CN110647857A - Gait recognition method and system - Google Patents
Gait recognition method and system Download PDFInfo
- Publication number
- CN110647857A CN110647857A CN201910932224.9A CN201910932224A CN110647857A CN 110647857 A CN110647857 A CN 110647857A CN 201910932224 A CN201910932224 A CN 201910932224A CN 110647857 A CN110647857 A CN 110647857A
- Authority
- CN
- China
- Prior art keywords
- gait
- module
- sample set
- information
- acceleration 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.)
- Pending
Links
- 230000005021 gait Effects 0.000 title claims abstract description 122
- 238000000034 method Methods 0.000 title claims abstract description 36
- 230000001133 acceleration Effects 0.000 claims abstract description 71
- 238000012549 training Methods 0.000 claims abstract description 25
- 238000012545 processing Methods 0.000 claims description 16
- 238000007781 pre-processing Methods 0.000 claims description 14
- 238000004891 communication Methods 0.000 claims description 13
- 238000012706 support-vector machine Methods 0.000 claims description 11
- 238000000605 extraction Methods 0.000 claims description 8
- 238000010606 normalization Methods 0.000 claims description 8
- 238000010276 construction Methods 0.000 claims description 7
- 238000001228 spectrum Methods 0.000 claims description 5
- 238000005516 engineering process Methods 0.000 abstract description 12
- 230000007613 environmental effect Effects 0.000 abstract description 4
- 230000006870 function Effects 0.000 description 28
- 239000013598 vector Substances 0.000 description 16
- 230000008569 process Effects 0.000 description 10
- 238000010586 diagram Methods 0.000 description 9
- 238000004364 calculation method Methods 0.000 description 8
- 238000012360 testing method Methods 0.000 description 8
- 230000009977 dual effect Effects 0.000 description 6
- 238000013507 mapping Methods 0.000 description 5
- 238000007635 classification algorithm Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 210000002683 foot Anatomy 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 206010021703 Indifference Diseases 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 239000013078 crystal Substances 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 210000002414 leg Anatomy 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000017105 transposition Effects 0.000 description 1
- 210000000689 upper leg Anatomy 0.000 description 1
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
- G06F2218/12—Classification; Matching
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
- A61B5/11—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb
- A61B5/112—Gait analysis
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
- A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B2562/00—Details of sensors; Constructional details of sensor housings or probes; Accessories for sensors
- A61B2562/02—Details of sensors specially adapted for in-vivo measurements
- A61B2562/0219—Inertial sensors, e.g. accelerometers, gyroscopes, tilt switches
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
Abstract
The invention discloses a gait recognition method and a system in the technical field of biological recognition, and aims to solve the technical problem that the application scene of the gait recognition technology based on images in the prior art is limited because the gait recognition technology based on images relies on external environmental factors and requires enough light and specific background conditions to acquire the gait images of people. The method comprises the following steps: constructing a sample set based on the collected gait information, wherein the gait information comprises gait acceleration data; training a classifier by using the sample set; and identifying the target gait by adopting a trained classifier.
Description
Technical Field
The invention relates to a gait recognition method and a gait recognition system, and belongs to the technical field of biological recognition.
Background
With the technological progress, especially the rise of the artificial intelligence technology, the transformation from the traditional password mode to the biological identification mode of the authentication technology is driven. The biometric identification technology takes the unique physiological characteristics of a human body as authentication information, and is one of the mainstream authentication technologies at present due to the characteristics of difficult simulation, long timeliness and the like.
Gait recognition is a biometric technique that recognizes the identity of a person based on gait information of the person while walking. The gait authentication technology has uniqueness because every person has a unique walking mode. At present, a mode of acquiring human gait information based on an image is generally regarded by researchers, and although the gait recognition technology based on the image is relatively mature and has a high recognition rate, the gait recognition technology based on the image depends on external environmental factors, and the human gait image can be acquired only by needing enough light and corresponding background conditions, so that the application of the gait recognition technology in some occasions is limited.
