CN111093163B - Passive target classification method based on channel state information - Google Patents
Passive target classification method based on channel state information Download PDFInfo
- Publication number
- CN111093163B CN111093163B CN201911333553.8A CN201911333553A CN111093163B CN 111093163 B CN111093163 B CN 111093163B CN 201911333553 A CN201911333553 A CN 201911333553A CN 111093163 B CN111093163 B CN 111093163B
- Authority
- CN
- China
- Prior art keywords
- training
- data
- neural network
- classification
- state information
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 44
- 238000012549 training Methods 0.000 claims abstract description 38
- 238000013528 artificial neural network Methods 0.000 claims abstract description 32
- 238000007781 pre-processing Methods 0.000 claims abstract description 8
- 230000004913 activation Effects 0.000 claims description 27
- 238000013145 classification model Methods 0.000 claims description 16
- 238000012360 testing method Methods 0.000 claims description 13
- 210000002569 neuron Anatomy 0.000 claims description 12
- 238000012544 monitoring process Methods 0.000 claims description 10
- 238000001514 detection method Methods 0.000 claims description 4
- 238000013480 data collection Methods 0.000 claims description 3
- 210000005036 nerve Anatomy 0.000 claims description 3
- 230000001537 neural effect Effects 0.000 claims description 3
- 238000003062 neural network model Methods 0.000 claims description 3
- 238000012795 verification Methods 0.000 claims description 3
- 210000004027 cell Anatomy 0.000 claims description 2
- 230000006870 function Effects 0.000 description 23
- 238000010586 diagram Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
Images
Classifications
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B17/00—Monitoring; Testing
- H04B17/30—Monitoring; Testing of propagation channels
- H04B17/309—Measuring or estimating channel quality parameters
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Quality & Reliability (AREA)
- Electromagnetism (AREA)
- Monitoring And Testing Of Transmission In General (AREA)
Abstract
The invention relates to a passive target classification method based on channel state information, which comprises the following steps in sequence: collecting data; preprocessing data; training a neural network; and (5) classifying results. The method has the advantages of low deployment cost, high classification accuracy, no need of manually extracting features and privacy protection; by using more stable channel state information as a base signal instead of the received signal strength which fluctuates greatly with time, for the height classification with fine granularity, the channel state information can better reflect the difference of the target height compared with the received signal strength indicating value.
Description
Technical Field
The invention relates to the technical field of passive sensing in a wireless network, in particular to a passive target classification method based on channel state information.
Background
The target classification has important application value in the fields of safety monitoring, emergency rescue, border patrol, intelligent interaction and the like. There are three methods widely used in the field of object classification, namely a video image-based method, a Synthetic Aperture Radar (SAR) -based method, and a sensor-based method. The most notable method is a video image and Synthetic Aperture Radar (SAR) based method. However, the cost of using either approach is relatively high, which hinders the possibility of deploying them for low-cost perception, and the use of video image-based approaches may carry the risk of privacy disclosure; the sensor-based method, which uses a mixture of various sensors such as acoustic, passive infrared, magnetic field and electrostatic sensors for object classification, can greatly reduce costs compared to the two aforementioned methods, however, the classification accuracy of the method needs to be further improved.
The passive sensing technology has no problems of high deployment cost and privacy disclosure. Currently, sensing methods based on wireless signals are receiving wide attention. The Passive sensing (DFP) senses the surrounding environment by using radio signals, senses the information of motion, position, height and the like of people by the influence of users on the radio signals, and can directly make up for the defects that a video image method has monitoring dead angles and monitoring cameras cannot be arranged in some places. Moreover, the technologies can be realized on common commercial Wi-Fi, so that the deployment complexity and the installation cost of detection devices such as video monitors and cameras and the labor cost of monitoring are directly reduced. Therefore, the passive sensing technology has great potential in the aspects of safety protection, smart home and the like. The theoretical basis for classification of objects using passive sensing is: the transmitted RF signal consists of several components and these components arrive at the receiving end from different paths forming a superposition. Since the RF signal can be easily blocked, reflected or scattered by different targets, each target will cause a different form of signal change. By identifying and interpreting these changes, the types of different objects can be detected and classified.
