CN109766951A - A kind of WiFi gesture identification based on time-frequency statistical property - Google Patents
A kind of WiFi gesture identification based on time-frequency statistical property Download PDFInfo
- Publication number
- CN109766951A CN109766951A CN201910048068.XA CN201910048068A CN109766951A CN 109766951 A CN109766951 A CN 109766951A CN 201910048068 A CN201910048068 A CN 201910048068A CN 109766951 A CN109766951 A CN 109766951A
- Authority
- CN
- China
- Prior art keywords
- time
- gesture
- frequency
- csi
- matrix
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Landscapes
- User Interface Of Digital Computer (AREA)
Abstract
The invention discloses a kind of WiFi gesture identification methods based on time-frequency statistical property.First, gesture data is received using Intel5300 network interface card, extract channel state information (Channel State Information, CSI) amplitude data and the mathematical model of CSI signal amplitude Yu dynamic communication change in path length is constructed, it was demonstrated that CSI amplitude is used for the validity of gesture identification；Secondly, pre-processing to CSI amplitude data by low-pass filter, interference brought by ambient noise such as random noise is reduced；Then, the amplitude signal after denoising is subjected to dimensionality reduction by singular value decomposition (Singular Value Decomposition, SVD) algorithm, removes the redundancy of data, reduce time overhead cost；Then, time-frequency characteristics are extracted the processing with feature normalization by statistical nature and obtain the statistical property that can be used for classifying by the time-frequency characteristics that signal is extracted by Short Time Fourier Transform (Short-Time Fourier Transform, STFT)；Classification judgement is carried out to gesture with the sorting algorithm for k- neighbouring (k-Nearest Neighbor, kNN) finally, utilizing.The present invention can effectively classify and identify gesture feature, solve the problems, such as the identification under complex environment indoors to gesture.
Description
Technical field
The invention belongs to fields of communication technology, and in particular in a kind of WiFi gesture identification method of time-frequency statistical property.
Background technique
With the rapid development of 21st century science and technology and universal, the human-computer interaction technology (Human- of computer
Computer Interaction, HCI) have become the objects of numerous state keys concern and research, so-called human-computer interaction,
Refer to and goes to complete by the preassigned interactive mode such as computer hardware, behavior act, sound between user and computer equipment
Specified task, to generate the process of information exchange.Nowadays, the interacting activity between people and computer is more and more closer, hands over
Mutual mode affects interactive complexity and interactive efficiency.Traditional man-machine interaction mode is mouse, keyboard, handle, data hand
The additional input devices such as set, do not meet the exchange habit of mankind's routine and there are one in terms of flexibility and the accurate control of realization
Fixed limitation.With the development of technology, the novel interactive mode such as vision, voice, tactile, power feel has become a hot topic of research, and
Essential characteristic one of of the gesture as vision, is most convenient and natural interactive mode.
Common 3 class gesture recognition systems are respectively gesture recognition system based on wearable sensing equipment, are based at present
The gesture recognition system of computer vision, the gesture recognition system based on radiofrequency signal.Wherein wearable sensing equipment is started to walk very
It is early, it is highly developed so far, but need to wear such as data glove, armlet, acceleration transducer sensing equipment and obtain accordingly
Parameter, cause the inconvenience of user；System based on computer vision obtains user gesture behavior using camera, by hand
The problem of gesture identification problem is converted to image procossing carries out attitude parameter to the key position information of detection target, thus complete
Detection at target with track and identify, but can only be operated under the more sufficient environment of light, under night and dim scene
Discrimination is low, and popularization is poor.
Summary of the invention
The object of the present invention is to provide a kind of WiFi gesture identification methods of time-frequency statistical property, can be effectively to gesture behavior
Carry out classification judgement.
WiFi gesture identification of the present invention based on time-frequency statistical property, specifically includes the following steps:
Step 1: receiving CSI data, and based on CSI data building CSI amplitude and dynamic communication change in path length
Mathematical model；
Step 2: pre-processing using low-pass filter to the CSI amplitude in the mathematical model, high frequency is filtered out
Noise；
Step 3: carrying out dimension-reduction treatment to the CSI amplitude after pretreatment using singular value decomposition algorithm；
Step 4: extracting the Time-Frequency Information of signal using Short Time Fourier Transform, every kind of behavior is extracted based on time-frequency characteristic
Statistical property；
Step 5: will be under eigentransformation to unified scale by min-max standardization；
Step 6: being divided using the sorting algorithm by taking kNN algorithm as an example the signal time-frequency statistical property after standardization
Class judgement.