Disclosure of Invention
Aiming at the defects of the prior art, the invention aims to provide a gait recognition method and a system, so as to solve the technical problem that the application scene of the gait recognition technology based on images is limited because the gait recognition technology in the prior art relies on external environmental factors and needs enough light and specific background conditions to acquire the gait images of people.
In order to solve the technical problems, the technical scheme adopted by the invention is as follows:
a gait recognition method comprises the following steps:
constructing a sample set based on the collected gait information, wherein the gait information comprises gait acceleration data;
training a classifier by using the sample set;
and identifying the target gait by adopting a trained classifier.
Preferably, constructing a sample set based on the collected gait information comprises:
carrying out Fourier transform on the gait information to obtain a spectrogram of gait acceleration data;
extracting gait features based on the spectrogram;
constructing a sample set based on the gait features.
Preferably, the fourier transform is a fast fourier transform.
Preferably, the fourier transform is performed on the gait information, further comprising: and preprocessing the gait information, wherein the preprocessing comprises at least any one of denoising processing, normalization processing and windowing processing.
Preferably, the classifier is a support vector machine.
In order to achieve the above object, the present invention further provides a gait recognition system, which comprises a lower computer acquisition module and an upper computer in communication connection therewith, wherein the lower computer acquisition module comprises an acceleration sensor module, the upper computer comprises a sample set construction module and an identification module which are electrically connected with each other,
the lower computer acquisition module: the gait information acquisition module is used for acquiring gait information and transmitting the gait information to the upper computer;
an acceleration sensor module: the gait acceleration data acquisition unit is used for acquiring gait acceleration data;
a sample set construction module: for constructing a sample set based on the collected gait information;
an identification module: and the classifier is used for training the classifier by using the sample set, and the trained classifier is used for identifying the target gait.
Preferably, the sample set construction module comprises an extraction module and a training module, which are electrically connected to each other, wherein,
an extraction module: the gait acceleration data acquisition device comprises a frequency spectrum graph used for carrying out Fourier transform on gait information to acquire gait acceleration data, and gait features are extracted based on the frequency spectrum graph, wherein the Fourier transform comprises fast Fourier transform;
a training module: for constructing a sample set based on gait characteristics.
Preferably, the gait information processing device further comprises a preprocessing module for preprocessing the gait information, wherein the preprocessing comprises at least any one of denoising processing, normalization processing and windowing processing.
Preferably, the acceleration sensor module is a MEMS acceleration sensor module.
Compared with the prior art, the invention has the following beneficial effects: the method of the invention collects the acceleration data of the person walking, preprocesses the acceleration data, extracts the gait characteristics through FFT, constructs a training sample set, constructs a support vector machine classifier, and utilizes the trained classifier to carry out gait recognition, thereby effectively overcoming the condition that the gait recognition based on the image is too dependent on external environmental factors. The system adopts the method to identify the gait, utilizes the portable electronic equipment to collect the acceleration data of the person during walking, is small, exquisite, convenient and practical, and can ensure that the identification result has higher accuracy.
Drawings
FIG. 1 is a schematic flow diagram of an embodiment of the method of the present invention;
FIG. 2 is a schematic diagram of a fast Fourier transform calculation process in an embodiment of the method of the present invention;
FIG. 3 is a diagram of a linear branch-able support vector machine according to an embodiment of the present invention;
FIG. 4 is a diagram illustrating a separating hyperplane of a low-dimensional to high-dimensional mapping structure in an embodiment of the method of the present invention;
fig. 5 is a schematic structural diagram of an embodiment of the system of the present invention.
Detailed Description
The invention is further described below with reference to the accompanying drawings. The following examples are only for illustrating the technical solutions of the present invention more clearly, and the protection scope of the present invention is not limited thereby.
The invention provides a gait recognition method, which utilizes MEMS (Micro-Electro-Mechanical System) acceleration sensors fixed on feet, legs and thighs to collect gait information of a wearer, wherein the MEMS acceleration sensors include but are not limited to mobile phones, sports foot rings and other portable electronic devices, so that the defect that image-based gait recognition is too dependent on external environment factors is overcome.
