CN115103438A - Wireless positioning method based on CIR peak value deviation and complex value deep neural network - Google Patents
Wireless positioning method based on CIR peak value deviation and complex value deep neural network Download PDFInfo
- Publication number
- CN115103438A CN115103438A CN202210527921.8A CN202210527921A CN115103438A CN 115103438 A CN115103438 A CN 115103438A CN 202210527921 A CN202210527921 A CN 202210527921A CN 115103438 A CN115103438 A CN 115103438A
- Authority
- CN
- China
- Prior art keywords
- cir
- csi
- wireless
- neural network
- target equipment
- 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
- 238000000034 method Methods 0.000 title claims abstract description 39
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 24
- 238000012549 training Methods 0.000 claims abstract description 20
- 238000004891 communication Methods 0.000 claims abstract description 8
- 238000001914 filtration Methods 0.000 claims abstract description 8
- 238000000342 Monte Carlo simulation Methods 0.000 claims abstract description 7
- 238000010606 normalization Methods 0.000 claims description 7
- 210000002569 neuron Anatomy 0.000 claims description 6
- 230000004044 response Effects 0.000 claims description 4
- 230000005540 biological transmission Effects 0.000 claims description 3
- 238000012544 monitoring process Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 description 12
- 230000006870 function Effects 0.000 description 9
- 230000008569 process Effects 0.000 description 7
- 238000012545 processing Methods 0.000 description 6
- 230000008901 benefit Effects 0.000 description 4
- 230000004927 fusion Effects 0.000 description 4
- 230000008054 signal transmission Effects 0.000 description 4
- 230000004913 activation Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000005259 measurement Methods 0.000 description 3
- 238000013508 migration Methods 0.000 description 3
- 230000005012 migration Effects 0.000 description 3
- 238000012358 sourcing Methods 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 239000007787 solid Substances 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 230000005358 geomagnetic field Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000007639 printing Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
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
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Position Fixing By Use Of Radio Waves (AREA)
Abstract
The invention relates to the technical field of wireless communication, in particular to a wireless positioning method based on CIR peak value deviation and a complex value deep neural network. The method comprises the following steps: s1, simulating the reflection and refraction of the WIFI signal by adopting a MonteCarlo method to generate a CSI data set; s2, converting the generated CSI into CIR through inverse Fourier transform; s3, normalizing the amplitude and the phase of the CIR; s4, using the normalized training set for cvDNN training, and obtaining a model, weight and bias of a Deep Neural Network (DNN) in an offline state; s5, detecting a data packet of the target device and the WIFI device by using a WIFI signal detector; s6, extracting CSI information of the target device from the data packet; s7, converting the CSI of the target equipment into CIR; s8, calculating the noise of the bottom plate, filtering the CIR, and marking the CIR peak position as an index; this technical scheme is used for solving current wireless location technique at the in-process of wireless location, uses the problem that has great limitation to and the lower problem of positioning accuracy.
Description
Technical Field
The invention relates to the technical field of wireless communication, in particular to a wireless positioning method based on CIR peak value deviation and a complex value deep neural network.
Background
The coverage and accuracy of a conventional satellite positioning system (e.g., GPS) are limited by the propagation environment. Outdoor and indoor wireless positioning has been widely used for outdoor and indoor navigation, underground mining, factory automation, wireless sensor networks, etc. based on ubiquitous wireless signals (e.g., bluetooth, radar, Radio Frequency Identification (RFID), Ultra Wideband (UWB), infrared, visible light, sound, geomagnetic field, etc.). Due to the popularity and widespread use of WiFi systems, there have been related inventions that provide WiFi-based positioning solutions. However, more studies are conducted to perform positioning based on a positioning protocol of ranging and Received Signal Strength (RSS), or to perform positioning by calculating euclidean distances of a measurement target device and a reference point through a location fingerprinting technique (finger printing system).
Chinese patent CN113015093A discloses "an indoor wireless positioning method based on three-dimensional depth residual error neural network", which is to construct an indoor wireless Channel State Information (CSI) map at an offline stage, extract CSI features by constructing a three-dimensional depth residual error neural network model, and implement online indoor wireless positioning.
