and carrying out normalization processing on the logic processing result, and carrying out logic processing on the normalization result and the output characteristic to obtain the intermediate characteristic.
In an exemplary embodiment of the disclosure, the determining a predicted position from the intermediate feature and the output feature includes: fusing the intermediate features and the output features to obtain fusion features; performing convolution operation on the fusion features, and extracting the fusion features to target time to determine target features of the target time; and mapping the target characteristics to determine the predicted position.
In an exemplary embodiment of the present disclosure, updating the model parameters of the position prediction model based on the predicted position and the actual position of the sample signal corresponding to the track with the preset length to obtain a trained position prediction model for fingerprint positioning includes: determining an error between the predicted position and the actual position of the sample signal corresponding to the track with the preset length; and adjusting model parameters of the position prediction model based on the gradient, and determining the position prediction model with the minimum error as a trained position prediction model.
In an exemplary embodiment of the present disclosure, the method further comprises: determining a minimum sample signal from the sample signals; a normalized minimum value is determined from the minimum sample signal and the sample signal is normalized based on the normalized minimum value.
According to one aspect of the present disclosure, there is provided an indoor positioning method, including: acquiring signals indoors through terminal equipment, and dividing the signals into signals corresponding to tracks with preset lengths; inputting the signal corresponding to the track with the preset length into a trained position prediction model for feature extraction to obtain a predicted position of the signal corresponding to the track with the preset length; the trained position prediction model is obtained by training according to the model training method.
According to one aspect of the present disclosure, there is provided a model training apparatus comprising: the signal acquisition module is used for acquiring sample signals indoors through the terminal equipment and dividing the sample signals into sample signals corresponding to the track with the preset length; the position prediction module is used for inputting the sample signals corresponding to the preset length tracks into a position prediction model for feature extraction to obtain the predicted positions of the sample signals corresponding to the preset length tracks at the last moment; and the parameter adjustment module is used for updating the model parameters of the position prediction model based on the predicted position and the actual position of the sample signal corresponding to the track with the preset length to obtain a trained position prediction model for fingerprint positioning.
According to one aspect of the present disclosure, there is provided an indoor positioning device including: the signal acquisition module is used for acquiring signals indoors through the terminal equipment and dividing the signals into signals corresponding to the tracks with the preset length; the position prediction module is used for inputting the signals corresponding to the preset length tracks into the trained position prediction model for feature extraction to obtain the predicted positions of the signals corresponding to the preset length tracks; the trained position prediction model is obtained by training according to the model training method.
According to one aspect of the present disclosure, there is provided an electronic device including: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to perform the model training method of any one of the above or the indoor positioning method of any one of the above via execution of the executable instructions.
According to the technical scheme provided by the embodiment of the disclosure, on one hand, the sample signals corresponding to the indoor collected preset length tracks are subjected to feature extraction by using the position prediction model to obtain the predicted positions, model training is further performed based on the predicted positions and the actual positions, the accuracy of the model is improved, on the basis of the model, when fingerprint positioning is performed based on the position prediction model, the problem that large errors exist in fingerprint positioning due to equipment differences and space ambiguity is avoided, and the positioning accuracy is improved. On the other hand, the problem of higher cost caused by the need of cooperative work of a plurality of base stations in the related technology can be avoided, the positioning cost is reduced, the dependence on other base stations is avoided, and the application range can be increased.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the disclosure and together with the description, serve to explain the principles of the disclosure. It will be apparent to those of ordinary skill in the art that the drawings in the following description are merely examples of the disclosure and that other drawings may be derived from them without undue effort.
Fig. 1 schematically illustrates a flow chart of a model training method in an embodiment of the present disclosure.
Fig. 2 schematically illustrates a flowchart of acquiring a predicted position in an embodiment of the present disclosure.
Fig. 3 schematically illustrates a structural schematic of an LSTM model based on an attention mechanism in an embodiment of the present disclosure.
Fig. 4 schematically illustrates a flow diagram for determining intermediate features in an embodiment of the present disclosure.
Fig. 5 schematically illustrates an overall flow diagram of model training of an embodiment of the present disclosure.
Fig. 6 schematically illustrates a flow diagram of indoor positioning in an embodiment of the disclosure.
Fig. 7 schematically illustrates a block diagram of a model training apparatus in an embodiment of the present disclosure.
Fig. 8 schematically illustrates a block diagram of an indoor positioning device in an embodiment of the disclosure.
Fig. 9 schematically illustrates a block diagram view of an electronic device of an embodiment of the disclosure.
Detailed Description
Example embodiments will now be described more fully with reference to the accompanying drawings. However, the exemplary embodiments may be embodied in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of the example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a thorough understanding of embodiments of the present disclosure. One skilled in the relevant art will recognize, however, that the aspects of the disclosure may be practiced without one or more of the specific details, or with other methods, components, devices, steps, etc. In other instances, well-known technical solutions have not been shown or described in detail to avoid obscuring aspects of the present disclosure.
Furthermore, the drawings are merely schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus a repetitive description thereof will be omitted. Some of the block diagrams shown in the figures are functional entities and do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software or in one or more hardware modules or integrated circuits or in different networks and/or processor devices and/or microcontroller devices.
In an embodiment of the disclosure, a model training method is provided so as to be capable of realizing accurate indoor positioning.
Next, a model training method in the embodiment of the present disclosure will be specifically described with reference to fig. 1.