2, CSI data are received in the step 1, and based on CSI data building CSI amplitude and dynamic communication path
The mathematical model of length variation, comprising:
2a, received CSI data, the CSI data include static propagation path PsWith dynamic communication path PdThe sum of, the
The path responses that j subcarrier is superimposed at t moment receiver are as follows:
2b, hypothesis are in kth propagation path with constant speed υ within actuation timekIt changes, target direction of motion
Angle between direct path isThe path initial length is dk(0), then t moment dynamic route is
2c, in conjunction with available j-th of the subcarrier of 2a and 2b CSI amplitude square | H (fj′,t)2Comprising a series of normal
The factor and cosine function are measured, cosine function is the function for portraying the transformation of dynamic communication path length, and formula is as follows:
3, the WiFi gesture identification method according to step 1, singular value decomposition algorithm is to the CSI width after pretreatment
Value carries out dimension-reduction treatment, removes the redundancy of data, reduces time overhead cost, comprising the following steps:
3a, CSI amplitude are made of 30 sub- carrier waves, are denoted as
WhereinIndicate the gesture motion duration, | H (fi, t) | indicate i-th of sub-carrier signal t moment amplitude,
First by matrix centralization；
3b, the matrix H after centralization is calculatedcSVD decompose；
Hc=U*S*VT
3c, preceding r non-zero singular value in S is taken, restores matrix H to greatest extentc；
Hc'=Hc*V(:,1:r)T
Hc' arrived for dimensionality reductionThe gesture behavioural characteristic of dimension.
4, in the step 4, comprising the following steps:
4a, window function is set as hamming window, window size 64, Fourier transformation points are 64, and windows overlay points are
52；
4b, to Hc' matrix every a line calculate gesture behavior time-frequency characteristics, can make every kind of behavior obtain size correspond to
Time-frequency characteristics matrix of consequence S.
4c, the statistical property for calculating the every a line of time-frequency matrix of consequence, constitute the vector of a r × length (S).Assuming that depositing
In M behavior, then time-frequency matrix of consequence forms a 6 × r of M row × length (S) column statistical matrix Hs, 6 indicate time-frequency knot
Six kinds of statistical properties of fruit are mean value, standard deviation, interquartile range, the 0.5 of probability distribution, 0.683 and lower point of 0.95 respectively
Digit.Wherein HsIn every a line indicate all features of gesture behavior a kind of, and each column indicate certain of all gesture behaviors
A kind of statistical property.
5, in the step 5, comprising the following steps:
5a, time-frequency statistical nature is transformed under unified scale
Wherein, Hs' it is time-frequency statistical nature matrix after data normalization.
The invention has the following advantages that the invention firstly uses Intel5300 network interface cards to receive CSI signal, CSI letter is extracted
Number amplitude information, construct CSI amplitude information and dynamic communication change in path length mathematical model, it was demonstrated that CSI amplitude information
Validity as gesture identification data；Aiming at the problem that CSI amplitude data redundancy, calculated using singular value decomposition (SVD) dimensionality reduction
The dimensionality reduction to data different dimensions can be achieved in method on the basis of not losing main feature information；Have between different gestures
The problem of different characteristic, proposes the time-frequency characteristics that gesture behavior is extracted using Short Time Fourier Transform (STFT)；Due to gesture speed
Time-frequency characteristics matrix dimensionality caused by degree, time etc. are inconsistent is inconsistent, and the sorting algorithm by taking kNN algorithm as an example is utilized to complete not
Classification with gesture is adjudicated.
Detailed description of the invention
Fig. 1 is present system block diagram；
Fig. 2 is CSI communication environments figure；
Fig. 3 is test scene schematic diagram；
Fig. 4 is the result figure that different gestures utilizes the test of this system method；
Fig. 5 is the confusion matrix figure of different gesture identification precision.
Specific embodiment
The present invention will be further explained below with reference to the attached drawings.