Most of the existing passive target classification systems adopt the received signal strength indication as a base signal, and professor Yangzhou of Qinghua university has pointed out that the measured values of the received signal strength indication at the same position and different moments under the same condition have great difference, which leads to the problem of insufficient classification precision caused by using the signal. Increasing the accuracy requires more wireless transceiver nodes to be deployed, which necessarily results in increased costs, and therefore creates a trade-off between classification accuracy and cost. In addition, most passive target classification systems adopt a method of manually extracting signal features, and find features which can easily distinguish different targets in a time domain, a frequency domain and the like, which can cause a problem of high cost.
Disclosure of Invention
The invention aims to provide a method for balancing classification precision and cost by using channel state information to replace a received signal strength indicator as a base signal; a passive object classification method based on channel state information reduces overhead by using a neural network model with autonomous learning data features instead of manually extracting features.
In order to achieve the purpose, the invention adopts the following technical scheme: a passive object classification method based on channel state information, the method comprising the sequential steps of:
(1) data collection:
(1a) deploying a pair of wireless signal transceivers in a monitoring area, fixing K positions for identifying and classifying in the monitoring area, and setting a packet sending rate of 100 HZ;
(1b) each target stands still at K positions for a period of time S respectively, and channel state information H of different targets is collected; the targets are persons standing with different heights and weights;
(2) data preprocessing:
(2a) for data collected by each target at each position, 4000 data packets of a middle stable state are taken, and channel state information comprises amplitude information and phase information;
(2b) carrying out denoising pretreatment on data by adopting continuous wavelet transform;
(3) training a neural network classification model: using 60% of the data after the step (2) as a training set, 20% as a verification set and 20% as a test set, and using the BP neural network to use the data of the training set for training the neural network to obtain a classification model;
(4) and (4) classification results: after the training of the classification model is finished, the data of the test set is sent into the classification model, and then the classification of the test target can be obtained.
In the step (1a), the wireless signal transceiver works in an ISM frequency band of 2.4GHZ, the wireless signal transceiver comprises a signal transmitter and a signal receiver, the wireless router is used as the signal transmitter, and a notebook computer equipped with a wireless network card is used as the signal receiver.
When the channel state information is collected in step (1b), the channel is divided into a plurality of subcarriers with different center frequencies by using the OFDM technique, and the number of the subcarriers is 30.
In the step (3), the method for obtaining the classification model by adopting the BP neural network training comprises the following steps:
the BP neural network comprises an input layer, a hidden layer and an output layer, the number of neurons of the input layer and the output layer is determined according to input training set data and classification requirements, the number of neurons of the hidden layer is two times or three times of the number of neurons of the input layer, an activation function between the layers is a sigmoid function, and the expression of the activation function is as follows:
the first derivative of the activation function can be derived:
the training steps are as follows:
firstly, a random initialization weight theta is a vector in a (0,1) range, and the dimensionality is determined by the number of front and rear nerve units;
secondly, forward propagation is achieved, i.e. for each input sample x(i)Calculating the activation value a layer by layer(l)Where l is the layer of the neural network;
hidden layer activation value: a is(2)=f((-θ(1))Tx(i)) (3)
In the formula, a(2)For input sample x(i)An activation value after the hidden layer is activated by the function;
output layer activation value: a is(3)=f((θ(2))Ta(2))=hθ(x(i)) (4)
a(3),hθ(x(i)) Are all input samples x(i)An activation value after the output layer has been activated by the activation function; theta is an expression form after vectorization, and the function represented by f (-) is a sigmoid function;
then, writing codes and calculating a cost function J (theta) according to the following formula;
in the above formula, m is the number of training samples, K is the number of neural units in the output layer, L is the total number of layers of the neural network, and SlThe number of the cells of the l layer is, and lambda is a regularization parameter;the sample value of the kth neuron of the ith sample in the output layer;
Partial derivatives D calculated with error backpropagation using gradient detection comparisonsij lAnd the error between the calculated partial derivative numgard is numerically estimated:
|Dij l-numgrad|<ε,ε∈(0,0.01) (6)
finally, find θ that minimizes the cost function using fmicg function:
at this point, the training of the entire neural network is completed.
