CN106264545B - Step recognition method and device - Google Patents

Step recognition method and device Download PDF

Info

Publication number
CN106264545B
CN106264545B CN201610639204.9A CN201610639204A CN106264545B CN 106264545 B CN106264545 B CN 106264545B CN 201610639204 A CN201610639204 A CN 201610639204A CN 106264545 B CN106264545 B CN 106264545B
Authority
CN
China
Prior art keywords
step recognition
paces
data
user
recognition model
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201610639204.9A
Other languages
Chinese (zh)
Other versions
CN106264545A (en
Inventor
付强
张小光
姜言言
贾雪静
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.)
Beijing Fengniao View Technology Co Ltd
Original Assignee
Beijing Fengniao View Technology Co Ltd
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 Beijing Fengniao View Technology Co Ltd filed Critical Beijing Fengniao View Technology Co Ltd
Priority to CN201610639204.9A priority Critical patent/CN106264545B/en
Publication of CN106264545A publication Critical patent/CN106264545A/en
Application granted granted Critical
Publication of CN106264545B publication Critical patent/CN106264545B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • 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
    • A61B5/103—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
    • A61B5/1121—Determining geometric values, e.g. centre of rotation or angular range of movement
    • 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
    • A61B5/1123—Discriminating type of movement, e.g. walking or running
    • 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
    • A61B5/7225—Details of analog processing, e.g. isolation amplifier, gain or sensitivity adjustment, filtering, baseline or drift compensation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
    • G06F18/24133—Distances to prototypes
    • G06F18/24137—Distances to cluster centroïds
    • G06F18/2414—Smoothing the distance, e.g. radial basis function networks [RBFN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/245—Classification techniques relating to the decision surface
    • G06F18/2451—Classification techniques relating to the decision surface linear, e.g. hyperplane

Landscapes

  • Health & Medical Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Public Health (AREA)
  • Heart & Thoracic Surgery (AREA)
  • General Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Biophysics (AREA)
  • Pathology (AREA)
  • Biomedical Technology (AREA)
  • Veterinary Medicine (AREA)
  • Medical Informatics (AREA)
  • Molecular Biology (AREA)
  • Surgery (AREA)
  • Animal Behavior & Ethology (AREA)
  • Artificial Intelligence (AREA)
  • Physiology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Dentistry (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Signal Processing (AREA)
  • Geometry (AREA)
  • Power Engineering (AREA)
  • Psychiatry (AREA)
  • Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)

Abstract

Present disclose provides a kind of step recognition methods, it include: to be trained the step recognition model of multistage paces training sample data input least square method supporting vector machine, and the initial value of each parameter of initial step recognition model is determined by cross validation, form initial step recognition model；Obtain the current pace data in user's certain time；The current pace data are pre-processed；And pretreated current pace data are inputted into the initial step recognition model, thus identify the paces classification of user.Furthermore the disclosure additionally provides a kind of device using the step recognition method.