It is the WiFi gesture recognition system block diagram based on time-frequency statistical property as shown in Figure 1, Fig. 2 is CSI under indoor environment
Communication environments schematic diagram, it is whole specifically includes the following steps:
Step 1: receiving CSI data, and the number based on CSI data building CSI amplitude and dynamic communication change in path length
Learn model, it was demonstrated that CSI amplitude is used for the validity of gesture identification.The CSI data received include static propagation path PsWith it is dynamic
State propagation path PdThe sum of, then the path responses that j-th of subcarrier is superimposed at t moment receiver are as follows:
It is assumed that in kth propagation path with constant speed υ within actuation timekChange, target direction of motion with
Angle between direct path isThe path initial length is dk(0), then t moment dynamic route is
Then the CSI amplitude of j-th of subcarrier square is
|H(fj', t) |2Comprising a series of constant factors and cosine function, cosine function is to portray dynamic communication path length
The function of transformation is spent, dynamic communication change in path length is related with the transformation of target frequency, therefore amplitude can be used as identification gesture
Characteristic information；
Step 2: pre-processing using low-pass filter to the CSI amplitude in the mathematical model, high frequency is filtered out
Noise；
Step 3: carrying out dimension-reduction treatment to the CSI amplitude after pretreatment using singular value decomposition algorithm, data are removed
Redundancy reduces time overhead cost, and specific implementation step is as follows:
The CSI amplitude data received is denoted as
WhereinIndicate the gesture motion duration, | H (fi, t) | indicate i-th of sub-carrier signal t moment amplitude,
First by matrix centralization
To the matrix H after centralizationcSVD decompose
Hc=U*S*VT
Preceding r non-zero singular value is taken in S to restore matrix H to greatest extent according to right singular matrixc；
Hc'=Hc*V(:,1:r)T
Hc' arrived for dimensionality reductionThe gesture behavioural characteristic of dimension.
Step 4: setting window function as hamming window, window size 64, Fourier transformation points are 64, windows overlay point
Number is 52；To Hc' matrix every a line calculate gesture behavior time-frequency characteristics, can make every kind of behavior obtain size it is corresponding when
Frequency characteristic results matrix S.
The statistical property for calculating the every a line of time-frequency matrix of consequence, constitutes the vector of a r × length (S).Assuming that there are M
A behavior, then time-frequency matrix of consequence forms a 6 × r of M row × length (S) column statistical matrix Hs, 6 indicate time-frequency result
Six kinds of statistical properties, be mean value, standard deviation, interquartile range, the 0.5 of probability distribution, 0.683 and lower point of 0.95 respectively
Digit.Wherein HsIn every a line indicate all features of gesture behavior a kind of, and each column indicate certain of all gesture behaviors
A kind of statistical property.Time-frequency statistical nature is transformed under unified scale
Hs' it is time-frequency statistical nature matrix after data normalization.
Step 5: mutually being existed together in the behavior of real-time testing gesture using such as preceding step 1 to the step of step 4
Reason extracts the time-frequency statistical nature matrix S' of real-time testing gesture.
Step 6: utilize sorting algorithm by taking kNN algorithm as an example according to online with off-lined signal time-frequency statistical property matrix
S' and S carries out classification judgement.
Test environment map of the invention is as shown in figure 3, environment size is 7.7m × 9m, between receiver and transmitter apart
2.0m.Five kinds of gestures of people are acquired in experiment, are lift hand respectively, are pushed and pulled, wave, drawing a circle, to the right.It is established in above-mentioned environment
Gesture database, amounts to 250 groups of behaviors by every kind 50 groups of behavior.Experiment is the frequency in 5.2G, bandwidth 40MHz, receives frequency
It is tested in the case where being 149 for 1000Hz, channel.
In order to verify reliability of the invention, Fig. 4 is the measured gesture court verdict figure of experiment.Fig. 5 is various actions
Confusion matrix figure, can be obtained by result figure, indoors under complex environment, the present invention be averaged accuracy of identification be 83.8%.
Claims (3)
1. a kind of WiFi gesture identification method based on time-frequency statistical property, which is characterized in that the described method comprises the following steps:
Step 1: receiving CSI data, and the number based on CSI data building CSI amplitude and dynamic communication change in path length
Learn model；
Step 2: being pre-processed using low-pass filter to the CSI amplitude in the mathematical model, filters out high frequency and make an uproar
Sound；
Step 3: carrying out dimension-reduction treatment to the CSI amplitude after pretreatment using singular value decomposition algorithm；
Step 4: extracting the Time-Frequency Information of signal using Short Time Fourier Transform, the system of every kind of behavior is extracted based on time-frequency characteristic
Count characteristic；
Step 5: will be under eigentransformation to unified scale by min-max standardization；
Sentence Step 6: carrying out classification to the signal time-frequency statistical property after standardization using the sorting algorithm by taking kNN algorithm as an example
Certainly.