The wireless signal transceiver supports an IEEE 802.11a/b/g/n wireless protocol.
According to the technical scheme, the beneficial effects of the invention are as follows: the method has the advantages of low deployment cost, high classification accuracy, no need of manually extracting features and privacy protection; the passive method has the challenges of selecting a base signal and extracting the characteristics of easily distinguished targets, and the method improves the classification precision by using more stable channel state information to replace the received signal strength which fluctuates greatly along with time as the base signal and simultaneously using a neural network method with the characteristics of self-learning data to extract and classify the characteristics.
Drawings
FIG. 1 is a flow chart of a method of the present invention;
FIG. 2 is a diagram of height and weight information of a volunteer participating in an experiment according to a first embodiment of the present invention;
FIG. 3 is a schematic diagram of specific locations of wireless transceivers and test point settings in an experimental scene office environment;
FIG. 4 is a table of confusion matrices for target classification results in an office environment;
FIG. 5 is a comparison result of whether to perform denoising in an experimental environment and the classification accuracy of the method based on the received signal strength indication.
Detailed Description
As shown in fig. 1, a passive object classification method based on channel state information includes the following sequential steps:
(1) data collection:
(1a) deploying a pair of wireless signal transceivers in a monitoring area, fixing K positions for identifying and classifying in the monitoring area, and setting a packet sending rate of 100 HZ;
(1b) each target stands still at K positions for a period of time S respectively, and channel state information H of different targets is collected; the targets are persons standing with different heights and weights;
(2) data preprocessing:
(2a) for data collected by each target at each position, 4000 data packets of a middle stable state are taken, and channel state information comprises amplitude information and phase information;
(2b) carrying out denoising pretreatment on data by adopting continuous wavelet transform;
(3) training a neural network classification model: using 60% of the data after the step (2) as a training set, 20% as a verification set and 20% as a test set, and using the BP neural network to use the data of the training set for training the neural network to obtain a classification model;
(4) and (4) classification results: after the training of the classification model is finished, the test set data is sent into the classification model, and then the classification of the test target can be obtained.
In the step (1a), the wireless signal transceiver works in an ISM frequency band of 2.4GHZ, the wireless signal transceiver comprises a signal transmitter and a signal receiver, the wireless router is used as the signal transmitter, and a notebook computer equipped with a wireless network card is used as the signal receiver.
In the step (1b), when the channel state information is collected, the channel is divided into a plurality of subcarriers with different center frequencies by using the OFDM technique, and the number of the subcarriers is 30.