Description

Step recognition method and device
Technical field
This disclosure relates to motion detection technique field more particularly to a kind of step recognition method and device.
Background technique
As the rapid proliferation of embedded device and corresponding signal processing technology continue to develop, the appearance of embedded device State identification technology is widely used in space flight/aviation, military investigation and civil field.Paces detection is a Xiang Ji in gesture recognition The technology of plinth, for example judge using the data that accelerometer in smart phone and angular speed measure to carry the personnel of the equipment Paces state, and the personnel's paces for carrying the equipment are counted based on the paces state of identification, or just The paces state of identification come judge the personnel for carrying the equipment be for ambulatory status or run state, and thus estimation carry The speed of travel of the personnel of the equipment.The identification of paces can be using the fields such as navigation, security protection indoors.
Existing step recognition method usually carries out peak detection using the acceleration information that sensor measures, and passes through judgement Whether acceleration peak value is more than that threshold value carries out paces counting.Such method depend on subjective experience, paces counting accuracy compared with It is low, it is difficult to judge that paces state is walking or runs, it is difficult to judge the speed of paces.In addition, the parameter of unified setting can be led Stringent standardization is caused, and the paces of different people are regular and different, simple judgment rule can not be carried out for different user Personalisation process.
Summary of the invention
Based on existing step recognition and step number metering aspect there are problems, present disclose provides a kind of step recognition sides Method, comprising: the step recognition model of multistage paces training sample data input least square method supporting vector machine is trained, and The initial value that each parameter of initial step recognition model is determined by cross validation forms initial step recognition model；It obtains Take the current pace data in the certain time of family；The current pace data are pre-processed；And it will be preprocessed Current pace data input the initial step recognition model, thus identify user paces classification.
According to the step recognition method of the disclosure, the kernel function of the least square method supporting vector machine is radial base core letter Number.
According to the step recognition method of the disclosure, the original paces data of the multistage paces training sample data are to pass through The 3-axis acceleration and three that 3-axis acceleration sensor and three axis angular rate sensors in the equipment of user's carrying obtain in real time Axis angular rate.
According to the step recognition method of the disclosure, the original paces data become paces training sample by pretreatment Notebook data.
According to the step recognition method of the disclosure, the pretreatment is low-pass filtering.
According to the step recognition method of the disclosure, the initial value of the parameter of the step recognition model includes regularization parameter Initial value and Radial basis kernel function width initial value.
According to the step recognition method of the disclosure, each section of correspondence in the original paces data of multistage is preset One in multiple paces classifications.
According to the step recognition method of the disclosure, further include: the paces that user is successfully identified in accumulation a period of time Data；And be trained the step recognition model of the paces data accumulated input least square method supporting vector machine, pass through Cross validation updates each parameter of the step recognition model of user, thus to obtain the step recognition model for being exclusively used in the user.
According to another aspect of the disclosure, a kind of step recognition device is additionally provided, comprising: parameter set unit obtains Take the step recognition mould of multistage training sample data input least square method supporting vector machine corresponding with completely original paces data Type is trained, and the initial value of each parameter of initial step recognition model is determined from there through cross validation；Step of user number According to acquiring unit, the current pace data in user's certain time are obtained；Pretreatment unit, to the current pace data into Row pretreatment；And step recognition unit, pretreated current pace data are inputted into the initial step recognition model, Thus the paces classification of user is identified.
According to the step recognition device of the disclosure, the kernel function of the least square method supporting vector machine is radial base core letter Number.
According to the step recognition device of the disclosure, the paces data are by the step of user data capture unit 3-axis acceleration sensor and three the axis angular rate sensors 3-axis acceleration and three axis angular rates that obtain in real time.
According to the step recognition device of the disclosure, the pretreatment includes carrying out low-pass filtering to paces data.
According to the step recognition device of the disclosure, the initial value of each parameter of the initial step recognition model is canonical Change the initial value of the initial value of parameter and the width of Radial basis kernel function.
According to the step recognition device of the disclosure, multistage training sample data corresponding with completely original paces data In each sample correspond to one in preset multiple paces classifications.
Further include parameter updating unit according to the step recognition device of the disclosure, accumulates user's quilt in a period of time The paces data and input minimum two for the paces data for the successful identification accumulated as training sample data that success identifies The step recognition model for multiplying support vector machines is trained, and each ginseng of the step recognition model of user is updated by cross validation Number, thus to obtain the step recognition model for being exclusively used in the user.
Detailed description of the invention