2. WiFi gesture identification method according to claim 1, which is characterized in that the singular value decomposition algorithm is to pre- place
CSI amplitude after reason carries out dimension-reduction treatment, removes the redundancy of data, reduces time overhead cost, comprising the following steps:
2a, CSI amplitude are made of 30 sub- carrier waves, are denoted as
WhereinIndicate the gesture motion duration, | H (fi, t) | indicate i-th of sub-carrier signal t moment amplitude, first
By matrix centralization；
2b, the matrix H after centralization is calculatedcSVD decompose；
Hc=U*S*VT
2c, preceding r non-zero singular value in S is taken, restores matrix H to greatest extentc；
Hc'=Hc* V (:, 1:r)T
Hc' arrived for dimensionality reductionThe gesture behavioural characteristic of dimension.
3. -2 described in any item WiFi gesture identification methods according to claim 1, which is characterized in that in the step 4, packet
Include following steps:
3a, window function is set as hamming window, window size 64, Fourier transformation points are 64, and windows overlay points are 52；
3b, to Hc' matrix every a line calculate gesture behavior time-frequency characteristics, every kind of behavior can be made to obtain corresponding time-frequency characteristics
Matrix of consequence S.
3c, the statistical property for calculating the every a line of time-frequency matrix of consequence S, constitute the vector of a r × length (S).Assuming that there are M
A behavior, then time-frequency matrix of consequence forms a 6 × r of M row × length (S) column statistical matrix Hs, 6 indicate time-frequency result
Six kinds of statistical properties, be respectively mean value, standard deviation, interquartile range, the 0.5 of probability distribution, 0.683 and 0.95 lower quartile
Number.Wherein HsIn every a line indicate all features of gesture behavior a kind of, and each column indicate a certain of all gesture behaviors
Kind statistical property.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910048068.XA CN109766951A (en) | 2019-01-18 | 2019-01-18 | A kind of WiFi gesture identification based on time-frequency statistical property |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910048068.XA CN109766951A (en) | 2019-01-18 | 2019-01-18 | A kind of WiFi gesture identification based on time-frequency statistical property |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109766951A true CN109766951A (en) | 2019-05-17 |
Family
ID=66454189
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910048068.XA Pending CN109766951A (en) | 2019-01-18 | 2019-01-18 | A kind of WiFi gesture identification based on time-frequency statistical property |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109766951A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113238659A (en) * | 2021-06-29 | 2021-08-10 | 中国科学技术大学 | Real-time behavior identification method and system based on WIFI signal |
| CN113449587A (en) * | 2021-04-30 | 2021-09-28 | 北京邮电大学 | Human behavior recognition and identity authentication method and device and electronic equipment |
| CN114764580A (en) * | 2022-06-15 | 2022-07-19 | 湖南工商大学 | Real-time human body gesture recognition method based on no-wearing equipment |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104951757A (en) * | 2015-06-10 | 2015-09-30 | 南京大学 | Action detecting and identifying method based on radio signals |
| US20160162027A1 (en) * | 2012-06-14 | 2016-06-09 | Immersion Corporation | Haptic effect conversion system using granular synthesis |
| CN105807935A (en) * | 2016-04-01 | 2016-07-27 | 中国科学技术大学苏州研究院 | Gesture control man-machine interactive system based on WiFi |
| CN106295684A (en) * | 2016-08-02 | 2017-01-04 | 清华大学 | A kind of the most continuous based on micro-Doppler feature/discontinuous gesture recognition methods |
| CN106604394A (en) * | 2016-12-28 | 2017-04-26 | 南京航空航天大学 | CSI-based indoor human body motion speed judgment model |
| CN106658703A (en) * | 2016-10-12 | 2017-05-10 | 南京邮电大学 | Cosine similarity based RSS (Received Signal Strength) detection difference compensation method |
| CN108924944A (en) * | 2018-07-19 | 2018-11-30 | 重庆邮电大学 | The dynamic optimization method of contention window value coexists in LTE and WiFi based on Q-learning algorithm |
| CN109188414A (en) * | 2018-09-12 | 2019-01-11 | 北京工业大学 | A kind of gesture motion detection method based on millimetre-wave radar |
-
2019
