CN113609977A - Pedestrian gait recognition method based on channel state information quotient distance - Google Patents

Pedestrian gait recognition method based on channel state information quotient distance Download PDF

Info

Publication number
CN113609977A
CN113609977A CN202110892189.XA CN202110892189A CN113609977A CN 113609977 A CN113609977 A CN 113609977A CN 202110892189 A CN202110892189 A CN 202110892189A CN 113609977 A CN113609977 A CN 113609977A
Authority
CN
China
Prior art keywords
quotient
csi
information
csid
data
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
Application number
CN202110892189.XA
Other languages
Chinese (zh)
Other versions
CN113609977B (en
Inventor
王林
张德安
刘文远
王新雨
厉斌斌
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.)
Yanshan University
Original Assignee
Yanshan University
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 Yanshan University filed Critical Yanshan University
Priority to CN202110892189.XA priority Critical patent/CN113609977B/en
Publication of CN113609977A publication Critical patent/CN113609977A/en
Application granted granted Critical
Publication of CN113609977B publication Critical patent/CN113609977B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/02—Preprocessing
    • G06F2218/04—Denoising
    • G06F2218/06—Denoising by applying a scale-space analysis, e.g. using wavelet analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
    • G06F21/30—Authentication, i.e. establishing the identity or authorisation of security principals
    • G06F21/31—User authentication
    • G06F21/32—User authentication using biometric data, e.g. fingerprints, iris scans or voiceprints
    • 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
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Data Mining & Analysis (AREA)
  • Software Systems (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computational Linguistics (AREA)
  • Computer Security & Cryptography (AREA)
  • Biomedical Technology (AREA)
  • Health & Medical Sciences (AREA)
  • Biophysics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Signal Processing (AREA)
  • Computer Hardware Design (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)

Abstract

The invention relates to the technical field of identity recognition, in particular to a pedestrian gait recognition method based on channel state information quotient distance, wherein the data acquisition comprises the following steps: acquiring CSI data of people walking in a sensing area by using one WiFi transmitting terminal and two WiFi receiving terminals; signal preprocessing: carrying out relevant denoising processing on the collected CSI data, and eliminating irrelevant noise information which is contained in the CSI data and is related to human walking; CSID extraction: the change information of the channel state information quotient on the complex plane contains the gait characteristics of human walking, and CSID is proposed to represent the change of the channel state information quotient on the complex plane; identity recognition: and combining the advantages of deep learning in the aspect of feature extraction, and using the LSTM network model to extract features so as to realize identity recognition. The method can realize the passive identity perception of people in the indoor environment only by using commercial WiFi equipment, and realizes a user identity recognition system by utilizing the unique gait characteristics of human walking and combining a deep learning model.