In step S110, network data to be processed is obtained, and spatial domain feature extraction is performed on the network data to be processed, so as to obtain corresponding spatial domain features.
In the embodiment of the disclosure, when the terminal equipment is indoor, a sample signal of the terminal equipment in the indoor can be collected. The indoor space can be any type of indoor environment, such as an office, a mall, a subway station, a teaching building, and the like. The rooms can be the rooms with the same plane or different planes, for example, the rooms with one floor or the rooms with multiple floors, and the rooms are determined according to actual requirements.
The collected sample signal may be an RSSI sample signal collected by the terminal device. RSSI (Received Signal Strength Indication, received sample signal strength indicator), which refers to a received sample signal strength indicator, is typically used to measure and indicate the strength of a wireless sample signal received by a receiver. In the embodiments of the present disclosure, the indoor sample signal may be collected through any type of network, which may be bluetooth or WiFi, and so on. In particular, a plurality of bluetooth beacons may be deployed indoors to collect a sample signal of a user indoors through bluetooth. In addition, a plurality of WAPs (Wireless Application Protocol ) may be deployed indoors to collect sample signals of users indoors through wireless network WiFi. The number of bluetooth beacons or WAPs can be set according to actual requirements. To improve the accuracy of indoor positioning, the number of bluetooth beacons or WAPs may be positively correlated with the size of the indoor area, e.g., the larger the indoor area, the greater the number of bluetooth beacons or WAPs; the smaller the area in the room, the fewer the number of bluetooth beacons and WAPs. The number of bluetooth beacons or WAPs deployed in each floor may be the same or different for the rooms of a multi-story building, and is not particularly limited herein.
Based on the type of network device deployed indoors, sample signals of the terminal device indoors can be acquired in real time through a network corresponding to the type of network device. That is, sample signal acquisition may be performed using bluetooth or WiFi.
The terminal device may be a terminal device carried by a user, may be an intelligent device with a positioning function, or may be a terminal device placed on another object, which is not specifically limited herein.
When the sample signal is acquired, the terminal equipment can be controlled to acquire and real-time position according to the preset moving track, so that the corresponding sample signal and real-time position are obtained. In order to improve accuracy and comprehensiveness, the preset moving track can cover all indoor maps.
After the sample signal is collected according to the moving track, in order to improve accuracy, normalization processing may be performed on the collected sample signal. The sample signal normalization process mainly comprises the following steps: traversing the sample signals acquired indoors to find out the minimum sample signals; and determining a normalized minimum value according to the minimum sample signal, and normalizing the acquired sample signal based on the normalized minimum value. Illustratively, the RSSI sample signal value acquired indoors is traversed to find the smallest sample signal, i.e., the smallest signal strength value p. In order to eliminate the influence caused by the special value, all normalized minimum values are set as c, and the normalized minimum values can be obtained by logically processing the minimum sample signal and a preset parameter value, wherein the preset parameter value can be any value, for example, can be 10, and the normalized minimum value c=p-10 based on the preset parameter value. Then, performing logic processing on all the collected sample signal RSSI values and normalized minimum values to normalize all the sample signals to obtain normalized sample signals, wherein the normalization process of the sample signals can be performed according to a formula (1):
Wherein R is i,t Representing the sample signal before normalization, R i ' ,t The normalized sample signal RSSI is represented, the index i represents the index of the sample signal access point, t represents the acquisition time, and e represents the natural logarithm.
After the normalization process, the normalized sample signal obtained by normalizing the collected sample signal can be expressed as
In order to reduce the time delay caused by the position prediction model in use, a section of track can be obtained by intercepting a sample signal acquired according to a moving track, and the shorter the length of the intercepted track is, the better the length of the intercepted track is. Based on this, in the embodiment of the present disclosure, the length of the truncated track may be set to a preset length. The preset length may be 4, or may be another smaller value. That is, the normalized sample signal may be divided into sample signals corresponding to a trace of a preset length, where the collected RSSI data may be intercepted using a window of a fixed length, and the sample signals corresponding to the intercepted trace of the preset length may be expressed as
It should be noted that, the track with a preset length may be any position in the collected sample signal as a starting point, for example, may be a track with a preset length including a starting position, may be a track with a preset length including an intermediate position, or may be a track with a preset length including an ending position, where the position of the track with a preset length is not specifically limited, as long as the length of the track is a preset length.
Next, in step S120, a sample signal corresponding to the preset length track is input to a position prediction model for feature extraction, so as to obtain a predicted position of the sample signal corresponding to the preset length track.
In the embodiment of the disclosure, after determining the sample signal corresponding to the track with the preset length, the intercepted sample signal corresponding to the track with the preset length may be divided into a training set, a verification set and a test set according to a certain proportion. Based on the method, the position prediction model can be trained by using sample signals corresponding to the tracks with the preset length in the training set until the trained position prediction model is obtained, so that fingerprint positioning is realized based on the trained position prediction model.
It should be noted that, the movement track of the user cannot be consistent with the training track, so that the movement track needs to be randomly disturbed in the training process of the position prediction model, so as to better help the network to perform gradient descent and enhance the generalization capability of the network model.