The drawings herein are incorporated into the specification and forms part of this specification, and shows the implementation for meeting the disclosure Example, and together with specification for explaining the principles of this disclosure.
It is discussed in detail the disclosure by embodiment below with reference to the accompanying drawings, in attached drawing:
Fig. 1 show the flow diagram of one embodiment of the step recognition method of the disclosure；
Fig. 2 show the real-time identification process schematic diagram of the step recognition method of the disclosure；
Fig. 3 show the functional block diagram of the step recognition method of the disclosure.
The initial data that Fig. 4 show the disclosure obtains and pretreated flow diagram.
Fig. 5 show the flow chart of the real-time paces of the disclosure specifically identified.
Fig. 6 show the schematic block diagram of the step recognition device of the disclosure.
Specific embodiment
Example embodiments are described in detail here, and the example is illustrated in the accompanying drawings.Following description is related to When attached drawing, unless otherwise indicated, the same numbers in different drawings indicate the same or similar elements.Following exemplary embodiment Described in embodiment do not represent all implementations consistent with this disclosure.On the contrary, they be only with it is such as appended The example of the consistent device and method of some aspects be described in detail in claims, the disclosure.
It is only to be not intended to be limiting and originally open merely for for the purpose of describing particular embodiments in the term that the disclosure uses.? The "an" of singular used in disclosure and the accompanying claims book, " described " and "the" are also intended to including most shapes Formula, unless the context clearly indicates other meaning.It is also understood that term "and/or" used herein refers to and includes One or more associated any or all of project listed may combine.
It will be appreciated that though various information, but this may be described using term first, second, third, etc. in the disclosure A little information should not necessarily be limited by these terms.These terms are only used to for same type of information being distinguished from each other out.For example, not departing from In the case where disclosure range, hereinafter, one of two classifications can be referred to as first category or be referred to as the second class Not, similarly, another of two classifications can be referred to as second category or be referred to as first category.Depending on context, As used in this word " if " can be construed to " ... when " or " when ... " or " in response to determination ".
In order to make those skilled in the art more fully understand the disclosure, with reference to the accompanying drawings and detailed description to this public affairs It opens and is described in further detail.
Fig. 1 show the flow chart of one embodiment of the step recognition method 100 of the disclosure.Method 100 may include Following steps 101 to 103.
In a step 101, preliminary step recognition model is established.
In one embodiment of the present disclosure, least square method supporting vector machine is had chosen as step recognition model.It supports Vector machine (SVM, Support Vector Machine) is a kind of machine learning method based on Statistical Learning Theory, is established On empirical risk minimization, it is widely applied in various classification problems, it is extensive compared to other machine learning methods Ability is stronger.Practical problem is transformed into the feature space of higher-dimension by SVM by nonlinear transformation, is constructed in higher dimensional space linear Decision realizes the non-linear decision problem in former space, is avoided that the defect of over-fitting and Local Extremum existing for neural network Problem is excellent in small sample, non-linear and high dimensional pattern identification problem, can be generalized in various classification problems.This public affairs Opening the least square used and holding vector machine (LS-SVM, Least Squares Support Vector Machine) is standard branch The extension of vector machine is held, the solution that it solves optimal hyperlane in support vector machines using improved least square method is asked Topic, makes the inequality constraints in the optimization problem become equality constraint, to substantially reduce the complexity of calculating, calculating speed compared with Fastly, it can apply in various real-time systems.Therefore, LS-SVM is utilized in step recognition and facilitates compared with sample for the disclosure Suitable model is established in the case where this, is improved the accuracy that paces count, is accurately identified paces state, reach simple rule institute Inaccessiable effect.
In one embodiment of the present disclosure, the kernel function of least square method supporting vector machine can be Radial basis kernel function. Radial basis function is that a value depends only on real-valued function from initial point distance, that is, Φ (x)=Φ (∥ x ∥), or Person can also be the distance to any point c, and c point is known as central point, that is, Φ (x, c)=Φ (∥ x-c ∥).Any one The function phi for meeting Φ (x)=Φ (∥ x ∥) characteristic is all called radial basis function, and standard generally (is also named using Euclidean distance Do European radial basis function), although other distance functions are also possible, such as most common radial basis function is Gaussian kernel letter Number.In neural network structure, the primary function of full articulamentum and ReLU layers can be used as.
In order to establish preliminary step recognition model, it can specify that seven class paces classifications:
Classification 1, it is unidentified to arrive new paces；
Classification 2 recognizes the paces of " walking ", " at a slow speed "；
Classification 3 recognizes the paces of " walking ", " middling speed "；
Classification 4 recognizes the paces of " walking ", " quick "；
Classification 5 recognizes the paces of " running ", " at a slow speed "；
Classification 6 recognizes the paces of " running ", " middling speed "；
Classification 7 recognizes the paces of " running ", " quick ".
Although define herein this in 7 paces classification, user paces classification can also be adjusted according to their own needs It is whole.