Description

Pedestrian gait recognition method based on channel state information quotient distance
Technical Field
The invention relates to the technical field of identity recognition, in particular to a pedestrian gait recognition method based on channel state information quotient distance.
Background
With the rise of the internet of things, the tide of all things interconnection is around the world, and a WiFi signal perception-based related technology is endless, so that the perception advantage of no equipment in the environment of the internet of things becomes a hot spot in the field of ubiquitous computing. The emergence of these key application technologies also provides research background and significance for the research of application security based on WiFi perception, and for many applications, it is necessary to perceive human identity information in advance.
Identity recognition is always the research focus in the fields of pervasive computing and man-machine interaction, and has extremely high economic value and wide application scenes in the fields of intelligent home, safety monitoring and the like. In the aspect of identification, a plurality of solutions are emerging, and by utilizing a plurality of modes including a camera, Wi-Fi, RFID, sound signals and the like, deep learning networks are increasingly integrated into a positioning system in recent research. Among them, the Wi-Fi signal-based identification technology has received much attention due to wide deployment and strong usability.
Identity recognition refers to the process of predefining identity information of a person and matching the identity information of the current person with a database of known identity information. The traditional identity identification methods such as identity authentication based on password input, gesture unlocking of a mobile phone screen and the like have low security because passwords and gestures can be stolen by others. At present, identity recognition based on unique biological characteristics of human is a high-safety mode, and the human biological characteristics can be roughly divided into two types in the biological characteristics, namely human physiological characteristics which mainly comprise human faces, irises, fingerprints, palmprints, DNA and the like. The other is human behavior characteristics, which mainly comprise gait, voice, key action, handwritten signature and the like. Identity authentication based on human biometrics has been widely used in the current society, and various devices (fingerprint check-in machines, DNA detectors, etc.) have been designed, which tend to be expensive. In addition, the identity of a person is authenticated through a face and a gait based on a camera, and the camera-based method has high accuracy, but can involve privacy problems in some environments and has dependence on illumination conditions. By combining the authentication modes, each identity recognition method has respective advantages and disadvantages due to different application scenes. People hope to get rid of the defects in the intelligent space perception application and hope to passively perceive the identity of people in the environment by a perception technology with the advantages of universality, easiness in deployment and the like.
Disclosure of Invention
In order to solve the problems, the pedestrian gait recognition method based on the channel state information quotient distance can realize the passive identity sensing of people in the indoor environment by only using commercial WiFi equipment, does not need to use a specific sensor, and does not need to actively interact with equipment. The method realizes a user identity recognition system by utilizing the unique gait characteristics of human walking and combining a deep learning model.
In order to achieve the purpose, the invention adopts the technical scheme that:
the pedestrian gait recognition method based on the channel state information quotient distance comprises the following steps:
step 1, data acquisition: collecting CSI data of people walking in a sensing area by using one WiFi transmitting terminal and two WiFi receiving terminals for model training and testing;
step 2, signal preprocessing: carrying out relevant denoising processing on the collected CSI data, and eliminating irrelevant noise information which is contained in the CSI data and is related to human walking;
step 3, CSID extraction: the change information of the channel state information quotient on the complex plane contains the gait characteristics of human walking, and CSID is proposed to represent the change of the channel state information quotient on the complex plane;
step 4, identity recognition: and combining the advantages of deep learning in the aspect of feature extraction, and using the LSTM network model to extract features so as to realize identity recognition.
Preferably, in the step 1, the CSI data on two vertical WiFi links are selected for system testing, each piece of CSI data is subjected to data preprocessing, initial denoising is carried out through two antenna operators to obtain a CSI quotient, then amplitude quotient and phase difference information of the CSI quotient are further extracted, the amplitude quotient and the phase difference are further denoised through hampel filtering and discrete wavelet transformation, and the denoised amplitude quotient and the denoised phase difference are reconstructed again to obtain the denoised CSI quotient; and extracting the CSID information containing the subcarriers from the movement of the CSI quotient in the complex plane, and encapsulating the CSID information with the label representing the real identity to form a sample in the data set.
Preferably, in step 2, the amplitude quotient and the phase difference information of the CSI quotient are extracted and further denoised, and then the denoised amplitude quotient and phase difference are subjected to signal inverse transformation to obtain the denoised CSI quotient information.
Preferably, step 2 comprises
Step 21, CSI quotient denoising: the CSI obtained by the two antennas is divided, the CSI is complex information which comprises amplitude and phase information of signals, the complex division operation can know that the amplitude of a CSI quotient is changed into the ratio of the amplitudes of the two antennas, the phase of the CSI quotient represents the difference of the phases of the two antennas, and the CSI quotient information obtained by the two antennas contains signal change information caused by walking of people;
step 22, Hampel filtering and denoising: removing abnormal values obviously existing in the signal waveform by using a Hampel filter;
step 23, denoising through discrete wavelet transform: the system further denoises amplitude quotient and phase difference information after the hampel filtering by adopting Discrete Wavelet Transform (DWT), and focuses signal waveforms on human walking behaviors to the maximum extent; when an original signal X with the frequency of F is decomposed by utilizing discrete wavelet transform, the original signal X is filtered by a low-pass filter and a high-pass filter; the wavelet coefficients are decomposed into 1 st-level approximate coefficients ACs through a low-pass filter, the 1 st-level detail coefficients DCs are decomposed through a high-pass filter, the approximate coefficients ACs and the detail coefficients DCs respectively correspond to the low-frequency part of 0-F/2 and the high-frequency part from F/2 to F of an original signal, and then the approximate coefficients obtained through each level of decomposition are subjected to iterative decomposition through the same decomposition process, so that more related wavelet coefficients with different frequency information and time scales can be obtained.
Preferably, in step 23, a db4 wavelet is selected to perform 3-level wavelet decomposition on the amplitude quotient and the phase difference of the selected sub-carriers; the sampling rate of the CSI collection is 1000Hz, the highest frequency of signals contained in CSI data is 500Hz, the 3 rd-level approximation coefficient is reconstructed to obtain signals containing 0-62.5Hz, and the discrete wavelet transformation can better remove signal noise irrelevant to human walking movement in waveforms, so that the denoised signal waveforms are better focused on signal waveform changes caused by human walking.
Preferably, in step 4, a two-layer long-and-short-term memory neural network is adopted, the gait features of each user to be identified are extracted by using an LSTM unit structure to realize robust identity identification, and the neural network model extracts the gait features contained in the CSID through an LSTM unit in a neural network hidden layer for the input CSID sequence.
Preferably, CSID obtained by signal preprocessing acquired CSI data is used as input of a network model, and a CSID sequence of 30 subcarriers in CSI information is expressed as the following formula
Figure BDA0003196344520000041