Chinese patent CN113207089A discloses a location fingerprint positioning method based on CSI and crowd sourcing migration self calibration update, which utilizes a crowd sourcing system comprising crowd sourcing participants, a central server and users. The crowdsourcing participants send respective position information and CSI data to a central server, and position fingerprints with position labels are updated frequently; the central server reconstructs the relation between the fingerprint and the position through an algorithm based on deep migration learning, updates a position fingerprint database and a positioning model, and sends a current CSI measured value as network input training to the central server in an online test stage to realize position positioning.
Chinese patent CN112911500A discloses a high-precision indoor positioning method based on multi-source data fusion, which corrects antenna delay of a positioning base station and collects TOF message data by using UWB positioning and ranging data receiving equipment; performing NLOS (non line of sight) feature identification on the ranging data through channel estimation, and performing reconstruction processing by adopting an EKF (extended Kalman filter) algorithm; and fusing the processed TOF ranging data with the signal intensity to obtain primary fusion data, and obtaining secondary fusion data through a corresponding data fusion rule and an estimator so as to solve the optimal matching coordinate information.
Chinese patent CN112867021A discloses an improved TrAdaBoost-based migration learning indoor positioning method, which takes an original scene as a source domain, defines a new scene as a target domain, utilizes a One-Hot algorithm to encode CSI amplitude, uses a One-vs-Rest algorithm to perform cross matching on the amplitude, and utilizes the TrAdaBoost algorithm to adjust the weight values of the data of the source domain and the target domain, train a final multi-classifier, and estimate the position of a test point through confidence regression.
Multipath effects and ubiquitous noise in the wireless signal transmission process reduce the accuracy and stability of the wireless positioning method. In the prior art, in the wireless positioning process, relevant data (such as position fingerprints and scene models) of a position area needs to be acquired in advance, when an application scene changes, the data of the position area needs to be acquired repeatedly for training, the operation efficiency is low, and the technology has great limitations.
Disclosure of Invention
In view of the above technical deficiencies, an object of the present invention is to provide a wireless positioning method based on CIR peak offset and complex-valued deep neural network, so as to solve the problems of great limitation and low positioning accuracy in the wireless positioning process of the existing wireless positioning technology.
In order to achieve the purpose, the technical scheme adopted by the invention is as follows:
a wireless positioning method based on CIR peak value deviation and complex value deep neural network comprises the following steps:
s1, simulating the reflection and refraction of the WIFI signal by adopting a MonteCarlo method to generate a CSI data set;
s2, converting the generated CSI into CIR through inverse Fourier transform;
s3, normalizing the amplitude and the phase of the CIR;
s4, using the normalized training set for cvDNN training, and obtaining a model, weight and bias of a Deep Neural Network (DNN) in an offline state;
s5, detecting a data packet of the target device and the WIFI device by using a WIFI signal detector;
s6, extracting CSI information of the target device from the data packet;
s7, converting the CSI of the target equipment into CIR;
s8, calculating the noise of the bottom plate, filtering the CIR, and marking the CIR peak position as an index;
s9, extracting two side lobes of CIR peak envelope by CIR peak center point, setting data point lower than 1% CIR peak center point amplitude value as 0;
s10, carrying out amplitude and phase normalization on the filtered CIR, and shifting the center point of the peak value to an index position;
s11, taking the normalized CIR as the input of the trained cvDNN network, and predicting TOA of the target equipment through DNN;
and S12, acquiring the position of the target equipment based on the arrival time difference of the unknown target equipment to the plurality of detectors, and realizing the positioning of the target equipment.
Further, in step S2, the inverse fourier transform function for converting the CSI into the CIR is:
CIR＝IFFT(CFR*W)
where W is the sampling width, CFR is the channel frequency response of the receiving end, and IFFT represents the inverse fourier transform.
Further, in step S3, the amplitude and phase normalization function of CIR is:
CIR_norm[N]＝CIR[N]/max(CIR[N])
in the formula, CIR _ norm [ N ] is the normalized value of the Nth CIR point, CIR [ N ] represents the value of the Nth CIR point, and max (CIR [ N ]) represents the maximum value of the whole CIR sequence.
Further, in step S4, the relationship between the weight and the offset is:
y＝f(x；W，b)＝f (1) (f (2) (f (3) (…f (L) (x))))
where L, W and b represent the number of layers, weights and offset values of the neural network, respectively.
Further, in step S8, the noise floor threshold is calculated as:
in the formula, Anoise represents a threshold of the noise floor, and Asignal _ max represents a maximum value of CIR.