Based on the method, the position prediction model can be trained according to sample signals corresponding to randomly disturbed tracks with preset lengths. The location prediction model may be an LSTM algorithm model that may pass through an attention-based mechanism. In order to avoid the problem of larger errors caused by the problems of equipment variability, space ambiguity and the like when indoor positioning is realized by utilizing Bluetooth or WiFi in the related technology, the characteristic extraction of a track sample signal corresponding to a preset track length can be performed through an LSTM algorithm model based on an attention mechanism, so that the position of the track sample signal at the last moment is determined, and indoor fingerprint positioning of a user is realized through an LSTM algorithm model improved based on the attention mechanism by utilizing RSSI (received signal strength indication) of the Bluetooth or WiFi received by the indoor user as a basis.
In the embodiment of the disclosure, the LSTM algorithm model based on the attention mechanism can comprise the attention mechanism and the LSTM algorithm. Wherein the attention mechanism allows the neural network to focus on relevant parts when processing the input data. By introducing an attention mechanism, the neural network can automatically learn and selectively focus on important information in the input, improving the performance and generalization capability of the model. LSTM (Long Short Term Memory, long and short term memory), is a special recurrent neural network that can analyze inputs using time series with the ability to memorize long and short term information.
Fingerprint positioning is a positioning technology based on wireless signals, a fingerprint library is established through signal intensity values around a mobile phone, and then the position of a user is determined by comparing data in the fingerprint library. The positioning principle of the position fingerprinting is based on the principle of signal intensity decay. The signal is affected by various factors in the transmission process, so that the signal strength is changed, and the signal strength values at different positions are different. By collecting signal values of different positions to establish a fingerprint library, the most matched position can be found from the fingerprint library according to the signal intensity value of the current position of the user, so that the position of the user can be determined.
A flow chart for obtaining a predicted position is schematically shown in fig. 2, and referring to the process shown in fig. 2, the process of obtaining a predicted position may include the steps of:
in step S210, feature extraction is performed according to a sample signal corresponding to a track with a preset length, so as to obtain an output feature corresponding to each moment;
in step S220, weighting is performed on the output characteristics at each moment to obtain a weight at each moment, and an intermediate characteristic is obtained according to the weight at each moment and the output characteristics;
in step S230, a predicted position is determined based on the intermediate feature and the output feature.
In the embodiment of the disclosure, the structure of the LSTM algorithm model based on the attention mechanism may be shown in fig. 3, and mainly includes the following network layers: input layer, full connection layer, active layer, LSTM layer, attention layer, residual addition layer, convolution layer, full connection layer, output layer. Wherein: the output characteristics may be determined by the output layer, the full connectivity layer, the activation layer, and the LSTM layer. Weighting the output features at each time instant to obtain the weight of each time instant, and obtaining intermediate features according to the weight of each time instant and the output features may be performed by an attention layer. The determination of the predicted position from the intermediate feature and the output feature may be performed by a residual addition layer, a convolution layer, and an output layer.
A specific process of determining a predicted position according to the position prediction model will be described with reference to the above model structure, in which:
input layer: for receiving an input sample signal corresponding to a predetermined length trajectory.
And the full connection layer is used for carrying out full connection processing on the sample signal RSSI value corresponding to the input track with the preset length, mapping the characteristic of the sample signal RSSI value corresponding to the input track with the preset length to a high-dimensional space, and obtaining the high-dimensional characteristic of the sample signal corresponding to the track with the preset length.
The activation layer can be a nonlinear activation layer to process the high-dimensional features and output the processed high-dimensional features to the LSTM layer.
LSTM layer: and the dimension transformation is used for carrying out dimension transformation on the high-dimensional characteristics of the sample signals corresponding to the input track with the preset length to obtain the output characteristics corresponding to each moment. For example, the output characteristics of each of the times t, t+1, t+2, t+3, t+4, and the like may be output. The output features at each time may be multi-dimensional features resulting from feature extraction of the high-dimensional features at each time of the input LSTM layer. For each moment, the dimension of the output features can be increased, i.e. the dimension can be larger than the dimension of the input high-dimensional features, e.g. the output R N ,t]、[R N ,t+1]Etc. The dimension of the output features can be specifically limited according to the actual scene, and the dimensions of the output features are different from each other according to the actual scene. In a positioning scenario, the resulting spatial feature of a high-dimensional space may be mapped for each time instant RSSI value.
Attention layer: and the intermediate characteristics are obtained according to the weight of each moment and the output characteristics of the LSTM layer. The intermediate features may be intermediate features obtained by fusing the weight of each moment and the output features of the LSTM layer.
Illustratively, referring to the model structure shown in fig. 3, the attention layer includes a full connection layer, a normalization layer, and a matrix multiplication operation layer. The full connection layer can score the weight of the output characteristics of the LSTM layer to obtain the weight of each moment, so that attention of different moments is paid according to the weight scoring. And then, logically processing the weight of each time after the scoring and the output characteristics of the LATM layer through an operation layer of matrix multiplication to obtain intermediate characteristics.
In some embodiments, referring to the structure of the position prediction model described above, the obtaining of the intermediate features may comprise the steps of:
In step S410, performing logic processing on the weight of each moment and the feature dimension of the high-dimensional feature to obtain a logic processing result;
in step S420, the normalization processing is performed on the logic processing result, and the normalization result and the output feature are logically processed, so as to obtain the intermediate feature.