As shown in fig. 1, is a schematic flow chart of an embodiment of the method of the present invention, and the method includes the following steps:
step one, a preparation stage. When the MEMS acceleration sensor is used for collecting gait acceleration data of a person, a place which is few and open and has a smooth road surface is selected, the person to be collected walks normally, and the gait needs to be natural and uniform.
And step two, acquiring the original data signal acquired by the MEMS acceleration sensor. The raw data signal includes: signals generated by human body movement, signals generated by human body shaking and noise signals of equipment;
the MEMS acceleration sensor is a piezoelectric acceleration sensor and works based on the piezoelectric effect of a piezoelectric crystal; preferably, the MEMS acceleration sensor in this embodiment is an MPU6050, which includes a three-axis acceleration sensor and a three-axis gyroscope (X, Y, Z three axes); the positive direction indicated in the MPU6050 is the X-axis positive direction, the Y-axis positive direction on the left side, and the Z-axis positive direction on the upper side, and the clockwise rotation direction around the X, Y, Z-axis positive direction, the roll angle (roll) around the X-axis rotation, the pitch angle (yaw) around the Y-axis rotation, and the yaw angle (pitch) around the Z-axis rotation are the positive directions, respectively. When the person to be collected starts to move, starting one-time data collection, immediately shutting down after the collection is finished, and storing the collected original data; the system cannot be started in advance or shut down after delay in the acquisition process; after the acquisition of one original data is finished, the computer is shut down and waits for a period of time, and then the computer is started and carries out the acquisition of the next original data.
And step three, preprocessing is carried out according to the original acceleration data acquired in the step two. The pretreatment comprises the following steps: denoising, normalizing and windowing;
3-1) firstly, carrying out denoising treatment on the original acceleration data by adopting a digital low-pass filter so as to reduce measurement noise and improve the measurement precision of the acceleration data. The differential equation of a common hardware RC low-pass filter is expressed by a differential equation to be solved, so that the function of hardware filtering can be simulated by adopting a software algorithm;
3-2) then, carrying out normalization processing on the acceleration data to change the acceleration data from an absolute value to a relative value relation. Let the acceleration data after de-noising be A (A ═ A)1,A2,...Ai...An]N ∈ C), normalized interval is [ X ∈ C ]1,X2](X1,X2E C), the normalized acceleration data is B (B ═ B)1,B2,...Bi...Bn]N ∈ C), the normalized calculation formula is as follows:
wherein A, B are all data sets, AiFor the i-th acceleration data element after de-noising, BiIs the normalized ith acceleration data element, n is the number of acceleration data elements in the set, C is a constant, AminIs the smallest one of A, AmaxIs the largest one of A, [ X ]1,X2]Generally take [0,1]Or [ -1,1 [)]. Normalizing the acceleration data to the interval [ -1,1 ] by a normalization process];
3-3) then, adopting a rectangular sliding window with the window length of being 50% overlapped of 512 sample points to carry out windowing processing on the normalized acceleration data, wherein the calculation formula of the window function is as follows:
in the formula, w (t)iIs the ith window function, i is a positive integer starting from 1, t is time (unit: seconds), n is a positive integer, and C represents a constant.
In this embodiment, the denoising process is to remove high-frequency random noise (mainly pulse noise) generated by the action of the gravitational field during the process of acquiring raw data by the MEMS acceleration sensor, so as to improve the signal-to-noise ratio. The normalization processing is to convert a dimensional expression into a dimensionless expression to make the dimensionless expression into a pure quantity, and aims to eliminate the influence of the absolute quantity of data signal parameters on gait recognition, simplify the recognition process and reduce the recognition difficulty; the gait acceleration data signals after normalization processing have the same minimum value and maximum value, and are unified in the same interval, so that a foundation is laid for feature selection and extraction and subsequent gait recognition. The windowing process is to truncate the signal using different truncation functions, which are also referred to as window functions. Because the input data is long, the input data needs to be cut off first, and then the features are extracted and classified, and therefore windowing processing is performed before the gait data features are extracted. By cutting the acceleration data signal into signals with the length of 512 sample points through the step 3-3), the frequency spectrum energy leakage and the error generated by Fast Fourier Transform (FFT) on non-integer periods are reduced. In this embodiment, since the sampling frequency of the adopted MEMS acceleration sensor is 100Hz, and each windowed data is 512 sample points, it takes 5.12 seconds to acquire 512 sample data, and about 5 cycles (i.e. 5 steps) can be included at the normal walking pace.