In the step (3), the method for obtaining the classification model by adopting the BP neural network training comprises the following steps:
the BP neural network comprises an input layer, a hidden layer and an output layer, the number of neurons of the input layer and the output layer is determined according to input training set data and classification requirements, the number of neurons of the hidden layer is two times or three times of the number of neurons of the input layer, an activation function between the layers is a sigmoid function, and the expression of the activation function is as follows:
the first derivative of the activation function can be derived:
the training steps are as follows:
firstly, a random initialization weight theta is a vector in a (0,1) range, and the dimensionality is determined by the number of front and rear nerve units;
secondly, forward propagation is achieved, i.e. for each input sample x(i)Calculating the activation value a layer by layer(l)Where l is the layer of the neural network;
hidden layer activation value: a is a(2)=f((-θ(1))Tx(i)) (3)
In the formula (I), the compound is shown in the specification,a(2)for input sample x(i)An activation value after the hidden layer is activated by the function;
output layer activation value: a is(3)=f((θ(2))Ta(2))=hθ(x(i)) (4)
a(3),hθ(x(i)) Are all input samples x(i)An activation value after the output layer has been activated by the activation function; theta is an expression form after vectorization, and the function represented by f (-) is a sigmoid function;
then, writing codes and calculating a cost function J (theta) according to the following formula;
in the above formula, m is the number of training samples, K is the number of neural units in the output layer, L is the total number of layers of the neural network, and SlIs the unit number of the l layer, and lambda is a regularization parameter;the sample value of the kth neuron of the ith sample in the output layer;
Partial derivatives D calculated with error backpropagation using gradient detection comparisonsij lAnd the error between the calculated partial derivative numgard is numerically estimated:
|Dij l-numgrad|<ε,ε∈(0,0.01) (6)
finally, find θ that minimizes the cost function using fmicg function:
at this point, the training of the entire neural network is completed.
The whole training process is to obtain a classification model and find theta which minimizes the cost function.
The wireless signal transceiver supports an IEEE 802.11a/b/g/n wireless protocol.
Example one
Firstly, collecting data, namely collecting data at four positions of a hexagonal mark of experimenters with different heights and weights shown in figure 2 in an experimental scene shown in figure 3, wherein an experimental platform selects a router as a transmitting end and is 80cm away from the ground; and a notebook computer provided with an Inter 5300 network card and an open source tool CSI-Tools is used as a receiving end and is 50cm away from the ground. Connect to the designated router in an unencrypted manner and generate approximately 100 packets per second using ping-i 0.01 commands. Each experimenter fixes the position for 1 minute, approximately 6000 data packets can be collected, 4000 data packets in a middle stable state are taken, channel state information comprises amplitude and phase information, and original phase information is disordered and irregular, so that the amplitude is only extracted from the 4000 data packets. The experiment is divided into 4 height segments, namely 150-.
Secondly, preprocessing the data of different experimenters at different positions, namely denoising by adopting a wavelet transform method. After preprocessing, 60% of the 4000 points were used as training set, 20% as test set, and the remaining 20% as validation set.
FIG. 4 is a diagram of a neural network model with a network structure of 30-90-4 obtained by preprocessing collected data, labeling the data of experimenters with different height segments with 1-4 labels, and sending the data to neural network training and repeatedly training. Through experimental tests, the ratio of the classification times of the correct height to the total times of the tests is counted to obtain the confusion matrix of different height segments shown in the figure 4, and the average classification accuracy reaches 92%.
Fig. 5 compares the influence of 3 base signals of RSSI, CSI original noisy signal and CSI denoised signal on the high classification performance, and the 3 types of data are respectively used as the data sets of the neural network, and the result is shown in fig. 5. for the RSSI signal, training is repeated for 10 times, and the average classification accuracy of the neural network is 41.24%. For the CSI noisy signal, the average classification accuracy of the neural network is 83.90%, and similarly, for the CSI denoised signal, the average classification accuracy is 90.30%. The result shows that the adoption of the CSI as the base signal is superior to the RSSI, and the feasibility of denoising by adopting wavelet transformation is further proved.
In conclusion, the method has the advantages of low deployment cost, high classification accuracy, no need of manually extracting features and privacy protection; the passive method has the challenges of selecting a base signal and extracting the characteristics of easily distinguished targets, and the method improves the classification precision by using more stable channel state information to replace the received signal strength which fluctuates greatly along with time as the base signal and simultaneously using a neural network method with the characteristics of self-learning data to extract and classify the characteristics.