Wherein i represents a subcarrier number, and j represents time series information; CSID indicates information of change of euclidean distance from a point on the CSI quotient complex plane to an initial point in a time series, and one sample data in a data set may be represented as D { (X)iY), i ═ 1,2, 3.., 30, where D includes CSID sequences extracted from 30 subcarriers in the CSI dataAnd an identity tag corresponding thereto; xiThe method comprises the steps of representing 30 CSID sequences obtained by processing 30 subcarrier information, wherein Y represents a label value and represents identity information; identification questions are created models that can be based on an input XiMatrix is used for predicting identity label Y;
for each hidden layer input, LSTM may set times t and i to [1,2 ═ m]Input of layers
Figure BDA0003196344520000042
Mapping as feature ftAs shown in formula:
Figure BDA0003196344520000043
wherein g represents an activation function, wherein,
Figure BDA0003196344520000044
representing input, containing current information of the sequence
Figure BDA0003196344520000045
Short-term information
Figure BDA0003196344520000051
And long term information
Figure BDA0003196344520000052
Wherein the dimension size is d-30, P represents a weight matrix, and b is an offset;
the output of the LSTM is matrix transformed by the full connectivity layer, and the probability of identifying each user is obtained by the following formula:
y＝Wo*o+bo；
wherein O is ∈ Rm×nRepresenting the output of the LSTM model, m representing the input batch size, n representing the output dimension size of the LSTM model, Wo∈Rn×7Parameter matrix representing a fully connected network, boDenotes the bias of the network, y ∈ Rm×7Representing the final probability matrix;
normalizing the probability of each user by utilizing a Softmax function; calculating the probability of each output category by using the following formula to represent the possibility of belonging to each category, wherein the classification result with the maximum probability is selected:
Figure BDA0003196344520000053
in the formula, yiThe output value of the ith node of the LSTM network is represented, k represents the number of output nodes, namely the number of classified categories, and the output value in the multi-classification problem is converted into [0,1 ] through a Softmax function]The probability distribution problem between;
the loss function used by the model is a cross-entropy loss function, which is expressed as the following equation:
Figure BDA0003196344520000054
wherein m is the number of samples participating in the calculation of the loss function, generally the number of samples loaded in a batch by the data set loader, Θ is a parameter of the model, Y isiFor the true, i.e. tag, value of the data, yiAnd performing gradient regression on the network model according to the condition of the loss function to obtain a predicted value of the network until the training is finished when the loss function is converged.
The beneficial effects of the invention are as follows:
1. the method uses a CSI quotient and discrete wavelet transform combined denoising method to effectively remove noise information irrelevant to human walking activity in signals, and uses time-frequency analysis of discrete wavelet transform to mainly concentrate signal frequency on frequency caused by walking, so that the method is more beneficial to extraction of subsequent CSID information, and the accuracy of identity recognition is obviously improved.
2. The method provides the CSID as the basis of identity recognition, and the CSID is analyzed through experiments to contain the unique walking gait characteristics of the human. The system realizes the identification of people on the basis.
3. The method adopts deep learning technology to extract the characteristics and realizes identity recognition. And (3) selecting an LSTM model to train and test on the existing public data set by combining the time sequence characteristics of the CSID. The gait characteristic information of the walking of the person is fully extracted, so that the accuracy rate of the identity recognition reaches more than 90 percent.
Drawings
Fig. 1 is a flowchart of a pedestrian gait recognition method based on the channel state information quotient distance according to the invention.
Fig. 2 is a diagram of original amplitude phase information on the antenna 1 and the antenna 2 in the pedestrian gait recognition method based on the channel state information quotient distance.
Fig. 3 is a CSI denoising result diagram in the pedestrian gait recognition method based on the channel state information quotient distance of the present invention.
Fig. 4 is a schematic diagram of a subcarrier with an abnormal value in the pedestrian gait recognition method based on the channel state information quotient distance.
Fig. 5 is a schematic diagram of the subcarriers after the Hampel filtering and denoising in the pedestrian gait recognition method based on the channel state information quotient distance.
Fig. 6 is a flow chart of discrete wavelet transform process in the pedestrian gait recognition method based on the channel state information quotient distance.
Fig. 7 is a schematic diagram of sub-carriers for discrete wavelet transform denoising in the pedestrian gait recognition method based on the channel state information quotient distance.
Fig. 8 is a schematic diagram of CSID information on subcarriers of a pedestrian gait recognition method based on channel state information quotient distance according to the present invention.
Fig. 9 is a schematic diagram of a network structure in the pedestrian gait recognition method based on the channel state information quotient distance according to the invention.
Fig. 10 is a confusion matrix diagram after the user data is identified in the verification of the pedestrian gait recognition method based on the channel state information quotient distance according to the invention.
Fig. 11 is a graph of score, accuracy and recall of identification of user data in verification of the pedestrian gait recognition method based on the channel state information quotient distance according to the present invention.
Fig. 12 is a diagram showing the change of the identification accuracy rate with the number of people in the verification of the pedestrian gait recognition method based on the channel state information quotient distance.
Fig. 13 is a diagram showing the influence of discrete wavelet transform denoising on user identification accuracy in the verification of the pedestrian gait identification method based on the channel state information quotient distance.
Detailed Description
In order to make the purpose, technical solution and advantages of the present technical solution more clear, the present technical solution is further described in detail below with reference to specific embodiments. It should be understood that the description is intended to be exemplary only, and is not intended to limit the scope of the present teachings.
As shown in fig. 1 to fig. 9, the pedestrian gait recognition method based on the channel state information quotient distance proposed by the present embodiment is divided into four parts, and the flow thereof is as shown in fig. 1. The whole system framework mainly comprises four parts, namely data acquisition, signal preprocessing, CSID extraction and identity recognition. The data acquisition utilizes a WiFi transmitting terminal and two WiFi receiving terminals to acquire CSI data of people walking in a sensing area for model training and testing. And performing signal preprocessing on the acquired CSI data to perform related denoising processing, and eliminating irrelevant noise information which is contained in the CSI data and is caused by human walking so as to better lay the next place. The change information of the channel state information quotient in the complex plane contains the walking characteristics of people, and CSID (Euclidean distance from an initial point of the complex plane to each subsequent point) is proposed to represent the change of the channel state information quotient in the complex plane. In the identity recognition, the LSTM network model is used for feature extraction to further realize the identity recognition by combining the advantages of deep learning in the aspect of feature extraction.
In a data acquisition part, the method uses the existing public data set to test and evaluate an identity recognition system provided in a paper, selects CSI data on two vertical WiFi links to perform system test, and the data set contains CSI information acquired when 7 persons walk in 8 directions on 4 walking paths. According to the method, each piece of CSI data is subjected to data preprocessing, initial denoising is carried out through a two-antenna operator to obtain a CSI quotient, then amplitude quotient and phase difference information of the CSI quotient are further extracted, the amplitude quotient and the phase difference are further denoised through hampel filtering and discrete wavelet transformation, and the denoised amplitude quotient and the denoised phase difference are reconstructed again to obtain the denoised CSI quotient. CSID information of 30 subcarriers is extracted from the movement of CSI quotient in a complex plane, and finally a matrix with the size of 2300 x 30 is formed. Where 2300 represents the sample sequence (sample rate 1000, 2.3 seconds of data), 30 represents the number of subcarriers, and each column of the matrix represents the CSID information for one subcarrier. The 2300 x 30 matrix is then encapsulated with a label representing the true identity to form a sample in the data set. The resulting dataset contained a total of 5392 samples, and the dataset was scaled by approximately 8: 2, the training set and the test set are divided into 4321 training sets and 1071 test sets.