Step four, for the acceleration data after the pretreatment, converting the acceleration data of the time domain into the acceleration data of the frequency domain through FFT, further acquiring a spectrogram of the acceleration data, and extracting gait features based on the spectrogram;
the Fast Fourier Transform (FFT) is a fast algorithm of Discrete Fourier Transform (DFT), as shown in fig. 2, and is a schematic diagram of a fast Fourier Transform calculation process in an embodiment of the method of the present invention, and the FFT is a process of converting a coefficient representation into a point value representation through DFT, multiplying the coefficient representation by the point value representation, and converting the coefficient representation into a coefficient representation through Inverse Discrete Fourier Transform (IDFT). Taking n sample points as an example, the multiplication calculation amount of DFT is n2Second, and the multiplication of FFT is nlog2n times, and the calculation result is the same as that of DET, so that the calculation efficiency of the fast Fourier transform is much higher than that of the discrete Fourier transform in the aspect of processing a large amount of data;
the gait feature extraction is to select the signal feature of the acceleration data. In this embodiment, 512 sampling points are provided in one direction of one piece of windowed gait data, and 1536 pieces of data are provided in three axial directions, and if a Support Vector Machine (SVM) is used to directly train the data, a lot of time is required. The characteristics which can represent the characteristics of the acceleration data signals are extracted, and the classification efficiency and accuracy are improved. Because the data volume obtained after the preprocessing is still large, the data needs to be described by using relatively simple characteristic parameters, that is, fast fourier transform is performed on each windowed acceleration data to obtain a corresponding spectrogram, and amplitudes corresponding to each frequency are extracted to serve as signal characteristics of the acceleration data. In this embodiment, the preprocessed acceleration data X, Y, Z is subjected to fast fourier transform on each of the three axes, the gait feature is the coefficient of the first 128 dimensions of each axis, the coefficient of the first dimension of each axis is the dc component, and the number of dimensions of the gait feature vector of each testee is 127 × 3 dimensions.
Step five, the gait characteristics are brought into a support vector machine for learning and training, and a binary classifier is designed, which comprises the following specific steps:
5-1) assuming that n testees exist, the gait data of the n testees correspondingly have n step characteristic vectors which are recorded asN sample sets are total, n-1 gait feature vectors are selected as a training set, and the rest 1 gait feature vectors are selected as a test set until the gait feature vectors selected as the test set traverse the n gait feature vectors;
5-2) Using the "one-to-one" classification algorithm, designing a classifier between any two classes of samples would requireA binary classifier. The two types of samples comprise gait feature vectors in a test set and gait feature vectors in a training set;
5-3) adopting a nonlinear support vector machine as a binary classifier model, adopting a radial basis function, namely a Gaussian function, as a kernel function, and adopting the Gaussian kernel function K (x)i,xj) Is given by the formulaIn the formula, xi,xjThe method comprises the following steps that any two different sample points in a sample set are provided, sigma is a bandwidth parameter of a kernel function, and | | · | | represents norm operation;
and constructing a Lagrangian function formula as follows:
wherein w is the normal vector of the hyperplane, b is the intercept of the hyperplane, λ is the Lagrange multiplier, Φ (x)i) Is xiMapping to the inner product space H, i is more than or equal to 1 and less than or equal to m, wherein m is the number of gait feature vectors, namely the number of testees;
the duality of the original problem is a very minimal problem in terms of duality, i.e.To solve the saddle point of the Lagrangian function (here, the maximum hyperplane distance), the partial derivative of L (w, b, lambda) with respect to w and b is first obtainedTaken into L (w, b, λ), then the functional formula for the dual problem is as follows:
in the formula, yi,yjFor class marking, λi,λjIs a Lagrange multiplier, s.t. represents a constraint condition, namely the constraint condition isλi,λj≥0i,j=1,2,...,m;