Claims (4)
1. A passive target classification method based on channel state information is characterized in that: the method comprises the following steps in sequence:
(1) data collection:
(1a) deploying a pair of wireless signal transceivers in a monitoring area, fixing K positions for identifying and classifying in the monitoring area, and setting a packet sending rate of 100 HZ;
(1b) each target stands still at K positions for a period of time S respectively, and channel state information H of different targets is collected; the targets are persons standing at different heights and weights;
(2) data preprocessing:
(2a) for data collected by each target at each position, 4000 data packets of a middle stable state are taken, and channel state information comprises amplitude information and phase information;
(2b) carrying out denoising pretreatment on data by adopting continuous wavelet transform;
(3) training a neural network classification model: using 60% of the data after the step (2) as a training set, 20% as a verification set and 20% as a test set, and using the BP neural network to use the data of the training set for training the neural network to obtain a classification model; dividing the height into 4 height segments, namely 150-159cm, 160-169cm, 170-179cm and 180-189cm, preprocessing the collected data, respectively marking the data of experimenters with different height segments with 1-4 labels, sending the data into a neural network for training, and obtaining a neural network model with a network structure of 30-90-4 through repeated training so as to realize the height classification of the target;
(4) and (4) classification results: after the training of the classification model is finished, the data of the test set is sent into the classification model to obtain the classification of the test target;
in the step (3), the method for obtaining the classification model by adopting the BP neural network training comprises the following steps:
the BP neural network comprises an input layer, a hidden layer and an output layer, the number of neurons of the input layer and the output layer is determined according to input training set data and classification requirements, the number of neurons of the hidden layer is two times or three times of the number of neurons of the input layer, an activation function between the layers is a sigmoid function, and the expression of the activation function is as follows:
the first derivative of the activation function can be found:
the training steps are as follows:
firstly, randomly initializing a weight theta to be a vector in a (0,1) range, wherein the dimensionality is determined by the number of front and rear nerve units;
secondly, forward propagation is achieved, i.e. for each input sample x(i)Calculating the activation value a layer by layer(l)Where l is the layer of the neural network;
hidden layer activation value: a is(2)=f((-θ(1))Tx(i)) (3)
In the formula, a(2)For input samples x(i)An activation value after the hidden layer is activated by the function;
output layer activation value: a is(3)=f((θ(2))Ta(2))=hθ(x(i)) (4)
a(3),hθ(x(i)) Are all input samples x(i)An activation value after the output layer has been activated by the activation function; theta is an expression form after vectorization, and the function represented by f (-) is a sigmoid function;
then, writing codes and calculating a cost function J (theta) according to the following formula;
in the above formula, m is the number of training samples, K is the number of neural units in the output layer, L is the total number of layers of the neural network, and SlThe number of the cells of the l layer is, and lambda is a regularization parameter;the sample value of the kth neuron of the ith sample in the output layer;
Partial derivatives D calculated with error backpropagation using gradient detection comparisonsij lAnd the error between the calculated partial derivative numgard is numerically estimated:
|Dij l-numgrad|<ε,ε∈(0,0.01) (6)
finally, find θ that minimizes the cost function using fmicg function:
at this point, the training of the entire neural network is completed.
2. The passive object classification method based on channel state information according to claim 1, characterized in that: in the step (1a), the wireless signal transceiver works in an ISM frequency band of 2.4GHZ, the wireless signal transceiver comprises a signal transmitter and a signal receiver, the wireless router is used as the signal transmitter, and a notebook computer equipped with a wireless network card is used as the signal receiver.
3. The passive object classification method based on channel state information according to claim 1, characterized in that: when the channel state information is collected in step (1b), the channel is divided into a plurality of subcarriers with different center frequencies by using the OFDM technique, and the number of the subcarriers is 30.