In the signal preprocessing part, noise caused by hardware and impulse noise in the signal are preliminarily removed in a CSI quotient mode on the two antennas. In order to further remove noise which is contained in the signal and is irrelevant to the walking information, amplitude quotient and phase difference information of the CSI quotient are extracted and are further denoised respectively, and then the denoised amplitude quotient and phase difference are subjected to signal inverse transformation to obtain the denoised CSI quotient information. In the signal preprocessing stage, denoising methods such as two antenna CSI quotients, hampel filtering, discrete wavelet transformation and the like on the same receiver are mainly used.
(1) And denoising the CSI quotient. As shown in fig. 2, which shows the amplitude and phase information of CSI received by antenna 1 and antenna 2, respectively, amplitude fluctuation caused by walking behavior can be observed through the amplitude information in the original CSI, but due to the influence of the device hardware itself and the environment, a large amount of noise information unrelated to walking behavior exists in the signal fluctuation. Meanwhile, it can be seen from the figure that the phases in the original CSI are randomly distributed in the range of [ -pi, pi ], and the information related to the walking behavior of the person cannot be directly obtained from the original phase information.
The CSI obtained by the two antennas is divided, the CSI is a complex information including amplitude and phase information of the signal, it can be known from the complex division that the amplitude of the CSI quotient becomes the ratio of the amplitudes of the two antennas, and the phase of the CSI quotient represents the difference between the phases of the two antennas. Fig. 4 shows the amplitude quotient and phase difference information in the CSI quotient. Compared with the original signal diagram, the noise in the amplitude is effectively eliminated, and the phase information which cannot be directly used originally becomes the usable phase difference information. As can be seen from fig. 3, walking activity has a significant influence on signal waveforms, that is, CSI quotient information obtained by quotient between two antennas includes information about signal changes caused by walking of a person.
(2) And (5) filtering and denoising by using hampel. Amplitude quotient and phase difference information are extracted from the CSI quotient, pulse noise in the amplitude information can be well removed by using the two antenna quotient, and meanwhile, the originally unavailable phase information is converted into the available phase difference information. However, when performing signal analysis, as shown in fig. 4, some abnormal noise still exists in the detected subcarrier waveform, and the abnormal noise is often caused by hardware defects and the existence of special noise in the environment.
Here, the system further utilizes a Hampel filter to remove outliers that are clearly present in the signal waveform. The Hampel filter has the main principle that the abnormal data value is removed through a moving average window, the average value of data is directly taken, and the purpose of removing abnormal noise is achieved through removing some points with larger square difference. From fig. 5, it can be seen that some data outliers existing in the amplitude quotient and the phase difference are effectively removed.
(3) And denoising by discrete wavelet transform. The most immediate benefit of using discrete wavelet transform is that wavelet decomposition can decompose signals at different scales, and different decomposition scales can be chosen according to actual needs. Often the low frequency information in many signals is of particular importance, which implies the main characteristics of the signal, while the high frequency content of the signal gives details or differences of the signal. The signal frequency corresponding to the human walking activity is mainly concentrated on low-frequency information, and the signal is decomposed and reconstructed through discrete wavelet transformation, so that the characteristic information generated by the walking activity in the signal can be better reserved.
The system further denoises amplitude quotient and phase difference information after the hampel filtering by adopting Discrete Wavelet Transform (DWT), and focuses signal waveforms on human walking behaviors to the maximum extent. The discrete wavelet transform enables multi-resolution analysis of discrete signals in both the time and frequency domains. As shown in fig. 6, when the original signal X with frequency F is decomposed by discrete wavelet transform, the original X signal is first filtered by low-pass and high-pass filters. Are decomposed into Approximation Coefficients of level 1 (ACs) via a low-pass filter and Detail Coefficients of level 1 (DCs) via a high-pass filter. ACs and DCs respectively correspond to the low-frequency part of 0-F/2 and the high-frequency part from F/2 to F of the original signal, and then the approximate coefficients obtained by each level of decomposition are subjected to iterative decomposition through the same decomposition process, so that more related wavelet coefficients with different frequency information and time scales can be obtained.
In the method, db4 wavelet is selected to carry out 3-level wavelet decomposition on the amplitude quotient and the phase difference of the selected subcarrier, so as to achieve the best denoising effect. As the sampling rate of the data set adopted in the method for collecting the CSI is 1000Hz, the Nyquist sampling theorem shows that the highest frequency of signals contained in the CSI data is 500Hz, the 3 rd-level approximate coefficient is reconstructed to obtain signals containing 0-62.5Hz, and the discrete wavelet transform can better remove signal noise which is irrelevant to the walking motion of a person in the waveforms, so that the denoised signal waveforms can be better focused on the signal waveform change caused by the walking of the person. For the collected walking CSI quotient data, fig. 7 shows the denoising effect of three-level wavelet approximation coefficient reconstruction of amplitude quotient and phase difference using db4 wavelet.
CSID extraction: in the previous section, a preprocessing process for CSI data is introduced, and in this section, change information of a CSID characterization denoised CSI quotient in a complex plane is proposed. In the CSI quotient model, there are mainly the following three properties.
(1) When the length of a reflection path of a moving object in the environment changes by a wavelength, the track of the CSI quotient on the complex plane is a perfect circle. When the length of the reflection path varies by a plurality of wavelengths, the circle of the CSI quotient on the complex plane is correspondingly rotated by the same number of turns.
(2) When the length of the reflected path varies by less than a wavelength, the CSI quotient forms an arc whose arc measure approximately matches the length variation of the reflected path.
(3) The rotation direction of the CSI quotient arc is closely related to the movement direction of the target in the Fresnel area.
Firstly, the method carries out relevant research on relevant properties of a CSI quotient aiming at human walking activities, the CSI quotient is obtained after signal preprocessing is carried out on collected CSI data, and then the change track of the CSI quotient on a complex plane is observed. In experiments, the CSI quotient is found to have no repeated circular motion in the complex plane in the complete circle, and the shape of the circle and the radius size are changed. This is because the human body is not an ideal source of reflection of the signal, and even if the relevant experimental variables are controlled, the leg movements and subtle movements of other parts of the body when the person is walking inevitably have some effect on the signal.
The gait characteristics of the human body movement comprise the motion characteristics of all parts of the human body when the human body walks, mainly the swing of arms, the stride step frequency information and the coordinated motion information of other parts of the human body. Because the walking behavior of each person is unique, the gait characteristics can be used as the basis for identifying the identity of the person. And generally, gait behaviors are difficult to imitate because the gait behaviors do not have unity, and the gait behaviors are characteristic information fusion of all parts of a human body.