User, which can adopt, carries out initialization training to step recognition model in various manners.There are two types of training methods, a kind of It is that the paces classification training data input step recognition model directly obtained by network is trained the model, thus directly Obtain the initial parameter value that the model carries out step recognition.Another way is by below obtaining the step of user data mentioned Unit is taken to directly acquire multistage training sample data corresponding with completely original paces data, it is then that the training sample data are defeated The step recognition model for entering least square method supporting vector machine is trained.
Above two mode can be certain for the moment to every part of training sample data of the input of least square method supporting vector machine K before quarter and the moment0The acceleration information and angular velocity data at a moment.Therefore, step recognition model can be expressed as
Y (k)=f (Α (k-k0),···,Α(k),Ω(k-k0),···,Ω(k))
It is a certain moment and its preceding k that it, which is inputted,0The acceleration information Α and angular velocity data Ω at a moment.
Its output y (k) can be one in preset multiple paces classifications.
By inputting the different training sample data of same paces classification repeatedly, certain can be determined together by cross validation The initial value of each parameter of the initial steps identification model of class paces.Thus initial steps identification model is established.
After establishing initial steps identification model, subsequent user (can will be explained) below by step recognition device Into practical step recognition step 102.In this step, user can identify the paces of oneself.
Shown in Fig. 2 is the detailed process of step recognition step 102.As shown in Fig. 2, firstly, being used in step S1021 Family is obtained in user's certain time by the step of user data capture unit (will explain below) in step recognition device Current pace data.Specifically, using 3-axis acceleration sensor and three shaft angles in step of user data capture unit Velocity sensor obtains user's 3-axis acceleration within a certain period of time and three axis angular rate data.In a reality of the disclosure Apply in example, can by the step recognition device (smart phone, intelligent wearable device etc.) entrained by pedestrian integrate 3-axis acceleration sensor and three axis angular rate sensors obtain the current pace data in user's certain time.These steps Cutting down data can separately include there are three the scalar size in dimension, and acceleration information therein can recorde as clock signal sequence Arrange ax(t)、ay(t)、az(t), angular velocity data therein can recorde as clock signal sequence ωx(t)、ωy(t)、ωz(t)。 Three axis accelerometer in step recognition device can be with the frequency acquisition acceleration information and angular velocity data of about 100Hz. The frequency can then can also increase paces data from main regulation for example, the possible cadence of some users is fast according to the needs of users Acquisition frequency.
Then, pretreatment unit (will explain below) is pre-processed (S1022) to the current pace data.Specifically For, which is, for example, low-pass filter described below.The low-pass filter carries out current pace data low The data obtained after pass filter can be expressed as ax′(t)、a′y(t)、az′(t)、ωx′(t)、ωy′(t)、ωz' (t), it may be assumed that
Α (k)={ ax′(k),a′y(k),az′(k)}
Ω (k)={ ωx′(k),ωy′(k),ωz′(k)}
Although above step S1021 and S1022 are current paces when identifying to the paces of user to user Data be acquired with pretreated step, but these steps can also be when carrying out initializing trained to step recognition model For obtaining training sample data set.Specifically, being under the original state of step recognition model, such as with smart phone Specific example is more than various knowings and doings in 200 minutes with the frequency acquisition of about 100Hz using the three axis accelerometer in mobile phone The acceleration and angular speed data under state are walked, carry out institute as above as raw sample data, and to these raw sample datas The pretreatment stated, to obtain training sample data.
Specifically, being directed to every kind of paces classification, the original paces sample data of certain amount is collected respectively, that is, a fixed number The acceleration and angular velocity data of amount, and these data are pre-processed.It then, will be corresponding to each section of complete paces Data time series and its paces state (paces classification) be used as one group of training data, step recognition model is instructed Practice, the initial value of the regularization parameter in model step recognition model and the width of Radial basis kernel function are determined by cross-validation method The initial value of degree, to obtain initial step recognition model f0(·)。
Referring back to Fig. 2.As shown in Fig. 2, step recognition unit will be at step S1022 through pre- at step S1023 The current pace data of processing input the initial step recognition model, thus identify the paces classification of user.Specifically, Regard the data in some regular length time before current time as a time window, time window is at any time constantly more Newly, the process of step recognition only carries out on actual time window.According to the step recognition model of least square method supporting vector machine The numerical value of time window parameter is set, inputs step recognition mould for pretreated paces data are passed through in actual time window Type, i.e., classification locating for the state of exportable current pace.If classification 1, representative is not detected paces, continues to identify, otherwise, Paces counter adds one, and output is as a result, wait a time interval t0After continue to identify.
Referring back to Fig. 1.As shown in Figure 1, successfully being identified at step S103 in user's Reusability step recognition model During the paces of oneself, the parameter updating unit (will be explained below) in step recognition device can will be used in a period of time The paces data that family is successfully identified are stored in local storage unit as training sample data, and after a certain time by institute There is the step recognition model of the training sample data input least square method supporting vector machine of accumulation to be trained, passes through cross validation Each parameter of the step recognition model of user is updated, thus to obtain the step recognition model for being exclusively used in the user.