Optimal solution to the dual problemTo solve λ in the dual problem, the formula is:where argmax is a function of solving parameters (set) for the function, and T represents the transposition of the vector;
setting a parameter w*,b*The optimal solution of the original problem satisfies the following formula:
a nonlinear classification decision function constructed from the sameWhere sgn is a sign function with a value of-1 or 1,for the ith element in the optimal solution to the dual problem,is b is*The ith element in (1);
5-4) based on the data of the n testees, taking the test set as the input of the decision function in the step 5-3) to carry out classification training to obtain the proportion of correct classification samples;
5-5) using the n sample sets obtained in the step 5-1), sequentially taking one of the n sample sets as a test set, taking the rest groups as a training set, repeating the step 5-3) and the step 5-4), obtaining a corresponding model with the highest correct classification proportion, and taking the corresponding model as a binary classifier between the two classes;
the support vector machine is a binary model and aims to find a hyperplane to segment a sample, and the segmentation principle is interval maximization. The models from simple to complex include: a linearly separable maximum interval model, an approximately linearly separable maximum interval model, and a non-linearly inseparable maximum interval model; fig. 3 is a schematic diagram of a linear branch-able support vector machine according to an embodiment of the present invention, which is divided into two-dimensional linear branches, i.e., a dividing plane y (x) ═ wTx + b), wherein w is hyperplaneThe normal vector of the surface, b is the intercept of the hyperplane, and the decision function for classification is denoted as f (x) sgn (y (x)), as can be seen from fig. 3, the straight line y (x) is not unique, only when y (x) is (w (x)), (x)TWhen x + b) is 0, the hyperplane has the best resistance to disturbance, namely the classification mark yiE { -1, +1 }; point corresponding classification mark y when y (x) is greater than 0iPoint corresponding to y (x) less than 0iThe dotted line is called interval boundary, and the point on the interval boundary is called support vector, namely the support vector satisfies yi(wTxi+ b) 1, in binary classification, the value of the classification mark is only two, i.e. yi1, (+ -.) corresponding to (x)i,yi) Are sample points. When the hyperplane has maximum spacing from the data on both sides of the dotted line, i.e. maximum spacing isThere is maximum "tolerance" for the limitations or noise of the data of the training set. If the interval is maximized, findThen w is minimized, i.e. solvedThen its equivalent solving equation is as follows:s.t.yi(wTxi+ b) is equal to or greater than 1, i is 1,2,. n, i.e. the constraint is yi(wTxi+b)≥1;
The lagrange function L (w, b, λ) is constructed by a constraint, as follows:
wherein λ isiI is more than or equal to 0, i is more than or equal to 1 and less than or equal to m, w is the normal vector of the hyperplane, b is the intercept of the hyperplane, yiFor class marking, xiIs an unknown quantity.
The duality of the original problem is a very minimal problem in terms of duality, i.e.Solving the saddle point of the lagrange function (here, taking the maximum hyperplane distance), i.e., taking the partial derivatives of L (w, b, λ) with respect to w and b, yields the following equation:
will be provided withTaken into L (w, b, λ), the functional formula for the dual problem is as follows:
i.e. with the constraint ofλi,λjMore than or equal to 0 i, j ═ 1, 2.., m; solving λ in the dual problem, the formula is as follows:
parameter w*,b*The optimal solution set of the original problem meets the following formula:
the formula of the constructed linearly separable classification decision function is as follows:
the maximum interval model of approximate linear divisibility is data allowing hyperplane to have division error, so relaxation variable xi is introducedi(ξiNot less than 0), the constraint is put to yi(wTxi+b)+ξiMore than or equal to 1, vector xi ═ xi (xi)1,ξ2,...,ξn)TReflecting the condition that the training set is allowed to be wrongly divided, and describing the degree of wrongly dividing the training set by xi; a penalty factor C is introduced, wherein C represents the penalty degree of classification errors under the condition of indifference linearity. The model equivalence solution equation is thus as follows:and constructing a classification decision function by constructing the approximately linearly separable maximum interval model, wherein the derivation process is similar to the construction of the linearly separable classification decision function.