In the embodiment of the disclosure, the feature dimension of the high-dimensional feature refers to the feature dimension of the input matrix represented by the high-dimensional feature of the input LSTM layer, and may be represented by d. The weight at each time instant may be divided by the square root of the feature dimension of the high-dimensional feature to yield a logical processing result. For example, the weight of each moment can be +.And obtaining a logic processing result.
Further, the logic processing result can be input to a normalization layer for normalization to obtain a normalization result. The normalization layer may be a softmax layer, or may be other types of normalization layers. Next, the normalized result may be matrix multiplied with the output features of the LSTM layer to obtain intermediate features, and the intermediate features are taken as the output of the attention layer.
After the intermediate features are obtained, a predicted position may be determined based on the intermediate features and the output features. The method specifically comprises the following steps:
Fusing the intermediate features and the output features to obtain fusion features;
performing convolution operation on the fusion features, and extracting the fusion features to target time to determine target features of the target time;
and mapping the target characteristics to determine the predicted position.
The fusion can be understood as an addition operation, or of course, a weighted summation, or the like. And fusing the intermediate features and the output features to obtain fused features, wherein the fused features can be executed by a residual adding layer. And carrying out convolution operation on the fusion characteristic, extracting the fusion characteristic to a target moment so as to determine a target characteristic of the target moment, wherein the target characteristic can be executed by a convolution layer. The target time may be the last time of the sample signal corresponding to the track with the preset length, for example, may be the time t+4. The target feature may be a feature at a target time, and since the target feature is extracted from a fusion feature at a plurality of times, the target feature may include all features at a plurality of times before the target time.
And the residual error adding layer is connected with the attention layer and the LSTM layer and can combine the intermediate characteristics output by the attention layer and the output characteristics of the LSTM layer to obtain the fusion characteristics. Specifically, the intermediate feature output by the attention layer and the output feature of the LSTM layer may be added to obtain a fusion feature, and the fusion feature is input to the next layer.
And the convolution layer can carry out convolution operation on the fusion characteristics output by the residual addition layer, and obtain the characteristics at the last moment as output. The convolution layer can change the input channel to extract the characteristic values of four moments to one moment.
In some embodiments, the convolution layer has a convolution kernel, and the convolution kernel may multiply and add the characteristic values of the RSSI at four times included in the input fusion characteristic to implement the convolution process, so as to obtain the target characteristic at the last time in the track with the preset length.
After determining the target feature, the target feature may be mapped based on the full connection layer to map the feature space onto the location feature to obtain the predicted location. The predicted position may further be input to the output layer such that the output layer outputs the predicted position as an output result of the entire position prediction model.
Based on the position prediction model, the sample signals corresponding to the preset length tracks in the training set can be input into the position prediction model for feature extraction, and the predicted positions of the sample signals corresponding to the preset length tracks at the last moment are obtained. The predicted position of the last moment may be the abscissa and the ordinate of the last moment, for example the predicted position may be expressed as [ x ] t+4 ,y t+4 ]。
It should be noted that, for the sample signals corresponding to all the tracks with preset lengths in the training set, the corresponding predicted positions may be determined by the method in step S120, which is not described herein.
Next, with continued reference to fig. 1, in step S130, model parameters of the position prediction model are updated based on the predicted position and the actual position of the sample signal corresponding to the track of the preset length, so as to obtain a trained position prediction model for fingerprint positioning.
In the embodiment of the disclosure, after the predicted position is obtained, the model parameters of the position prediction model can be updated by combining the predicted position of the sample signal corresponding to the preset length track and the actual position of the sample signal corresponding to the preset length track, so as to realize a model training process, thereby obtaining a trained position prediction model, and the trained position prediction model can perform fingerprint positioning based on the obtained RSSI sample signal.
In some embodiments, the model training process may include the steps of:
determining an error between the predicted position and the actual position of the sample signal corresponding to the track with the preset length;
And adjusting model parameters of the position prediction model based on the gradient to perform model training, and taking the position prediction model with the minimum error as a trained position prediction model.
In the embodiment of the disclosure, the actual position may be a position coordinate of a target time recorded when the sample signal is acquired according to the movement track. The actual position may be an actual tag corresponding to the input data (sample signal corresponding to the preset length track). The error between the predicted position (predicted tag) and the actual position of the sample signal corresponding to the track of the preset length can be determined. The error here may be a positioning error. Further, the model training process can be realized by calculating the gradient so that the parameters of the position prediction model are updated according to the gradient. Model training may be performed using a gradient descent method.
Further, sample signals corresponding to the tracks with preset lengths corresponding to the verification set can be used for inputting the trained position prediction models, and the position prediction model with the minimum error is used as the trained position prediction model.
After the trained position prediction model is obtained, a divided test set can be loaded, and signals corresponding to the preset length tracks in the test set are input into the trained position prediction model to output corresponding preset positions, namely predicted position coordinates.
An overall flow chart of model training is schematically shown in fig. 5, and with reference to fig. 5, mainly comprises the following steps:
in step S510, the RSSI signals are collected along different movement tracks using the terminal device;
in step S520, the RSSI signal is preprocessed; intercepting RSSI data by using a window with a fixed length, and acquiring a position coordinate at the last moment as a corresponding label; dividing the data into a training set, a verification set and a test set according to a certain proportion after interception; executing a training process through the training set, and executing a testing process through the testing set;
in step S530, a training set is loaded, and the track data and the corresponding position labels are randomly disturbed;
in step S540, the randomly disturbed training set is input to the LSTM model based on the attention mechanism to obtain a predicted position;
in step S550, comparing the predicted position with the label corresponding to the input data to calculate an error, and turning to step S540 to continue to execute, so as to obtain a trained LSTM model based on the attention mechanism, so as to execute the test process from step S560 to step S590;
in step S560, using the verification set input model, selecting the model with the smallest positioning error as the trained LSTM model based on the attention mechanism;
In step S570, a test set is loaded;
in step S580, the test set is input to the trained attention-mechanism-based LSTM model;
in step S590, the corresponding predicted position coordinates are output.