Further, in step S4, the network structure of the cvDNN includes 6 hidden layers, and the 6 hidden layers are composed of complex-valued layers and real-valued layers, where the number of the complex-valued layers is 3, and each layer has 256 neurons.
Further, the operation of the wireless location system and the WASP system includes:
the wireless access point comprises a plurality of antennas for transmitting and receiving WIFI signals and realizes the transmission and the reception of the WIFI signals;
the wireless detectors are used for monitoring the WiFi network flow and equipment communication;
the target equipment sends a WiFi data packet to the AP;
and the wireless positioning terminal calculates the TDOA of the target equipment so as to solve the position of the target equipment.
The technical effect achieved by the technical scheme is as follows:
compared with the prior art, the method has the advantages that the required communication information is directly extracted from an application scene, a Saleh-Valenzuela channel model (hereinafter referred to as an S-V model) is simulated by adopting a Monte Carlo method, millions of CSI data sets are generated, the CSI is converted into CIRs for training a cvDNN model, and key parameters (weight and bias of a hidden layer) of the DNN model are obtained, so that the scene model and field data do not need to be collected in advance, the cost is reduced, and the operation efficiency is improved; the noise floor threshold value and peak value offset technology can reduce the influence of multipath effect and noise on DNN model training in the wireless signal transmission process; and a DNN model based on a complex value domain respectively processes a real part and an imaginary part of a signal in a hidden layer, so that the accuracy of TOA prediction is improved. Therefore, the complexity of a wireless positioning system is reduced, the method can be widely suitable for most indoor and outdoor wireless positioning scenes, and can also effectively improve the accuracy of wireless positioning, and secondly, compared with the MUSIC and ESPRIT methods which are commonly used for wireless positioning at present, the CIR filtering and peak value drifting method provided by the invention has the advantage that the positioning accuracy is improved to 0.5 meter from 2 meters, so that the positioning accuracy is higher.
Drawings
FIG. 1 is a flow chart of wireless positioning based on CIR peak shift and complex-valued deep neural network;
FIG. 2 is a schematic diagram of a wireless positioning system;
FIG. 3 is a schematic diagram of a fully-connected cvDNN structure;
FIG. 4 is a flow chart of the operation of a complex-valued deep neural network;
fig. 5 shows the CIR peak shift and the effect of complex deep neural network wireless positioning.
Detailed Description
The following is further detailed by way of specific embodiments:
as shown in fig. 1, a wireless positioning method based on CIR peak offset and complex-valued deep neural network includes the following steps:
s1, simulating the reflection and refraction of the WIFI signal by adopting a MonteCarlo method to generate a CSI data set;
s2, converting the generated CSI into CIR through inverse Fourier transform;
s3, normalizing the amplitude and the phase of the CIR;
s4, using the normalized training set for cvDNN training, and obtaining a model, weight and bias of a Deep Neural Network (DNN) in an offline state;
s5, detecting a data packet of the target device and the WIFI device by using a WIFI signal detector;
s6, extracting CSI information of the target device from the data packet;
s7, converting the CSI of the target equipment into CIR;
s8, calculating the noise of the bottom plate, filtering the CIR, and marking the CIR peak position as an index;
s9, extracting two side lobes of CIR peak envelope by CIR peak center point, setting data point lower than 1% CIR peak center point amplitude value as 0;
s10, carrying out amplitude and phase normalization on the filtered CIR, and shifting the center point of the peak value to an index position;
s11, taking the normalized CIR as the input of the trained cvDNN network, and predicting TOA of the target equipment through DNN;
and S12, acquiring the position of the target equipment based on the arrival time difference of the unknown target equipment to the plurality of detectors, and realizing the positioning of the target equipment.
Fig. 2 illustrates a schematic diagram of a wireless location system based on the WASP system. The wireless positioning system runs in the WASP system and mainly comprises a WIFI signal detector, a wireless Access Point (AP) and a wireless positioning processing terminal.
The WASP system is based on 802.11a/g/n protocol, adopts Orthogonal Frequency Division Multiplexing (OFDM) to operate in 5.8GHz ISM band, uses 125MHz bandwidth to realize TOA-based distance measurement, and can be divided into 8 sub-bands, and the maximum bandwidth of each sub-band is 18.75 MHz. The MAC layer is based on a Time Division Multiple Access (TDMA) protocol and its applications, supporting high TOA measurement rates.