An approximately linearly separable classification decision function is constructed as follows:
for a nonlinear inseparable maximum interval model, by introducing a kernel function, recording the kernel function as K (x, z) to (phi (x) · phi (z)) wherein phi (x) represents the mapping of x to an inner product space H, mapping an input space to a high-dimensional feature space, and finally constructing an optimal separation hyperplane in the high-dimensional feature space, so as to separate nonlinear data which are not easily separated on a plane per se; as shown in fig. 4, which is a schematic diagram of a separated hyperplane constructed by low-dimensional to high-dimensional mapping in the embodiment of the method of the present invention, a stack of data cannot be divided in a two-dimensional space, and thus is mapped into a three-dimensional space for division, and a solution formula of a nonlinear indivisible model is as follows:constructed nonlinear classification decision function of
The one-to-one classification algorithm is to design a binary classifier between any two classes of samples (training set and test set), so that samples of n classes are designedA binary classifier. When classifying an unknown sample, let the classification decision function between classes i, j be fij(x) (i ≠ j, j ≠ 1.. multidot.n), if fij(x) If x is greater than 0, x is considered to belong to the class i, i.e. the class label Fij(x) 1 is ═ 1; otherwise, belong to the class j, i.e. Fij(x) Is-1. According to the principle, when the samples x to be identified are classified, calculation is carried outThe category i corresponding to the maximum value is taken as the category of x, namely the category with the most votes is the category of the unknown sample.
And step six, selecting gait data of the tested person, repeating the steps one to four to obtain the step characteristics of the tested person, bringing the step characteristics into the SVM classifier after learning and training in the step five, and selecting whether the gait characteristics are the identity of the corresponding person or not, thereby achieving the effect of gait recognition.
The specific embodiment of the present invention provides a gait recognition system, as shown in fig. 5, which is a schematic structural diagram of an embodiment of the system of the present invention, the system includes a lower computer acquisition module and an upper computer in communication connection therewith, the lower computer acquisition module includes a power module, an MEMS acceleration sensor module, and a master control module, the upper computer includes a preprocessing module, an extraction module, a generation module, a training module, and a recognition module, the lower computer acquisition module and the upper computer are in communication connection through a communication module, wherein,
(1) the lower computer acquisition module is used for acquiring the original acceleration data of the MEMS acceleration sensor and transmitting the data to the upper computer, wherein,
the power supply module is used for supplying power to the lower computer acquisition module;
the MEMS acceleration sensor module is used for acquiring the original acceleration data of the current movement and sending the original acceleration data to the main control module through the integrated circuit bus;
the main control module is used for receiving the original data transmitted by the MEMS acceleration sensor module and transmitting the original data to the upper computer through the communication module;
(2) and the communication module is used for realizing data communication in the lower computer and communication between the upper computer and the lower computer. The communication mode comprises an integrated circuit bus (IIC) and Bluetooth communication, wherein the preferred embodiment of the Bluetooth communication is an HC-05 Bluetooth module;
(3) the upper computer is used for constructing a sample set based on the collected original acceleration data, training a classifier by using the sample set, identifying a target gait by using the trained classifier, wherein,
the preprocessing module is used for processing the raw acceleration data transmitted from the MEMS acceleration sensor;
the extraction module is used for extracting the gait characteristics of the person to be identified;
the training module is used for constructing a training set sample according to the gait characteristics of each tested person;
and the identification module is used for matching the gait characteristics of the person to be identified with the test set samples, comparing the test samples with the training samples according to a one-to-one classification algorithm, and finally determining the identity information of the other person.
As a preferred embodiment, the MEMS acceleration sensor module employs an MPU 6050;
in a preferred embodiment, the main control module adopts STM32F103VET 6;
in a preferred embodiment, the bluetooth communication employs an HC-05 bluetooth module.