The reason for the CSI quotient to move in the complex plane is that each dynamic part changes the length of a dynamic reflection path in the environment when the human body walks. The walking gait of each person is different, so that different walking modes of different persons can be analyzed and obtained to generate different influences on the change of the CSI quotient on the complex plane, and the change of the CSI quotient on the complex plane caused by the walking of the person contains unique walking gait information of the person, so that the movement information of the CSI quotient on the complex plane can be used as a basis for personal identification.
Through analyzing the data acquired in the real walking experiment, the movement of the CSI quotient in the complex plane can well catch the beginning and the end of the walking behavior of the human body. Mainly based on the following experimental observation, when a person is stationary, the environment only contains a static reflection path, the movement track of the CSI quotient in the complex plane is often only moved in a small range, when the person starts to walk, the environment has a dynamic reflection path, and the movement range of the CSI quotient in the complex plane is enlarged.
In order to quantitatively represent the analysis result of the movement of the CSI quotient on the complex plane, CSID is provided, and the CSID represents the distance information from an initial point to each subsequent point when the CSI quotient moves on the complex plane. The movement of the CSI quotient in the complex plane is quantified by the distance information from a point which continuously moves along with the time to the initial point.
CSID information acquired by one subcarrier in CSI data is shown in fig. 8. As can be seen from the figure, the CSID has the same characteristic of sensing the change of the surrounding environment as the CSI information, and is characterized by having stability in a static environment and certain sensitivity in a dynamic environment. When a person is at rest in the environment, i.e. when only static reflection paths exist in the environment, it can be seen from the figure that the fluctuation of the CSID is very smooth, and the fluctuation range of the CSID becomes significantly larger when the person starts walking, i.e. when dynamic reflection paths exist in the environment.
The CSID well quantifies the motion trajectory of the CSI quotient in the complex plane. A perfect peak in CSID represents the process of the CSI quotient moving in a circular-like trajectory in the complex plane, because in the case of determining an initial point, when the point starts to move in a circular-like motion, it is easy to think that the trajectory of the point is farther and farther from the initial point until it is closer to the initial point after forming a semicircle until it coincides with the initial point. The abscissa of the peak value in the graph represents the time when the point moves to half of a complete circle, and the change of the ordinate of the peak value in the graph is just caused by that the human body is not a perfect reflector and each dynamic part of the human body influences the dynamic reflection path in the environment. Meanwhile, in the time dimension, the number of wave crests contains information of the change speed of the length of the reflection path.
From the above analysis, the following conclusions can be drawn: the CSID provided by the method can be used for sensing the human body motion in the environment, and the time domain change of the CSID is caused by the dynamic change of the length of a transmitting path caused by each part when the human body walks. The CSID contains a large amount of human walking gait characteristics, and theoretically, the CSID can provide basis for identity recognition. The CSID provided by the system provides a new idea for gait recognition based on WiFi signals, and in the following work, the powerful feature extraction capability of deep learning is utilized to extract the walking gait features of people contained in the CSID, so that identity recognition is realized. Meanwhile, the good sensing capability of the CSID is simultaneously benefited by the effective preprocessing and reconstruction of the CSI quotient signals by the signal preprocessing of the system.
Model training: the track of the CSI quotient on the complex plane can reflect gait characteristics of walking of a person, the track on the CSI quotient on the complex plane is changed along with time, and CSID information extracted from a denoised CSI quotient complex matrix is a sequence changed along with time. The method is expected to extract gait features from the time sequence CSID sequence and identify, and the RNN particularly processes the time sequence in the deep learning field.
The human walking behavior is a natural continuous behavior, and the signal sequence generated by walking is often larger than 1s in the time dimension. The traditional RNN model has the problem of gradient disappearance, has limited extraction capability of long-term information, and cannot extract complete gait features under a time sequence. LSTM, a variant of RNN network, can learn long-term dependencies and is currently widely used for handling time-series sequences.
In order to capture all previous information in the sequence, including long-term and short-term information of walking behavior, the method uses a double-layer long-term and short-term memory neural network when extracting gait features. And (3) extracting the gait features of each user to be identified by using an LSTM unit structure instead of a neural unit structure in the traditional RNN model so as to realize robust identity identification. As shown in fig. 9, which shows the basic structure of the network, the neural network model will extract the gait features contained in the CSID through the LSTM unit in the hidden layer of the neural network for the input CSID sequence.
The CSID obtained by preprocessing the collected CSI data is used as the input of a network model, and the CSID sequence of 30 subcarriers in the CSI information is expressed as the following formula
Figure BDA0003196344520000131
Where i represents a subcarrier number, which includes thirty subcarriers in total, j represents time series information, here referred to as a sampling sequence in the CSI raw data, and CSID represents information of euclidean distance variation from a point on the CSI quotient complex plane to an initial point in the time series. One sample data in a dataset may be denoted as D { (X)iY), i ═ 1,2, 3., 30, where D includes CSID sequences extracted from 30 subcarriers in CSI data and their corresponding one identity tag. XiIndicating 30 CSID sequences obtained by processing 30 subcarrier information, and Y indicates a tag value representing identity information. In the method, the identification problem is actually creating a model which can be based on the input XiThe matrix predicts identity label Y.
For each hidden layer input, LSTM may set times t and i to [1,2 ═ m]Input of layers
Figure BDA0003196344520000141
Mapping as feature ftAs shown in formula:
Figure BDA0003196344520000142
wherein g represents an activation function, wherein,
Figure BDA0003196344520000143
representing input, containing current information of the sequence
Figure BDA0003196344520000144
Short-term information
Figure BDA0003196344520000145
And long term information
Figure BDA0003196344520000146
Where the dimension size is d-30, P denotes the weight matrix, and b is the offset.
The identification problem is actually a multi-classification problem, so the system performs matrix transformation on the output of the LSTM through the full connection layer, and the probability of identifying each user is obtained through the following formula.
y＝Wo*o+bo
O∈Rm×nRepresents the output of the LSTM model, m represents the input batch size (batch _ size), n represents the output dimension size of the LSTM model, Wo∈Rn×7Parameter matrix representing a fully connected network, boDenotes the bias of the network, y ∈ Rm×7Representing the final probability matrix.
The probability for each user is then normalized using the Softmax function. The essence of the Softmax function is that any real vector input is mapped to another real vector of the same dimension, each number in the mapped vector is between (0,1) and the sum of the additions is 1 (normalized). And calculating the probability of each output category by using the following formula to represent the possibility of belonging to each category, wherein the classification result is the maximum selected probability.
Figure BDA0003196344520000147
In the formula, yiThe output value of the ith node of the LSTM network is shown, and k represents the number of output nodes, namely the number of classified categories. The output value in the multi-classification problem can be converted into [0,1 ] by the Softmax function]The probability distribution problem between.
The loss function used by the model is a cross-entropy loss function, which is expressed as the following equation:
Figure BDA0003196344520000151