In the embodiment of the disclosure, the accuracy of the model can be improved by training the position prediction model by combining the attention mechanism and the LSTM model. In addition, the user movement track data is randomly disturbed in the training process, so that model training is realized according to the randomly disturbed data, and the network can be helped to realize gradient descent better.
The scheme proposed by the embodiment of the present disclosure is verified in the following two-dimensional and three-dimensional scenes, respectively: 24 Bluetooth beacons are deployed in an office area (total area 195.84 square meters) of a certain floor of an office building, two different devices are used for acquiring Bluetooth RSSI signal values along a specific moving track, and corresponding position coordinates are recorded.
436 WAPs are deployed in a building comprising three floors with a total area of 9564 square meters. And acquiring the RSSI values of WiFi along different movement tracks in the same floor, and recording corresponding position coordinates. In addition, some movement trajectories can be moved between different floors to simulate real indoor movement scenarios.
After preprocessing the collected RSSI signals, data is input into the position prediction model proposed in the embodiments of the present disclosure for training, and then the position prediction model is tested. According to the execution result, the position prediction model can effectively realize indoor positioning in two-dimensional and three-dimensional scenes, and the positioning accuracy is improved compared with that of a KNN algorithm and a simple LSTM network structure.
After model training is completed, signals collected indoors can be predicted based on the trained position prediction model, so that corresponding predicted positions are obtained, and indoor positioning is achieved. A flow chart of indoor positioning is schematically shown in fig. 6, and referring to fig. 6, the method mainly comprises the following steps:
in step S610, signals are collected indoors through a terminal device, and the signals are divided into signals corresponding to tracks with preset lengths;
in step S620, inputting the signal corresponding to the preset length track to the trained position prediction model for feature extraction, so as to obtain a predicted position of the signal corresponding to the preset length track; the trained position prediction model is obtained by training according to the model training method.
In the embodiment of the disclosure, when the terminal equipment is indoor, the signal of the terminal equipment in the indoor can be collected according to the moving track of the terminal equipment. The indoor space can be any type of indoor environment, such as any type of indoor space of an office building, a mall, a subway station, a teaching building, and the like. The rooms can be the rooms with the same plane or different planes, for example, the rooms with one floor or the rooms with multiple floors, and the rooms are determined according to actual requirements. When the terminal equipment is a smart phone carried by a user, indoor RSSI signals can be collected according to the movement track of the user.
The signal collected in the room may still be an RSSI signal collected by the terminal device. In the embodiment of the disclosure, the indoor signal may be collected through any type of network, and the network may be bluetooth or WiFi, etc. In particular, a plurality of bluetooth beacons may be deployed indoors to collect signals of a user indoors through bluetooth. In addition, a plurality of WAPs (Wireless Application Protocol ) may be deployed indoors to collect signals indoors through wireless network WiFi. The number of bluetooth beacons or WAPs can be set according to actual requirements. To improve the accuracy of indoor positioning, the number of bluetooth beacons or WAPs may be positively correlated with the size of the indoor area, e.g., the larger the indoor area, the greater the number of bluetooth beacons or WAPs; the smaller the area in the room, the fewer the number of bluetooth beacons and WAPs. For example, 24 bluetooth beacons may be deployed in an office area (total area 195.84 square meters) in a floor of an office building. Alternatively, 436 WAPs are deployed in a building containing three floors with a total area of 9564 square meters.
The terminal device may collect the signal of the terminal device in the room in real time through a network corresponding to the type of the network device based on the type of the network device deployed in the room. That is, bluetooth or WiFi may be utilized for signal acquisition. The terminal equipment can be terminal equipment carried by a user, for example, a mobile phone can be held by the user; the intelligent device with a positioning function, such as a robot, can also be used; but also can be a terminal device placed on other objects, such as a mobile phone placed on a transmission device or a mobile vehicle, etc., which is not particularly limited herein.
After the signals are acquired according to the moving track, normalization processing can be performed on the acquired signals. The sample signal normalization process mainly comprises the following steps: traversing the sample signals acquired indoors to find out the minimum sample signal value; and determining a normalized minimum value according to the minimum sample signal value, and normalizing the acquired sample signal based on the normalized minimum value. The specific normalization process is the same as the normalization process in the training process, and will not be described here again.
Then, the signals acquired according to the moving track can be divided into signals corresponding to the track with the preset length, and the signals corresponding to the track with the preset length are input into the trained position prediction model. Performing feature extraction on signals corresponding to the track with the preset length according to the trained position prediction model to obtain output features corresponding to each moment; weighting scoring is carried out on the output characteristics at each moment to obtain the weight at each moment, and intermediate characteristics are obtained according to the weight at each moment and the output characteristics; and determining the predicted position of the signal corresponding to the track with the preset length at the target moment according to the intermediate characteristic and the output characteristic. Specifically, the intermediate feature and the output feature can be fused to obtain a fused feature; performing convolution operation on the fusion features, and extracting the fusion features to target time to determine target features at the target time; and mapping the target characteristics to determine the predicted position of the target moment. The target time may be the last time at which the signal corresponding to the track of the preset length is located. For the user, the trained position prediction model outputs coordinates of the indoor position where the user is currently located.