The wireless Access Point (AP) comprises a plurality of antennas for transmitting and receiving WIFI signals, and realizes the transmission and reception of the WIFI signals;
the WIFI signal detector is installed at a known position and used for monitoring the WiFi network flow and equipment communication. The WIFI signal detector cannot interfere with a WIFI system of the existing standard, and has the function of a WIFI access point. The WIFI signal detector estimates the clock offset by measuring the timestamp of the access point, and the clock synchronization problem of various devices is solved.
In the WASP system, when target equipment needing positioning is accessed to WiFi, the target equipment sends a WiFi data packet to an AP, a signal detector measures a timestamp, the measured data is sent to a wireless positioning processing terminal, and TDOA of the target equipment is calculated, so that the position of the target equipment is calculated.
Further, in order to simulate the reflection and refraction processes of the WIFI signal, a Saleh-Valencuela channel model (hereinafter referred to as an S-V model) is simulated by adopting a MonteCarlo method, and million CSI data sets are generated and used for training a cvDNN model, so that parameters such as weight, bias and the like of the deep neural network model are obtained.
For a CSI data set generated based on an S-V model, the generated CSI is converted into a Channel Impulse Response (CIR) by inverse fourier transform, that is:
CIR＝IFFT(CFR*W)
where W is the sampling width, CFR is the channel frequency response of the receiving end, and IFFT represents the inverse fourier transform. In the present invention, W is greater than xxx.
In order to improve the accuracy and convergence rate of DNN prediction, the amplitude and phase of the CIR are normalized, wherein the normalization mode is to obtain the ratio of each CIR value to the CIR maximum value, and the specific expression is as follows:
CIR_norm[N]＝CIR[N]/max(CIR[N])
in the formula, CIR _ norm [ N ] is the normalized value of the Nth CIR point, CIR [ N ] represents the value of the Nth CIR point, and max (CIR [ N ]) represents the maximum value of the whole CIR sequence.
The normalized CIR amplitude and phase are used for training cvDNN in full-connection form to obtain DNN model parameters (weight and bias of neural network) with high accuracy and good convergence, and the specific mode is as follows:
first, a fully-connected cvDNN network architecture includes an input layer, an hidden layer, and an output layer. In this network configuration, the input layer is CSI information (the number of CSI for each input layer is 10000), and the output layer is the TOA of the target device. There are M hidden layers, each with i neural units. The output y of the neural network is the nonlinear transformation of the input x in a cascade form, and the system structure is shown in fig. 3. The mathematical expression is:
y＝f(x；W，b)＝f (1) (f (2) (f (3) (…f (L) (x))))
wherein, L, W and b represent the number of layers, weights and offset values of the neural network, respectively.
In the present invention, the cvDNN contains 6 hidden layers, including 3 complex-valued layers, and others are real-valued layers, each layer having 256 neurons.
In the WASP system based on the 802.11a/g/n protocol, the complex value h (n) of CSI can be obtained from the target device and the wireless signal detector to the nth subcarrier of the access point, and is expressed as h ═ x + iy; where x is the real part and y is the imaginary part. The constructed weight matrix of the hidden layer of the DNN (deep neural network) is: w ═ a + iB, where a and B represent the real and imaginary parts, respectively. For forward propagation of DNN:
W*h＝(A*x-B*y)+i(Bx+Ay)
the real part R (W × h) and imaginary part S (W × h) of CSI may be expressed as a matrix:
in order to improve the fitting performance of cvDNN, the bias is constructed as follows: and b is m + in, wherein the dimension of the weight W and the offset b is the same as the complex value CSI. Thus, the input and output relationships for the n-th layer of neurons are:
in fact, the operation of the complex-valued hidden layer is as shown in FIG. 4. Where CR and CI represent the real and imaginary parts of CSI, WR and WI represent the real and imaginary parts of weight W, respectively, and CIWI is the result of the neuron. The activation function CReLU of the complex-valued hidden layer is defined as the association of the ReLU function in the real part and the imaginary part respectively, that is, the complex-valued vector h of the CSI is decomposed into the parameters of the ReLU function:
CIR＝IFFT(CFR*W)
the activation function of the real-valued hidden layer is a parameter-corrected linear unit PReLU, which is defined as:
f(h i )＝max(0，h i )+a i min(0，h i )
here hi denotes the input to the ith neuron of the ith hidden layer and ai is a self-learnable parameter used to control the coefficient of the slope of the negative part. Compared with the commonly used activation functions such as Sigmoid and tanh, the PReLU has the advantages of high operation speed, difficulty in gradient disappearance and the like.