wherein m is ginsengThe number of samples calculated from the loss function, typically the number of samples loaded in a batch by the dataset loader, Θ being a parameter of the model, YiFor the true, i.e. tag, value of the data, yiIs a predicted value of the network. And the network model performs gradient regression according to the condition of the loss function until the training is finished when the loss function is converged.
It is noted that the deep learning model structure used in the system comprises only two layers of LSTM network structure, since the method performs relevant preprocessing on the raw data before data input. The CSI quotient preliminarily and effectively removes some impulse noises in the CSI original data by using a two-antenna quotient mode, and the signal-to-noise ratio of the original signal is greatly improved. And then extracting amplitude quotient and phase difference in the CSI quotient, and respectively performing abnormal noise removal and discrete wavelet denoising on the amplitude quotient and the phase difference, wherein due to the distance of the CSI quotient represented by the CSID at a complex plane point, an abnormal value can similarly cause an obvious abnormal value to appear in the CSID, and the time-frequency analysis characteristic based on the discrete wavelet transformation better reserves the signal frequency influenced by human walking, so that the noise of the unrelated human walking motion in the signal is further removed. And performing signal reconstruction on the amplitude quotient and the phase difference containing the human walking information to obtain a denoised CSI quotient, then extracting the CSID from the CSI quotient, and reserving the signal characteristics generated by human walking in the finally obtained CSID to the maximum extent. The two-layer network structure adopted by the method can be used for extracting the gait features contained in the CSID, and more complex network models often represent more time cost. The distance information represented by the CSID is changed along with time, and the traditional method for extracting the statistical features cannot fully utilize the features in a long-time sequence and can destroy some original feature information in the signal.
Based on the analysis, the method adopts the deep learning model only comprising two layers of LSTMs to carry out feature extraction and identify, achieves the average identification precision of 97% when identifying the identities of 4 persons, and achieves the average identification precision of 90% when identifying the identities of 7 persons.
The invention provides a pedestrian gait recognition method based on channel state information quotient distance, which mainly comprises the following steps:
(1) denoising method combining CSI quotient with discrete wavelet transform
The CSI data directly acquired from the WiFi equipment contain a large amount of noise and environmental noise caused by the equipment, so that the CSI cannot be directly applied to identity sensing work. Amplitude quotient and phase difference information of the CSI quotient are extracted, wavelet decomposition and reconstruction layer number are determined by utilizing time-frequency analysis characteristics of discrete wavelet transformation and combining sampling frequency of data acquisition, and waveform of signals is better focused on human walking behaviors.
(2) Extraction of CSID information
Most of the previous work of identifying identity by utilizing WiFi signals is directly sensing work based on CSI amplitude or phase information, and in the method, European distances (CSID) from an initial point of a complex plane to later points of a CSI quotient are provided as the basis of identity sensing, and the reason that the CSID comprises walking gait characteristics is explained in detail.
(3) Feature extraction and identity recognition for CSID based on LSTM neural network
The method uses the LSTM neural network to extract gait features contained in the CSID so as to realize identity recognition. As CSID is distance information which changes along with time and is time sequence information, as is known, an RNN model has natural advantages in processing time sequence signals, but the problems of gradient disappearance, gradient explosion and the like are easily generated when long-term information is subjected to feature extraction, continuous signals generated by walking behaviors usually last for more than 1s, the long-term gait features of human walking are extracted when the gait features are extracted, and the problems of gradient disappearance and gradient explosion can be effectively avoided by using LSTM as a variant of the RNN, so that the method adopts the LSTM network to extract the gait features and identify the identity.
Example 1
The feasibility and the effect of the process are explained in detail by the examples below.
1. Accuracy of identification
Seven persons of data were selected from the data set used for test analysis, fig. 10 shows the confusion matrix after identifying the data of 7 users, and fig. 11 shows the corresponding accuracy, recall and F1 scores.
According to the confusion matrix, the identity recognition is carried out on 7 users, each row in the confusion matrix represents a real identity label of data, each column represents a predicted identity label of the data, and the average accuracy of the identity recognition of seven users reaches 90.66%. As can be seen from fig. 1, the recognition confusion of the users 5, 6, and 7 is high, and when the user 7 is recognized, 13% of samples of the user 7 are recognized as the user 6, which indicates that the user 7 has a similar body type to the user 6, and the difference of the walking posture is small, so that there is a false recognition phenomenon during the recognition. The identification accuracy of the user 3 reaches 100%, because the walking gait characteristics of the user 3 are obviously different from those of the other 6 people, the feasibility of the gait identification scheme principle adopted by the method is further verified.
2. Influence of training population on experimental results
In this section, how the accuracy rate of the method used by the system changes with the number of recognized people is mainly analyzed, the method uses the data of 7 users in the public data set to perform testing, and the existence of potential intruders is not considered in the testing, namely, users which do not belong to the 7 users do not exist.
In the experiment process, the identification accuracy rate when the number of the target people is increased from 2 to 7 is tested and analyzed, as shown in fig. 12, experiments show that the accuracy of identity identification is slightly reduced along with the increase of the number of the target people, and similar gait features exist in the feature clustering process in the gait feature space along with the increase of the number of the classified people.
As can be seen from fig. 12, the user identification model proposed by the system maintains the identification accuracy above 97% when the number of people is not more than 4, and reflects that the extracted gait features are different in identification of different people on the side. When the number of the recognized target people is 5, the recognition accuracy rate is reduced to 94%, and when the number of the recognized people is increased to 7, the identity recognition accuracy rate can still reach 90%. The recognition accuracy rate of the system reaches 97% when the number of the recognized people is four, which shows the application potential of the method in the intelligent home scene with only a few users.
3. Influence of denoising method on experimental result
CSID is proposed based on the relevant properties of CSI quotient in complex plane motion to perform gait recognition on people in the environment. The CSID is closely related to the CSI quotient, and noise information which is contained in the CSI quotient and is irrelevant to human walking activities is further eliminated in a signal preprocessing stage by combining with the time-frequency analysis characteristic of discrete wavelet transformation. In order to verify the effectiveness of the denoising method provided by the system, CSID is extracted from a CSI quotient which is not subjected to discrete wavelet denoising and is used as an original data set for testing, and other experimental parameters are kept unchanged.
As shown in fig. 13, it can be seen from the figure that, compared with the identification by the CSID extracted from the denoised CSI quotient, the identification accuracy of the other six persons is affected to different degrees except that the identification accuracy of the user 4 is not affected, which indicates that the method provided for denoising the amplitude quotient and the phase difference in the CSI quotient by using the discrete wavelet transform, and then reconstructing the CSI quotient effectively retains the walking gait feature while removing the noise, thereby improving the accuracy of the identification.
The foregoing is only a preferred embodiment of the present invention, and many variations in the specific embodiments and applications of the invention may be made by those skilled in the art without departing from the spirit of the invention, which falls within the scope of the claims of this patent.