The invention relates to a human body gait recognition method and a human body gait recognition system based on an acceleration sensor. The system carries out gait recognition based on the method, adopts portable electronic equipment to collect acceleration data of people during walking, is small, exquisite, convenient and practical, and can ensure that the recognition result has higher accuracy.
The above description is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, several modifications and variations can be made without departing from the technical principle of the present invention, and these modifications and variations should also be regarded as the protection scope of the present invention.
Claims (9)
1. A gait recognition method is characterized by comprising the following steps:
constructing a sample set based on the collected gait information, wherein the gait information comprises gait acceleration data;
training a classifier by using the sample set;
and identifying the target gait by adopting a trained classifier.
2. The gait recognition method according to claim 1, wherein constructing a sample set based on the collected gait information comprises:
carrying out Fourier transform on the gait information to obtain a spectrogram of gait acceleration data;
extracting gait features based on the spectrogram;
constructing a sample set based on the gait features.
3. The gait recognition method according to claim 2, wherein the fourier transform is a fast fourier transform.
4. The gait recognition method according to claim 2, wherein the step of performing fourier transform on the gait information further comprises: and preprocessing the gait information, wherein the preprocessing comprises at least any one of denoising processing, normalization processing and windowing processing.
5. The gait recognition method according to any one of claims 1 to 4, wherein the classifier is a support vector machine.
6. A gait recognition system is characterized by comprising a lower computer acquisition module and an upper computer which is in communication connection with the lower computer acquisition module, wherein the lower computer acquisition module comprises an acceleration sensor module, the upper computer comprises a sample set construction module and a recognition module which are electrically connected with each other,
the lower computer acquisition module: the gait information acquisition module is used for acquiring gait information and transmitting the gait information to the upper computer;
an acceleration sensor module: the gait acceleration data acquisition unit is used for acquiring gait acceleration data;
a sample set construction module: for constructing a sample set based on the collected gait information;
an identification module: and the classifier is used for training the classifier by using the sample set, and the trained classifier is used for identifying the target gait.
7. The gait recognition system of claim 6, wherein the sample set construction module includes an extraction module and a training module electrically connected to each other, wherein,
an extraction module: the gait acceleration data acquisition device comprises a frequency spectrum graph used for carrying out Fourier transform on gait information to acquire gait acceleration data, and gait features are extracted based on the frequency spectrum graph, wherein the Fourier transform comprises fast Fourier transform;
a training module: for constructing a sample set based on gait characteristics.
8. The gait recognition system according to claim 6 or 7, further comprising a preprocessing module for preprocessing the gait information, the preprocessing including at least any one of denoising, normalization, and windowing.
9. The gait recognition system of claim 6 or 7, wherein the acceleration sensor module is a MEMS acceleration sensor module.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910932224.9A CN110647857A (en) | 2019-09-29 | 2019-09-29 | Gait recognition method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910932224.9A CN110647857A (en) | 2019-09-29 | 2019-09-29 | Gait recognition method and system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110647857A true CN110647857A (en) | 2020-01-03 |
Family
ID=68993112
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910932224.9A Pending CN110647857A (en) | 2019-09-29 | 2019-09-29 | Gait recognition method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110647857A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111544005A (en) * | 2020-05-15 | 2020-08-18 | 中国科学院自动化研究所 | Parkinson's disease dyskinesia quantification and identification method based on support vector machine |
CN111950437A (en) * | 2020-08-07 | 2020-11-17 | 平安科技(深圳)有限公司 | Gait recognition method and device based on deep learning model and computer equipment |