Based on the received RSSI signals, the RSSI signals are input into a trained position prediction model, and fingerprint positioning is carried out on the terminal equipment. The problem that a plurality of base stations need to be deployed to cooperatively work in the related technology is avoided, and the deployment cost is reduced. In addition, the problem that larger positioning errors can exist due to equipment difference and control ambiguity in the related technology is avoided, the fingerprint positioning accuracy is improved, positioning blind areas are avoided, and the positioning reliability is improved.
The disclosure also provides a model training device. Referring to fig. 7, the model training apparatus 700 may include:
the signal acquisition module 701 is configured to acquire a sample signal indoors through a terminal device, and divide the sample signal into sample signals corresponding to a track with a preset length;
the position prediction module 702 is configured to input a sample signal corresponding to the preset length track to a position prediction model for feature extraction, so as to obtain a predicted position of the sample signal corresponding to the preset length track at a last moment;
and the parameter adjustment module 703 is configured to update model parameters of the position prediction model based on the predicted position and an actual position of the sample signal corresponding to the track with the preset length, so as to obtain a trained position prediction model for fingerprint positioning.
In an exemplary embodiment of the present disclosure, inputting the sample signal corresponding to the preset length track to a position prediction model for feature extraction, to obtain a predicted position of the sample signal corresponding to the preset length track includes: extracting features according to sample signals corresponding to the tracks with preset lengths to obtain output features corresponding to each moment; weighting scoring is carried out on the output characteristics at each moment to obtain the weight at each moment, and intermediate characteristics are obtained according to the weight at each moment and the output characteristics; and determining a predicted position according to the intermediate characteristic and the output characteristic.
In an exemplary embodiment of the present disclosure, the scoring the weights of the output features at each time to obtain weights at each time, and obtaining the intermediate features according to the weights at each time and the output features includes: carrying out logic processing on the weight of each moment and the feature dimension of the high-dimensional feature to obtain a logic processing result; and carrying out normalization processing on the logic processing result, and carrying out logic processing on the normalization result and the output characteristic to obtain the intermediate characteristic.
In an exemplary embodiment of the disclosure, the determining a predicted position from the intermediate feature and the output feature includes: fusing the intermediate features and the output features to obtain fusion features; performing convolution operation on the fusion features, and extracting the fusion features to target time to determine target features of the target time; and mapping the target characteristics to determine the predicted position.
In an exemplary embodiment of the present disclosure, updating the model parameters of the position prediction model based on the predicted position and the actual position of the sample signal corresponding to the track with the preset length to obtain a trained position prediction model for fingerprint positioning includes: determining an error between the predicted position and the actual position of the sample signal corresponding to the track with the preset length; and adjusting model parameters of the position prediction model based on the gradient, and determining the position prediction model with the minimum error as a trained position prediction model.
In an exemplary embodiment of the present disclosure, the method further comprises: determining a minimum sample signal from the sample signals; a normalized minimum value is determined from the minimum sample signal and the sample signal is normalized based on the normalized minimum value.
The disclosure also provides an indoor positioning device. Referring to fig. 8, the indoor positioning device 800 may include:
the signal acquisition module 801 is configured to acquire a signal indoors through a terminal device, and divide the signal into signals corresponding to a track with a preset length;
the position prediction module 802 is configured to input a signal corresponding to the preset length track to a trained position prediction model for feature extraction, so as to obtain a predicted position of the signal corresponding to the preset length track; the trained position prediction model is obtained by training according to the model training method.
It should be noted that, the specific details of each module in the above model training device and the indoor positioning device have been described in detail in the corresponding methods, so that the details are not repeated here.
It should be noted that although in the above detailed description several modules or units of a device for action execution are mentioned, such a division is not mandatory. Indeed, the features and functionality of two or more modules or units described above may be embodied in one module or unit in accordance with embodiments of the present disclosure. Conversely, the features and functions of one module or unit described above may be further divided into a plurality of modules or units to be embodied.
Furthermore, although the steps of the methods in the present disclosure are depicted in a particular order in the drawings, this does not require or imply that the steps must be performed in that particular order or that all illustrated steps be performed in order to achieve desirable results. Additionally or alternatively, certain steps may be omitted, multiple steps combined into one step to perform, and/or one step decomposed into multiple steps to perform, etc.
In an exemplary embodiment of the present disclosure, an electronic device capable of implementing the above method is also provided.
Those skilled in the art will appreciate that the various aspects of the present disclosure may be implemented as a system, method, or program product. Accordingly, various aspects of the disclosure may be embodied in the following forms, namely: an entirely hardware embodiment, an entirely software embodiment (including firmware, micro-code, etc.) or an embodiment combining hardware and software aspects may be referred to herein as a "circuit," module "or" system.
An electronic device 900 according to such an embodiment of the present disclosure is described below with reference to fig. 9. The electronic device 900 shown in fig. 9 is merely an example and should not be construed to limit the functionality and scope of use of embodiments of the present disclosure in any way.