- 2019-01-18 CN CN201910048068.XA patent/CN109766951A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160162027A1 (en) * | 2012-06-14 | 2016-06-09 | Immersion Corporation | Haptic effect conversion system using granular synthesis |
| CN104951757A (en) * | 2015-06-10 | 2015-09-30 | 南京大学 | Action detecting and identifying method based on radio signals |
| CN105807935A (en) * | 2016-04-01 | 2016-07-27 | 中国科学技术大学苏州研究院 | Gesture control man-machine interactive system based on WiFi |
| CN106295684A (en) * | 2016-08-02 | 2017-01-04 | 清华大学 | A kind of the most continuous based on micro-Doppler feature/discontinuous gesture recognition methods |
| CN106658703A (en) * | 2016-10-12 | 2017-05-10 | 南京邮电大学 | Cosine similarity based RSS (Received Signal Strength) detection difference compensation method |
| CN106604394A (en) * | 2016-12-28 | 2017-04-26 | 南京航空航天大学 | CSI-based indoor human body motion speed judgment model |
| CN108924944A (en) * | 2018-07-19 | 2018-11-30 | 重庆邮电大学 | The dynamic optimization method of contention window value coexists in LTE and WiFi based on Q-learning algorithm |
| CN109188414A (en) * | 2018-09-12 | 2019-01-11 | 北京工业大学 | A kind of gesture motion detection method based on millimetre-wave radar |
Non-Patent Citations (1)
| Title |
|---|
| 鲁勇: "基于WiFi信号的人体行为感知技术研究综述", 《计算机学报》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113449587A (en) * | 2021-04-30 | 2021-09-28 | 北京邮电大学 | Human behavior recognition and identity authentication method and device and electronic equipment |
| CN113238659A (en) * | 2021-06-29 | 2021-08-10 | 中国科学技术大学 | Real-time behavior identification method and system based on WIFI signal |
| CN114764580A (en) * | 2022-06-15 | 2022-07-19 | 湖南工商大学 | Real-time human body gesture recognition method based on no-wearing equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11763599B2 (en) | Model training method and apparatus, face recognition method and apparatus, device, and storage medium | |
| Zhang et al. | Widar3. 0: Zero-effort cross-domain gesture recognition with wi-fi | |
| CN107633227B (en) | CSI-based fine-grained gesture recognition method and system | |
| Lee et al. | Kinect-based Taiwanese sign-language recognition system | |
| Liu | Fingerprint classification based on Adaboost learning from singularity features | |
| CN105844216A (en) | Detection and matching mechanism for recognition of handwritten letters using WiFi signals | |
| CN111103976B (en) | Gesture recognition method and device and electronic equipment | |
| CN109766951A (en) | A kind of WiFi gesture identification based on time-frequency statistical property | |
| CN111027378A (en) | Pedestrian re-identification method, device, terminal and storage medium | |
| US11630518B2 (en) | Ultrasound based air-writing system and method | |
| CN108564040B (en) | Fingerprint activity detection method based on deep convolution characteristics | |
| CN107491729B (en) | Handwritten digit recognition method based on cosine similarity activated convolutional neural network | |
| Tang et al. | WiFi CSI gesture recognition based on parallel LSTM-FCN deep space-time neural network | |
| CN112733609A (en) | Domain-adaptive Wi-Fi gesture recognition method based on discrete wavelet transform | |
| Liu et al. | Novel algorithm for hand gesture recognition utilizing a wrist-worn inertial sensor | |
| Uysal et al. | RF-Wri: An efficient framework for RF-based device-free air-writing recognition | |
| CN111444488A (en) | Identity authentication method based on dynamic gesture | |
| Chen et al. | Dynamic gesture recognition using wireless signals with less disturbance | |
| Boyali et al. | A robust gesture recognition algorithm based on sparse representation, random projections and compressed sensing | |
| CN105354532A (en) | Hand motion frame data based gesture identification method | |
| Ogris et al. | Continuous activity recognition in a maintenance scenario: combining motion sensors and ultrasonic hands tracking | |
| Mezari et al. | Gesture recognition using symbolic aggregate approximation and dynamic time warping on motion data | |
| Zha et al. | LT-SEI: Long-tailed specific emitter identification based on decoupled representation learning in low-resource scenarios | |
| Elshenaway et al. | On-air hand-drawn doodles for IoT devices authentication during COVID-19 | |
| Infantino et al. | A framework for sign language sentence recognition by commonsense context |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190517 |