CN113591552A (en) * | 2021-06-18 | 2021-11-02 | 新绎健康科技有限公司 | Method and system for identity recognition based on gait acceleration data |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103886341A (en) * | 2014-03-19 | 2014-06-25 | 国家电网公司 | Gait behavior recognition method based on feature combination |
CN104323780A (en) * | 2014-10-30 | 2015-02-04 | 上海交通大学 | Support vector machine-based pedestrian gait classifying system and method |
CN106529499A (en) * | 2016-11-24 | 2017-03-22 | 武汉理工大学 | Fourier descriptor and gait energy image fusion feature-based gait identification method |
-
2019
- 2019-09-29 CN CN201910932224.9A patent/CN110647857A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103886341A (en) * | 2014-03-19 | 2014-06-25 | 国家电网公司 | Gait behavior recognition method based on feature combination |
CN104323780A (en) * | 2014-10-30 | 2015-02-04 | 上海交通大学 | Support vector machine-based pedestrian gait classifying system and method |
CN106529499A (en) * | 2016-11-24 | 2017-03-22 | 武汉理工大学 | Fourier descriptor and gait energy image fusion feature-based gait identification method |
Non-Patent Citations (1)
Title |
---|
赵琦琳 等: "《人工神经网络在环境科学与工程中的设计应用》", 上海交通大学出版社, pages: 156 - 157 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111544005A (en) * | 2020-05-15 | 2020-08-18 | 中国科学院自动化研究所 | Parkinson's disease dyskinesia quantification and identification method based on support vector machine |
CN111544005B (en) * | 2020-05-15 | 2022-03-08 | 中国科学院自动化研究所 | Parkinson's disease dyskinesia quantification and identification method based on support vector machine |
CN111950437A (en) * | 2020-08-07 | 2020-11-17 | 平安科技(深圳)有限公司 | Gait recognition method and device based on deep learning model and computer equipment |
WO2021139337A1 (en) * | 2020-08-07 | 2021-07-15 | 平安科技(深圳)有限公司 | Deep learning model-based gait recognition method and apparatus, and computer device |
CN113591552A (en) * | 2021-06-18 | 2021-11-02 | 新绎健康科技有限公司 | Method and system for identity recognition based on gait acceleration data |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Kumar et al. | Envisioned speech recognition using EEG sensors | |
Ravi et al. | Deep learning for human activity recognition: A resource efficient implementation on low-power devices | |
CN110647857A (en) | Gait recognition method and system | |
CN105841961A (en) | Bearing fault diagnosis method based on Morlet wavelet transformation and convolutional neural network | |
CN109299751B (en) | EMD data enhancement-based SSVEP electroencephalogram classification method of convolutional neural model | |
Ng et al. | Iris recognition using rapid Haar wavelet decomposition | |
CN111062250A (en) | Multi-subject motor imagery electroencephalogram signal identification method based on depth feature learning | |
CN111461176A (en) | Multi-mode fusion method, device, medium and equipment based on normalized mutual information | |
CN105117708A (en) | Facial expression recognition method and apparatus | |
CN112294341B (en) | Sleep electroencephalogram spindle wave identification method and system based on light convolutional neural network | |
CN103646255A (en) | Face detection method based on Gabor characteristics and extreme learning machine | |
CN101815466A (en) | A non-invasive device NADI TARANGINI useful for quantitative detection of arterial NADI pulse waveform | |
CN111950437A (en) | Gait recognition method and device based on deep learning model and computer equipment | |
Bu | Human motion gesture recognition algorithm in video based on convolutional neural features of training images | |
US20160192863A1 (en) | Method for sensor orientation invariant gait analysis using gyroscopes | |
CN111476158A (en) | Multi-channel physiological signal somatosensory gesture recognition method based on PSO-PCA-SVM | |
Tao et al. | Finger-vein recognition using bidirectional feature extraction and transfer learning | |
Wu et al. | Activity recognition from mobile phone using deep CNN | |
Qaroush et al. | Smart, comfortable wearable system for recognizing Arabic Sign Language in real-time using IMUs and features-based fusion | |
CN108563939B (en) | Human body identity recognition based on gait track curve characteristics | |
CN112580486B (en) | Human behavior classification method based on radar micro-Doppler signal separation | |
CN107045624A (en) | A kind of EEG signals pretreatment rolled into a ball based on maximum weighted and sorting technique | |
Liu et al. | Dual-domain and multiscale fusion deep neural network for PPG biometric recognition | |
CN108470182A (en) | A kind of brain-computer interface method enhanced for asymmetric brain electrical feature with identification | |
CN109993132B (en) | Pattern recognition generation method and system based on electroencephalogram signals |
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 |