As shown in fig. 9, the electronic device 900 is embodied in the form of a general purpose computing device. Components of electronic device 900 may include, but are not limited to: the at least one processing unit 910, the at least one storage unit 920, a bus 930 connecting the different system components (including the storage unit 920 and the processing unit 910), and a display unit 940.
Wherein the storage unit stores program code that is executable by the processing unit 910 such that the processing unit 910 performs steps according to various exemplary embodiments of the present disclosure described in the above-described "exemplary methods" section of the present specification. For example, the processing unit 910 may perform the steps as shown in fig. 1.
The storage unit 920 may include readable media in the form of volatile storage units, such as Random Access Memory (RAM) 9201 and/or cache memory 9202, and may further include Read Only Memory (ROM) 9203.
The storage unit 920 may also include a program/utility 9204 having a set (at least one) of program modules 9205, such program modules 9205 include, but are not limited to: an operating system, one or more application programs, other program modules, and program data, each or some combination of which may include an implementation of a network environment.
The bus 930 may be one or more of several types of bus structures including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.
The electronic device 900 may also communicate with one or more external devices 1000 (e.g., keyboard, pointing device, bluetooth device, etc.), with one or more devices that enable a user to interact with the electronic device 900, and/or with any device (e.g., router, modem, etc.) that enables the electronic device 900 to communicate with one or more other computing devices. Such communication may occur through an input/output (I/O) interface 950. Also, electronic device 900 may communicate with one or more networks such as a Local Area Network (LAN), a Wide Area Network (WAN), and/or a public network, such as the Internet, through network adapter 960. As shown, the network adapter 960 communicates with other modules of the electronic device 900 over the bus 930. It should be appreciated that although not shown, other hardware and/or software modules may be used in connection with electronic device 900, including, but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, data backup storage systems, and the like.
From the above description of embodiments, those skilled in the art will readily appreciate that the example embodiments described herein may be implemented in software, or may be implemented in software in combination with the necessary hardware. Thus, the technical solution according to the embodiments of the present disclosure may be embodied in the form of a software product, which may be stored in a non-volatile storage medium (may be a CD-ROM, a U-disk, a mobile hard disk, etc.) or on a network, including several instructions to cause a computing device (may be a personal computer, a server, a terminal device, or an electronic device, etc.) to perform the method according to the embodiments of the present disclosure.
In an exemplary embodiment of the present disclosure, a computer-readable storage medium having stored thereon a program product capable of implementing the method described above in the present specification is also provided. In some possible implementations, various aspects of the disclosure may also be implemented in the form of a program product comprising program code for causing a terminal device to carry out the steps according to the various exemplary embodiments of the disclosure as described in the "exemplary methods" section of this specification, when the program product is run on the terminal device.
A program product for implementing the above-described method according to an embodiment of the present disclosure may employ a portable compact disc read-only memory (CD-ROM) and include program code, and may be run on a terminal device, such as a personal computer. However, the program product of the present disclosure is not limited thereto, and in this document, a readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.
The program product may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium can be, for example, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium would include the following: an electrical connection having one or more wires, a portable disk, a hard disk, random Access Memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
The computer readable signal medium may include a data signal propagated in baseband or as part of a carrier wave with readable program code embodied therein. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination of the foregoing. A readable signal medium may also be any readable medium that is not a readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.
Program code embodied on a readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.
Program code for carrying out operations of the present disclosure may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, C++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device, partly on a remote computing device, or entirely on the remote computing device or server. In the case of remote computing devices, the remote computing device may be connected to the user computing device through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computing device (e.g., connected via the Internet using an Internet service provider).
Furthermore, the above-described figures are only schematic illustrations of processes included in the method according to the exemplary embodiments of the present disclosure, and are not intended to be limiting. It will be readily appreciated that the processes shown in the above figures do not indicate or limit the temporal order of these processes. In addition, it is also readily understood that these processes may be performed synchronously or asynchronously, for example, among a plurality of modules.
Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any adaptations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.

Claims (10)

1. A method of model training, comprising:
collecting sample signals indoors through terminal equipment, and dividing the sample signals into sample signals corresponding to tracks with preset lengths;
inputting the sample signals corresponding to the preset length tracks into a position prediction model for feature extraction to obtain the predicted positions of the sample signals corresponding to the preset length tracks;
And updating model parameters of the position prediction model based on the predicted position and the actual position of the sample signal corresponding to the track with the preset length to obtain a trained position prediction model for fingerprint positioning.
2. The model training method according to claim 1, wherein the inputting the sample signal corresponding to the preset length track into a position prediction model for feature extraction, to obtain the predicted position of the sample signal corresponding to the preset length track, includes:
extracting features according to sample signals corresponding to the tracks with preset lengths to obtain output features corresponding to each moment;
weighting scoring is carried out on the output characteristics at each moment to obtain the weight at each moment, and intermediate characteristics are obtained according to the weight at each moment and the output characteristics;
and determining a predicted position according to the intermediate characteristic and the output characteristic.
3. The model training method according to claim 2, wherein the weighting scoring the output feature at each time to obtain the weight at each time, and obtaining the intermediate feature according to the weight at each time and the output feature comprises:
Carrying out logic processing on the weight of each moment and the feature dimension of the high-dimensional feature to obtain a logic processing result;
and carrying out normalization processing on the logic processing result, and carrying out logic processing on the normalization result and the output characteristic to obtain the intermediate characteristic.