4. The passive object classification method based on channel state information according to claim 1, characterized in that: the wireless signal transceiver supports an IEEE 802.11a/b/g/n wireless protocol.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911333553.8A CN111093163B (en) | 2019-12-23 | 2019-12-23 | Passive target classification method based on channel state information |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911333553.8A CN111093163B (en) | 2019-12-23 | 2019-12-23 | Passive target classification method based on channel state information |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111093163A CN111093163A (en) | 2020-05-01 |
CN111093163B true CN111093163B (en) | 2022-06-17 |
Family
ID=70396514
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911333553.8A Active CN111093163B (en) | 2019-12-23 | 2019-12-23 | Passive target classification method based on channel state information |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111093163B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114376517A (en) * | 2021-11-30 | 2022-04-22 | 西安交通大学 | Physiological index measuring method and system based on wireless radio frequency signal |
US11784693B1 (en) | 2022-08-18 | 2023-10-10 | Corning Research & Development Corporation | Coverage cluster-based beamforming in a wireless node in a wireless communications system (WCS) |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10476730B2 (en) * | 2008-09-23 | 2019-11-12 | Origin Wireless, Inc. | Methods, apparatus, servers, and systems for human identification based on human radio biometric information |
US10045717B2 (en) * | 2016-06-10 | 2018-08-14 | The Regents Of The University Of California | WiFi-based person-identification technique for use in smart spaces |
CN108833036B (en) * | 2018-05-31 | 2020-11-03 | 湖南大学 | Human body falling detection method |
CN110062379B (en) * | 2019-04-15 | 2022-10-28 | 哈尔滨工程大学 | Identity authentication method based on channel state information under human behavior scene |
CN110475221B (en) * | 2019-07-15 | 2021-05-04 | 南京邮电大学 | Personnel action identification and position estimation method based on channel state information |
-
2019
- 2019-12-23 CN CN201911333553.8A patent/CN111093163B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN111093163A (en) | 2020-05-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108629380B (en) | Cross-scene wireless signal sensing method based on transfer learning | |
JP7154295B2 (en) | Motion detection based on machine learning of radio signal characteristics | |
CN109672485B (en) | Indoor personnel real-time invasion and movement speed detection method based on channel state information | |
Medaiyese et al. | Wavelet transform analytics for RF-based UAV detection and identification system using machine learning | |
CN105807935B (en) | A kind of gesture control man-machine interactive system based on WiFi | |
CN107290741B (en) | Indoor human body posture identification method based on weighted joint distance time-frequency transformation | |
CN110502105B (en) | Gesture recognition system and method based on CSI phase difference | |
Huang et al. | Machine-learning-based data processing techniques for vehicle-to-vehicle channel modeling | |
CN110337066A (en) | Based on channel state information indoor occupant activity recognition method, man-machine interactive system | |
CN107480699A (en) | A kind of intrusion detection method based on channel condition information and SVMs | |
CN111093163B (en) | Passive target classification method based on channel state information | |
Huang et al. | An accurate and efficient device-free localization approach based on sparse coding in subspace | |
CN107992882A (en) | A kind of occupancy statistical method based on WiFi channel condition informations and support vector machines | |
CN109474890A (en) | A kind of indoor objects passive tracking method based on WiFi Doppler frequency shift | |
Yang et al. | People counting based on CNN using IR-UWB radar | |
Yang et al. | Door-monitor: Counting in-and-out visitors with COTS WiFi devices | |
Xia et al. | Person identification with millimeter-wave radar in realistic smart home scenarios | |
CN113609976A (en) | Direction-sensitive multi-gesture recognition system and method based on WiFi (Wireless Fidelity) equipment | |
Zhu et al. | NotiFi: A ubiquitous WiFi-based abnormal activity detection system | |
CN106570490A (en) | Pedestrian real-time tracking method based on fast clustering | |
Shang et al. | LSTM-CNN network for human activity recognition using WiFi CSI data | |
CN113033654A (en) | Indoor intrusion detection method and system based on WiFi channel state information | |
CN114818916B (en) | Road target classification method based on millimeter wave radar multi-frame point cloud sequence | |
CN112949487A (en) | WiFi-based ship cab personnel number detection method and system | |
CN114781463A (en) | Cross-scene robust indoor tumble wireless detection method and related equipment |
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 |