Real-valued Mean Square Error (MSE) loss function
Is defined as:
the complex-valued Mean Square Error (MSE) loss function is defined as:
in fact, in the present invention, the training of the cvDNN network is performed in a supervised manner by MSE, with a learning rate of 0.001, 10000 back-propagation iterations. Through continuous training, the network model parameters with the best TOA precision, namely the weight W and the bias b, are obtained, and TOA prediction of the target equipment is realized.
In the WASP online system, a WIFI signal detector is adopted to detect a data packet of communication between target equipment and WIFI equipment. And extracting the CSI information of the target equipment from the data packet, and converting the CSI information into CIR information by the formula (1).
Due to multipath effect and ubiquitous noise pollution in the wireless signal transmission process, filtering processing is carried out on CIR information in order to eliminate the influence of noise, and background noise is eliminated.
In the invention, the calculation formula of the noise floor threshold is as follows:
in the formula, Anoise represents a threshold of the noise floor, and Asignal _ max represents a maximum value of the CIR. Through the above formula, the CIR value smaller than the noise floor threshold is set to 0, and the separation of the noise and the CIR is realized.
Finding the peak point of the filtered CIR value, and marking the peak point as an index, for example, marking the time position of the first peak as an index 0, marking the time position of the second peak as an index 1, and so on. In order to reduce the multipath effect, two side lobes of the CIR peak envelope are respectively extracted from the CIR peak center point, and data points which are 1% lower than the CIR peak envelope center point amplitude value are set to be 0. And (3) performing amplitude and phase normalization on the filtered CIR by adopting an equation (2), and offsetting the central point of the peak value to the marked index position.
Through the processing of the step, on one hand, the influence of noise is reduced, on the other hand, the influence of multipath effect on the CIR is reduced, so that CIR envelope energy is more concentrated, and the precision and the efficiency of subsequent cvDNN network training can be effectively improved
And taking the normalized CIR information as an input layer of the cvDNN network, wherein an output layer is TOA time, and the network model parameters are DNN model parameters (including weights and offsets of a real part and an imaginary part of a hidden layer) generated by the training, so that TOA prediction of the target equipment is realized. And finally, acquiring the position of the target equipment by using the time difference of arrival (TDOA) of the target equipment to the plurality of wireless detectors, thereby realizing the positioning of the target equipment.
The specific embodiment is as follows:
in order to verify the effect of the invention, in a field with the area of 900 square meters, 22 network nodes are distributed by taking 1 AP (wireless signal access point) as the center, wherein the 22 network nodes comprise 6 WIFI signal detectors and 1 dongle connected with a wireless network. The dongle is connected to a notebook computer by a USB interface and is used as target equipment for effect verification of the invention. The AP is configured under an 802.11ac mode, and the bandwidth is 80 MHz. The notebook computer is in wireless communication with the AP at regular time through the dongle. The AP is also used for WIFI signal transmission and delay calibration. The target device (notebook computer) is placed at different positions at different times, and the positioning performance of the content of the invention is evaluated through the processing.
Fig. 5 shows the wireless positioning effect according to the present invention. The solid red square represents a wireless signal detector, the solid green square represents the position of the AP, the "+" red sign represents the real position of the target equipment, and the "+" blue sign represents the position of the target equipment predicted by the CIR peak value deviation and the complex value deep neural network. It can be seen that the positioning error of 80% of the points in the positioning of the target device is 0.5m, while the positioning error of the MUSIC and ESPRIT methods commonly used for wireless positioning is 2 m. Therefore, the method has better positioning precision and popularization and application values.
In the technical scheme, a MonteCarlo method is adopted to simulate a Saleh-Vallenzuela channel model (hereinafter referred to as an S-V model) to generate a million CSI data sets, and the CSI is converted into a CIR for training a cvDNN model to obtain key parameters (weight and bias of a hidden layer) of the DNN model. The CIR peak value offset method and the filtering threshold value are provided, the influence of multipath effect and noise on DNN training is reduced, the accuracy of TOA time prediction is improved, and more accurate wireless positioning is realized. In addition, the invention does not need to collect the relevant data (such as position fingerprints, scene models and the like) of the position area in advance, and has better economy and universality. Secondly, compared with the MUSIC and ESPRIT methods which are commonly used for wireless positioning at present, the CIR filtering and peak value drifting method provided by the invention has the advantages that the positioning accuracy is improved to 0.5 meter from 2 meters, and therefore, the positioning accuracy is higher.