Claims (7)

1. The pedestrian gait recognition method based on the channel state information quotient distance is characterized in that: the method comprises the following steps:
step 1, data acquisition: collecting CSI data of people walking in a sensing area by using one WiFi transmitting terminal and two WiFi receiving terminals for model training and testing;
step 2, signal preprocessing: carrying out relevant denoising processing on the collected CSI data, and eliminating irrelevant noise information which is contained in the CSI data and is related to human walking;
step 3, CSID extraction: the change information of the channel state information quotient on the complex plane contains the gait characteristics of human walking, and CSID is proposed to represent the change of the channel state information quotient on the complex plane;
step 4, identity recognition: and combining the advantages of deep learning in the aspect of feature extraction, and using the LSTM network model to extract features so as to realize identity recognition.
2. The pedestrian gait recognition method based on the channel state information quotient distance as claimed in claim 1, characterized in that: in the step 1, selecting CSI data on two vertical WiFi links to carry out system test, firstly carrying out data preprocessing on each piece of CSI data, carrying out preliminary denoising through two antenna operators to obtain a CSI quotient, then further extracting amplitude quotient and phase difference information of the CSI quotient, further denoising the amplitude quotient and the phase difference by using hampel filtering and discrete wavelet transform, and reconstructing the denoised amplitude quotient and phase difference again to obtain the denoised CSI quotient; and extracting the CSID information containing the subcarriers from the movement of the CSI quotient in the complex plane, and encapsulating the CSID information with the label representing the real identity to form a sample in the data set.
3. The pedestrian gait recognition method based on the channel state information quotient distance as claimed in claim 1, characterized in that: in step 2, the amplitude quotient and the phase difference information of the CSI quotient are extracted and further denoised respectively, and then the denoised amplitude quotient and the denoised phase difference are subjected to signal inverse transformation to obtain the denoised CSI quotient information.
4. The pedestrian gait recognition method based on the channel state information quotient distance as claimed in claim 3, characterized in that: in step 2, comprising
Step 21, CSI quotient denoising: the CSI obtained by the two antennas is divided, the CSI is complex information which comprises amplitude and phase information of signals, the complex division operation can know that the amplitude of a CSI quotient is changed into the ratio of the amplitudes of the two antennas, the phase of the CSI quotient represents the difference of the phases of the two antennas, and the CSI quotient information obtained by the two antennas contains signal change information caused by walking of people;
step 22, Hampel filtering and denoising: removing abnormal values obviously existing in the signal waveform by using a Hampel filter;
step 23, denoising through discrete wavelet transform: the system further denoises amplitude quotient and phase difference information after the hampel filtering by adopting Discrete Wavelet Transform (DWT), and focuses signal waveforms on human walking behaviors to the maximum extent; when an original signal X with the frequency of F is decomposed by utilizing discrete wavelet transform, the original signal X is filtered by a low-pass filter and a high-pass filter; the wavelet coefficients are decomposed into 1 st-level approximate coefficients ACs through a low-pass filter, the 1 st-level detail coefficients DCs are decomposed through a high-pass filter, the approximate coefficients ACs and the detail coefficients DCs respectively correspond to the low-frequency part of 0-F/2 and the high-frequency part from F/2 to F of an original signal, and then the approximate coefficients obtained through each level of decomposition are subjected to iterative decomposition through the same decomposition process, so that more related wavelet coefficients with different frequency information and time scales can be obtained.
5. The pedestrian gait recognition method based on the channel state information quotient distance as claimed in claim 4, characterized in that: in step 23, a db4 wavelet is selected to perform 3-level wavelet decomposition on the amplitude quotient and the phase difference of the selected sub-carriers; the sampling rate of the CSI collection is 1000Hz, the highest frequency of signals contained in CSI data is 500Hz, the 3 rd-level approximation coefficient is reconstructed to obtain signals containing 0-62.5Hz, and the discrete wavelet transformation can better remove signal noise irrelevant to human walking movement in waveforms, so that the denoised signal waveforms are better focused on signal waveform changes caused by human walking.
6. The pedestrian gait recognition method based on the channel state information quotient distance as claimed in claim 4, characterized in that: in step 4, a double-layer long-and-short time memory neural network is adopted, the gait features of each user to be identified are extracted by using an LSTM unit structure so as to realize robust identity identification, and the input CSID sequence is extracted by a neural network model through the LSTM unit in a neural network hidden layer.
7. The pedestrian gait recognition method based on the channel state information quotient distance as claimed in claim 6, characterized in that: CSID obtained by preprocessing the collected CSI data is used as the input of a network model, and CSID sequences of 30 subcarriers in the CSI information are expressed as the following formula
Figure FDA0003196344510000031
Wherein i represents a subcarrier number, and j represents time series information; CSID indicates information of change of euclidean distance from a point on the CSI quotient complex plane to an initial point in a time series, and one sample data in a data set may be represented as D { (X)iY) }, i ═ 1,2, 3.., 30, where D includes CSID sequences extracted from 30 subcarriers in CSI data and a corresponding identity tag; xiThe method comprises the steps of representing 30 CSID sequences obtained by processing 30 subcarrier information, wherein Y represents a label value and represents identity information; identification questions are created models that can be based on an input XiMatrix is used for predicting identity label Y;
for each hidden layer input, LSTM may set times t and i to [1,2 ═ m]Input of layers
Figure FDA0003196344510000032
Mapping as feature ftAs shown in formula:
Figure FDA0003196344510000033
wherein g represents an activation function, wherein,
Figure FDA0003196344510000034
representing input, containing current information of the sequence
Figure FDA0003196344510000035
Short-term information
Figure FDA0003196344510000036
And long term information
Figure FDA0003196344510000037
Wherein the dimension size is d-30, P represents a weight matrix, and b is an offset;
the output of the LSTM is matrix transformed by the full connectivity layer, and the probability of identifying each user is obtained by the following formula:
y＝Wo*o+bo；
wherein O is ∈ Rm×nRepresenting the output of the LSTM model, m representing the input batch size, n representing the output dimension size of the LSTM model, Wo∈Rn×7Parameter matrix representing a fully connected network, boDenotes the bias of the network, y ∈ Rm×7Representing the final probability matrix;
normalizing the probability of each user by utilizing a Softmax function; calculating the probability of each output category by using the following formula to represent the possibility of belonging to each category, wherein the classification result with the maximum probability is selected:
Figure FDA0003196344510000041
in the formula, yiThe output value of the ith node of the LSTM network is represented, k represents the number of output nodes, namely the number of classified categories, and the output value in the multi-classification problem is converted into [0,1 ] through a Softmax function]The probability distribution problem between;
the loss function used by the model is a cross-entropy loss function, which is expressed as the following equation:
Figure FDA0003196344510000042
wherein m is the number of samples participating in the calculation of the loss function, generally the number of samples loaded in a batch by the data set loader, Θ is a parameter of the model, Y isiFor the true, i.e. tag, value of the data, yiAnd performing gradient regression on the network model according to the condition of the loss function to obtain a predicted value of the network until the training is finished when the loss function is converged.
CN202110892189.XA 2021-08-04 2021-08-04 Pedestrian gait recognition method based on channel state information quotient distance Active CN113609977B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110892189.XA CN113609977B (en) 2021-08-04 2021-08-04 Pedestrian gait recognition method based on channel state information quotient distance