4. The model training method of claim 2, wherein the determining a predicted position from the intermediate feature and the output feature comprises:
fusing the intermediate features and the output features to obtain fusion features;
performing convolution operation on the fusion features, and extracting the fusion features to target time to determine target features of the target time;
and mapping the target characteristics to determine the predicted position.
5. The model training method according to claim 2, wherein updating model parameters of the position prediction model based on the predicted position and an actual position of the sample signal corresponding to the track of the preset length to obtain a trained position prediction model for fingerprint positioning comprises:
determining an error between the predicted position and the actual position of the sample signal corresponding to the track with the preset length;
And adjusting model parameters of the position prediction model based on the gradient, and determining the position prediction model with the minimum error as a trained position prediction model.
6. The model training method of claim 1, wherein the method further comprises:
determining a minimum sample signal from the sample signals;
a normalized minimum value is determined from the minimum sample signal and the sample signal is normalized based on the normalized minimum value.
7. An indoor positioning method, comprising:
acquiring signals indoors through terminal equipment, and dividing the signals into signals corresponding to tracks with preset lengths;
inputting the signal corresponding to the track with the preset length into a trained position prediction model for feature extraction to obtain a predicted position of the signal corresponding to the track with the preset length; the trained position prediction model is obtained by training according to the model training method of any one of claims 1-6.
8. A model training device, comprising:
the signal acquisition module is used for acquiring sample signals indoors through the terminal equipment and dividing the sample signals into sample signals corresponding to the track with the preset length;
The position prediction module is used for inputting the sample signals corresponding to the preset length tracks into a position prediction model for feature extraction to obtain the predicted positions of the sample signals corresponding to the preset length tracks at the last moment;
and the parameter adjustment module is used for updating the model parameters of the position prediction model based on the predicted position and the actual position of the sample signal corresponding to the track with the preset length to obtain a trained position prediction model for fingerprint positioning.
9. An indoor positioning device, comprising:
the signal acquisition module is used for acquiring signals indoors through the terminal equipment and dividing the signals into signals corresponding to the tracks with the preset length;
the position prediction module is used for inputting the signals corresponding to the preset length tracks into the trained position prediction model for feature extraction to obtain the predicted positions of the signals corresponding to the preset length tracks; the trained position prediction model is obtained by training according to the model training method of any one of claims 1-6.
10. An electronic device, comprising:
a processor; and
A memory for storing executable instructions of the processor;
wherein the processor is configured to perform the model training method of any one of claims 1 to 6 or the indoor positioning method of claim 7 via execution of the executable instructions.
CN202311267236.7A 2023-09-27 2023-09-27 Model training method, indoor positioning method and device and electronic equipment Pending CN117295156A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311267236.7A CN117295156A (en) 2023-09-27 2023-09-27 Model training method, indoor positioning method and device and electronic equipment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311267236.7A CN117295156A (en) 2023-09-27 2023-09-27 Model training method, indoor positioning method and device and electronic equipment

Publications (1)

Publication Number Publication Date
CN117295156A true CN117295156A (en) 2023-12-26

Family

ID=89247661

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311267236.7A Pending CN117295156A (en) 2023-09-27 2023-09-27 Model training method, indoor positioning method and device and electronic equipment

Country Status (1)

Country Link
CN (1) CN117295156A (en)

Similar Documents

Publication Publication Date Title
US10930281B2 (en) Method, apparatus and system for testing intelligent voice device
CN108919177B (en) Positioning map construction method based on virtual information source estimation and track correction
CN108668249B (en) Indoor positioning method and device for mobile terminal
TWI608744B (en) Estimation devices and methods for estimating communicaiton quality of wireles network and method for installing meters thereof
CN111698774B (en) Indoor positioning method and device based on multi-source information fusion
Sun et al. Improving fingerprint indoor localization using convolutional neural networks
CN109379711B (en) positioning method
US11468349B2 (en) POI valuation method, apparatus, device and computer storage medium
CN106162559A (en) For estimating equipment and the method for position in a wireless communication system
CN110909804B (en) Method, device, server and storage medium for detecting abnormal data of base station
CN116528282B (en) Coverage scene recognition method, device, electronic equipment and readable storage medium
CN114449439A (en) Method and device for positioning underground pipe gallery space
CN103702411B (en) Method and device for accurately positioning under mobile environment
CN110320493B (en) Indoor positioning method, device, electronic equipment and computer storage medium
CN117295156A (en) Model training method, indoor positioning method and device and electronic equipment
CN108132458B (en) Indoor distance measuring method, device, equipment and storage medium
CN114187524B (en) Household air conditioner identification method, device, equipment, storage medium and product
CN113063420B (en) Indoor positioning method, device, electronic equipment and storage medium
CN113596757A (en) Rapid high-precision indoor fingerprint positioning method based on integrated width learning
CN108495262B (en) Sparse representation and matching positioning method for indoor space ubiquitous positioning signal fingerprint database
CN112866628B (en) Image collector name determining method, device, server and storage medium
CN111126120A (en) Urban area classification method, device, equipment and medium
CN114666734B (en) Positioning method and device for deep learning radio frequency signal fingerprint identification algorithm
CN111125272A (en) Regional feature acquisition method and device, computer equipment and medium
CN107525508A (en) Processing method, device, equipment and the computer-readable recording medium of indoor and outdoor state

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