It should be noted that, in the present invention, unless otherwise explicitly specified or limited, the terms "mounted," "connected," "fixed," and the like are to be construed broadly, and for example, may be fixedly connected, detachably connected, or integrally connected; they may be connected directly or indirectly through intervening media, or they may be interconnected between two elements. The specific meanings of the above terms in the present invention can be understood by those skilled in the art according to specific situations.
The above description is only an example of the present invention, and the detailed structure and characteristics of the common general knowledge in the scheme are not described too much. It should be noted that, for those skilled in the art, without departing from the structure of the present invention, several changes and modifications can be made, which should also be regarded as the protection scope of the present invention, and these will not affect the effect of the implementation of the present invention and the practicability of the patent. The scope of the claims of the present application shall be defined by the claims, and the description of the embodiments and the like in the specification shall be used to explain the contents of the claims.
Claims (7)
1. A wireless positioning method based on CIR peak value deviation and complex value deep neural network is characterized by comprising the following steps:
s1, simulating the reflection and refraction of the WIFI signal by adopting a MonteCarlo method to generate a CSI data set;
s2, converting the generated CSI into CIR through inverse Fourier transform;
s3, normalizing the amplitude and the phase of the CIR;
s4, using the normalized training set for cvDNN training, and obtaining a model, weight and bias of a Deep Neural Network (DNN) in an off-line state;
s5, detecting a data packet of the target device and the WIFI device by using a WIFI signal detector;
s6, extracting CSI information of the target device from the data packet;
s7, converting the CSI of the target equipment into CIR;
s8, calculating the noise of the bottom plate, filtering the CIR, and marking the CIR peak position as an index;
s9, extracting two side lobes of CIR peak envelope by CIR peak center point, setting data point lower than 1% CIR peak center point amplitude value as 0;
s10, carrying out amplitude and phase normalization on the filtered CIR, and shifting the center point of the peak value to an index position;
s11, taking the normalized CIR as the input of the trained cvDNN network, and predicting TOA of the target equipment through DNN;
and S12, acquiring the position of the target equipment based on the arrival time difference of the unknown target equipment to the plurality of detectors, and realizing the positioning of the target equipment.
2. The method of claim 1, wherein in step S2, the inverse fourier transform function for converting CSI into CIR is:
CIR＝IFFT(CFR*W)
where W is the sampling width, CFR is the channel frequency response of the receiving end, and IFFT represents the inverse fourier transform.
3. The method as claimed in claim 1, wherein in step S3, the CIR magnitude and phase normalization function is:
CIR_norm[N]＝CIR[N]/max(CIR[N])
in the formula, CIR _ norm [ N ] is the normalized value of the Nth CIR point, CIR [ N ] represents the value of the Nth CIR point, and max (CIR [ N ]) represents the maximum value of the whole CIR sequence.
4. The method as claimed in claim 1, wherein in step S4, the relation between the weight and the offset is:
y＝f(x；W，b)＝f (1) (f (2) (f (3) (…f (L) (x))))
where L, W and b represent the number of layers, weights and offset values of the neural network, respectively.
5. The CIR peak shift and complex deep neural network-based wireless positioning method of claim 1, wherein in step S8, the noise floor threshold is calculated as:
in the formula, Anoise represents a threshold of the noise floor, and Asignal _ max represents a maximum value of the CIR.
6. The method of claim 1, wherein in step S4, the network structure of the cvDNN comprises 6 hidden layers, and 6 hidden layers are composed of a complex layer and a real layer, wherein the number of the complex layers is 3, and each layer has 256 neurons.