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110892189.XA CN113609977B (en) 2021-08-04 2021-08-04 Pedestrian gait recognition method based on channel state information quotient distance

Publications (2)

Publication Number Publication Date
CN113609977A true CN113609977A (en) 2021-11-05
CN113609977B CN113609977B (en) 2023-10-17

Family

ID=78306853

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110892189.XA Active CN113609977B (en) 2021-08-04 2021-08-04 Pedestrian gait recognition method based on channel state information quotient distance

Country Status (1)

Country Link
CN (1) CN113609977B (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114707563A (en) * 2022-06-07 2022-07-05 湖南工商大学 Human body action recognition method based on Wi-Fi channel state information
CN114840834A (en) * 2022-04-14 2022-08-02 浙江大学 Implicit identity authentication method based on gait characteristics
CN115119146A (en) * 2022-05-19 2022-09-27 大连理工大学 Isolated person identity authentication method and device based on wifi signal
CN116597521A (en) * 2023-07-19 2023-08-15 山东科技大学 Human track and trunk activity recognition method and system
CN116821778A (en) * 2023-08-30 2023-09-29 之江实验室 Gait recognition method and device based on WIFI signals and readable storage medium

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20170119268A (en) * 2016-04-22 2017-10-26 권순태 Information exchange system between a plurality of terminals.
US20190087654A1 (en) * 2017-09-15 2019-03-21 Huazhong University Of Science And Technology Method and system for csi-based fine-grained gesture recognition
CN110288018A (en) * 2019-06-24 2019-09-27 桂林电子科技大学 A kind of WiFi personal identification method merging deep learning model
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

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20170119268A (en) * 2016-04-22 2017-10-26 권순태 Information exchange system between a plurality of terminals.
US20190087654A1 (en) * 2017-09-15 2019-03-21 Huazhong University Of Science And Technology Method and system for csi-based fine-grained gesture recognition
CN110288018A (en) * 2019-06-24 2019-09-27 桂林电子科技大学 A kind of WiFi personal identification method merging deep learning model
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

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
朱奕鸿;李嘉成;: "信道状态信息在动作识别中的应用与异常分析", 信息网络安全, no. 02 *

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114840834A (en) * 2022-04-14 2022-08-02 浙江大学 Implicit identity authentication method based on gait characteristics
CN114840834B (en) * 2022-04-14 2024-06-11 浙江大学 Implicit identity authentication method based on gait characteristics
CN115119146A (en) * 2022-05-19 2022-09-27 大连理工大学 Isolated person identity authentication method and device based on wifi signal
CN114707563A (en) * 2022-06-07 2022-07-05 湖南工商大学 Human body action recognition method based on Wi-Fi channel state information
CN116597521A (en) * 2023-07-19 2023-08-15 山东科技大学 Human track and trunk activity recognition method and system
CN116597521B (en) * 2023-07-19 2023-09-19 山东科技大学 Human track and trunk activity recognition method and system
CN116821778A (en) * 2023-08-30 2023-09-29 之江实验室 Gait recognition method and device based on WIFI signals and readable storage medium
CN116821778B (en) * 2023-08-30 2024-01-09 之江实验室 Gait recognition method and device based on WIFI signals and readable storage medium

Also Published As

Publication number Publication date
CN113609977B (en) 2023-10-17

Similar Documents

Publication Publication Date Title
CN113609977B (en) Pedestrian gait recognition method based on channel state information quotient distance
Cao et al. Radar‐ID: human identification based on radar micro‐Doppler signatures using deep convolutional neural networks
CN106899968B (en) Active non-contact identity authentication method based on WiFi channel state information
CN103093234B (en) Based on the personal identification method of ground reaction force during walking
CN104573458A (en) Identity recognition method, device and system based on electrocardiogram signals
Al-Naffakh et al. Unobtrusive gait recognition using smartwatches
CN110647788B (en) Human daily behavior classification method based on micro-Doppler characteristics
CN110929242B (en) Method and system for carrying out attitude-independent continuous user authentication based on wireless signals
Zhao et al. Trustworthy authorization method for security in Industrial Internet of Things
Avola et al. Person re-identification through Wi-Fi extracted radio biometric signatures
Zhu et al. Fingergan: a constrained fingerprint generation scheme for latent fingerprint enhancement
Andrews et al. Biometric authentication and stationary detection of human subjects by deep learning of passive infrared (PIR) sensor data
CN114578963B (en) Electroencephalogram identity recognition method based on feature visualization and multi-mode fusion
Sarin et al. Cnn-based multimodal touchless biometric recognition system using gait and speech
CN106650685B (en) Identity recognition method and device based on electrocardiogram signal
CN113742669A (en) User authentication method based on twin network
CN115905945A (en) Pedestrian action recognition method, device, equipment and storage medium
Anchal et al. GMM-UBM based person verification using footfall signatures for smart home applications
Chockaian et al. New approach for identifying hereditary relation using primary fingerprint patterns
CN115035608A (en) Living body detection method, device, equipment and system
Choudhary et al. An Audio-Seismic Fusion Framework for Human Activity Recognition in an Outdoor Environment
Raghavendra Sparse representation for accurate person recognition using hand vein biometrics
CN114840834A (en) Implicit identity authentication method based on gait characteristics
Lathika et al. Artificial neural network based multimodal biometrics recognition system
Mohan et al. A nature‐inspired meta‐heuristic paradigm for person identification using multimodal biometrics

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