7. A wireless location system for performing the wireless location method of claim 1, wherein the wireless location system operates in a WASP system and comprises:
the wireless access point comprises a plurality of antennas for transmitting and receiving WIFI signals and realizes the transmission and the reception of the WIFI signals;
the wireless detectors are used for monitoring the WiFi network flow and equipment communication;
the target equipment sends a WiFi data packet to the AP;
and the wireless positioning terminal calculates the TDOA of the target equipment so as to solve the position of the target equipment.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210527921.8A CN115103438B (en) | 2022-05-16 | 2022-05-16 | Wireless positioning method based on CIR peak value offset and complex value deep neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210527921.8A CN115103438B (en) | 2022-05-16 | 2022-05-16 | Wireless positioning method based on CIR peak value offset and complex value deep neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115103438A true CN115103438A (en) | 2022-09-23 |
| CN115103438B CN115103438B (en) | 2024-07-16 |
Family
ID=83287003
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210527921.8A Active CN115103438B (en) | 2022-05-16 | 2022-05-16 | Wireless positioning method based on CIR peak value offset and complex value deep neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115103438B (en) |
Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9883511B1 (en) * | 2012-12-05 | 2018-01-30 | Origin Wireless, Inc. | Waveform design for time-reversal systems |
| CN109450830A (en) * | 2018-12-26 | 2019-03-08 | 重庆大学 | Channel estimation methods based on deep learning under a kind of high-speed mobile environment |
| CN111031477A (en) * | 2019-12-25 | 2020-04-17 | 中山大学 | Millimeter wave indoor positioning method based on deep learning |
| CN111212379A (en) * | 2020-01-06 | 2020-05-29 | 天津工业大学 | Novel CSI indoor positioning method based on convolutional neural network |
| CN111551893A (en) * | 2020-06-19 | 2020-08-18 | 天津大学 | Deep learning and neural network integrated indoor positioning method |
| US20200322868A1 (en) * | 2015-07-17 | 2020-10-08 | Origin Wireless, Inc. | Method, apparatus, and system for improving topology of wireless sensing systems |
| CN111819892A (en) * | 2018-03-06 | 2020-10-23 | 三星电子株式会社 | Method and apparatus for AI-based UE velocity estimation using uplink SRS measurements |
| CN112333653A (en) * | 2020-09-22 | 2021-02-05 | 西安电子科技大学 | Identity intelligent identification method and system based on WiFi channel state information |
| CN112712557A (en) * | 2020-12-17 | 2021-04-27 | 上海交通大学 | Super-resolution CIR indoor fingerprint positioning method based on convolutional neural network |
| US20210136515A1 (en) * | 2015-07-17 | 2021-05-06 | Feng Zhang | Method, apparatus, and system for wireless monitoring with motion localization |
| CN112867021A (en) * | 2021-01-13 | 2021-05-28 | 合肥工业大学 | Improved TrAdaBoost-based indoor positioning method for transfer learning |
| CN112911500A (en) * | 2021-01-15 | 2021-06-04 | 河海大学 | High-precision indoor positioning method based on multi-source data fusion |
| CN113015093A (en) * | 2021-01-29 | 2021-06-22 | 辽宁大学 | Indoor wireless positioning method based on three-dimensional depth residual error neural network |
| CN113383501A (en) * | 2019-01-18 | 2021-09-10 | 苹果公司 | Frequency domain Channel State Information (CSI) compression |
-
2022
- 2022-05-16 CN CN202210527921.8A patent/CN115103438B/en active Active
Patent Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9883511B1 (en) * | 2012-12-05 | 2018-01-30 | Origin Wireless, Inc. | Waveform design for time-reversal systems |
| US20200322868A1 (en) * | 2015-07-17 | 2020-10-08 | Origin Wireless, Inc. | Method, apparatus, and system for improving topology of wireless sensing systems |
| US20210136515A1 (en) * | 2015-07-17 | 2021-05-06 | Feng Zhang | Method, apparatus, and system for wireless monitoring with motion localization |
| CN111819892A (en) * | 2018-03-06 | 2020-10-23 | 三星电子株式会社 | Method and apparatus for AI-based UE velocity estimation using uplink SRS measurements |
| CN109450830A (en) * | 2018-12-26 | 2019-03-08 | 重庆大学 | Channel estimation methods based on deep learning under a kind of high-speed mobile environment |
| CN113383501A (en) * | 2019-01-18 | 2021-09-10 | 苹果公司 | Frequency domain Channel State Information (CSI) compression |
| CN111031477A (en) * | 2019-12-25 | 2020-04-17 | 中山大学 | Millimeter wave indoor positioning method based on deep learning |
| CN111212379A (en) * | 2020-01-06 | 2020-05-29 | 天津工业大学 | Novel CSI indoor positioning method based on convolutional neural network |
| CN111551893A (en) * | 2020-06-19 | 2020-08-18 | 天津大学 | Deep learning and neural network integrated indoor positioning method |
| CN112333653A (en) * | 2020-09-22 | 2021-02-05 | 西安电子科技大学 | Identity intelligent identification method and system based on WiFi channel state information |
| CN112712557A (en) * | 2020-12-17 | 2021-04-27 | 上海交通大学 | Super-resolution CIR indoor fingerprint positioning method based on convolutional neural network |
| CN112867021A (en) * | 2021-01-13 | 2021-05-28 | 合肥工业大学 | Improved TrAdaBoost-based indoor positioning method for transfer learning |
| CN112911500A (en) * | 2021-01-15 | 2021-06-04 | 河海大学 | High-precision indoor positioning method based on multi-source data fusion |
| CN113015093A (en) * | 2021-01-29 | 2021-06-22 | 辽宁大学 | Indoor wireless positioning method based on three-dimensional depth residual error neural network |
Non-Patent Citations (3)
| Title |
|---|
| MINGYANG ZHENG; SHUFANG LI; LI DENG; MEIJUN QU; CHEN ZHANG; CHENWEI CAI: ""CSI-Based Indoor High-Precision Localization System"", 《2018 ASIA-PACIFIC MICROWAVE CONFERENCE (APMC)》, 17 January 2019 (2019-01-17) * |
| 胡云冰;童世华: ""一种便携式智能跌倒检测仪的设计与实现"", 《自动化仪表》, 20 February 2015 (2015-02-20) * |
| 陈恩庆;陶然;张卫强: ""一种基于分数阶傅立叶变换的时变信道参数估计方法"", 《电子学报》, 25 December 2005 (2005-12-25) * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115103438B (en) | 2024-07-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hsieh et al. | Deep learning-based indoor localization using received signal strength and channel state information | |
| Dvorecki et al. | A machine learning approach for Wi-Fi RTT ranging | |
| CN110351655B (en) | Indoor positioning method and system based on signal multipath propagation measurement | |
| CN104812061B (en) | A kind of indoor ranging and localization method based on MIMO-OFDM channel state information | |
| EP2798366B1 (en) | Method and system for localization | |
| CN110290491B (en) | Indoor target positioning method based on multipath assistance | |
| Yang et al. | Decimeter level indoor localization using WiFi channel state information | |
| Subhan et al. | Experimental analysis of received signals strength in Bluetooth Low Energy (BLE) and its effect on distance and position estimation | |
| Liu et al. | LC-DNN: Local connection based deep neural network for indoor localization with CSI | |
| Nguyen et al. | Deep gated recurrent unit-based 3D localization for UWB systems | |
| CN105611629A (en) | 60GHz millimeter wave non-line of sight identification and wireless fingerprint positioning method based on energy detection | |
| CN105554882B (en) | The identification of 60GHz non line of sight and wireless fingerprint positioning method based on energy measuring | |
| He et al. | Multi-antenna array-based aoa estimation using bluetooth low energy for indoor positioning | |
| Tan et al. | UAV localization with multipath fingerprints and machine learning in urban NLOS scenario | |
| CN103096465A (en) | Environment self-adaption multi-target direct locating method | |
| Van Marter et al. | Support vector regression for bluetooth ranging in multipath environments | |
| CN105611628A (en) | High precision pulse 60GHz wireless fingerprint positioning method based on energy detection | |
| CN105738866A (en) | 60GHz Non-Line-of-Sight identification and wireless fingerprint positioning method based on energy detection | |
| CN115103438B (en) | Wireless positioning method based on CIR peak value offset and complex value deep neural network | |
| Li et al. | Robust kernel-based machine learning localization using NLOS TOAs or TDOAs | |
| Hedley et al. | Wireless sensor network using hybrid TDOA/RSS tracking of uncooperative targets | |
| CN115908547A (en) | Wireless positioning method based on deep learning | |
| US11815591B2 (en) | Distance estimation method in an LPWA network and a position estimation method associated therewith | |
| Lv et al. | Accessorial locating for internet of vehicles based on doa estimation in industrial transportation | |
| Qin et al. | A ML-based direct localization method for multiple sources with moving arrays |
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 |