Specifically, the paces rule due to different user is not exclusively the same, there is respectively different modes, use unification Mode carries out identification to the paces of different user and is easy to produce mistake.For example, if initial steps identification model is ft(), such as Oneself common cadence of a certain user of fruit is higher, then the paces for the type that skelps may be identified as jogging, or cannot identify This paces type, therefore it is easy to produce identification mistake.For this purpose, user is by being f in initial steps identification modeltOn () not The disconnected identification for carrying out paces, all automatically stores data after identifying successfully every time.After a period of time, when needing to carry out When model parameter updates, using its data re -training step recognition model that memory stores up for the previous period, to obtain newly Parameter, so that training new step recognition model is ft+1(·).Since the paces rule of different user is not exclusively the same, have Respectively different mode carries out identification using paces of the unified mode to different user and is easy to produce mistake, therefore, according to not Rule progress parameter with user updates the accuracy having using paces state recognition is improved.
Shown in Fig. 3 is the schematic illustration for step recognition method shown in FIG. 1.As shown, entirely identifying Journey is divided into initial training stage and online recognition stage.In initial training stage, by establishing least square method supporting vector machine Step recognition model, using original acquisition training paces data sample to the step recognition model carry out parameter initialization, It is f thus to obtain initial steps identification modelt(·).Subsequently entering the initial steps identification model is ftThe step recognition of () Application stage, i.e., online step recognition stage.In the online step recognition stage, user's current pace data are obtained in real time, are passed through The classification of user's current pace is obtained after identification, and paces data corresponding to the paces classification that success is identified are as sample number According to being stored in local sample database, and after a period of time has passed, using the sample data accumulated to least square branch It is f that the step recognition model for holding vector machine, which carries out re -training thus to obtain new step recognition model,t+1(·)。
Shown in Fig. 4 is according to the disclosure for carrying out pretreated process schematic to original paces data.Although This only shows 3-axis acceleration and three axis angular rates in terms of being in acquisition paces data, but can also acquire other numbers According to, such as other human body physiological datas, it is used for auxiliary judgment paces classification.
Shown in fig. 5 is the flow diagram according to the online step recognition of the disclosure.As shown in figure 5, in step S501 Place obtains the paces data of user in real time and pre-processes.Then, at step S502, judge acquired real-time paces Whether data are one of paces classification of defined.If it is not, then continuing paces data back to step S501 Real-time acquisition.If it is determined that being the paces of one of classification, then to the increased number of category paces at step S503 One, the thus lasting quantity of this classification paces of counting user, and export the data of such paces.Selectively, it walks here The counting cut down is also possible to the metering to the sum of all total class paces, so that user counts their own total step interior for a period of time Number.Then, at S504, judge whether that the parameter to current step recognition model is needed to be updated.The judgement is based on each Kind of standard carries out, for example, the number for every kind of paces classification being identified based on active user, being identified based on current pace Time span that the parameter of model has been used, the inquiry that step recognition device is issued the user with when being activated based on user Affirmative acknowledgement asked, etc..If it is determined that needing to be updated, then at step S505, step recognition device is using a timing Between in section as the paces data of the identified paces of training data sample cumulative to step recognition model progress re -training, and New model parameter is obtained, by cross validation thus to obtain the step recognition model for being exclusively used in the user.Carrying out parameter more After new, ask whether to continue step recognition at step S506, but judgement does not need to carry out step recognition, then directly Terminate step recognition process.If it is required, then carrying out the acquisition process of original paces data next time back to step S501.
Shown in fig. 6 is the structural schematic diagram according to the step recognition device 600 of the disclosure.As shown in fig. 6, step recognition Device 600 includes parameter set unit 601, paces data capture unit 602, pretreatment unit 603 and step recognition unit 604.Training sample data input in sample database 606 is inputted least square method supporting vector machine by parameter set unit 601 Step recognition model be trained, the initial of each parameter of initial step recognition model is determined from there through cross validation Value.Selectively, in the case where not having available sample data in sample database 606, parameter set unit 601 can also benefit Acquired in paces data capture unit 602 and by the pretreated multistage of pretreatment unit 603 and complete original paces data The step recognition model of corresponding training sample data input least square method supporting vector machine is trained, and is tested from there through intersection Card determines the initial value of each parameter of initial step recognition model.Step of user data capture unit 602 both can be offline Obtained under physical training condition in user's certain time various paces data (acceleration and angular speed data including paces and Paces categorical data) it is used as training sample data, it can also be in online step recognition for obtaining user for identified number According to.Paces data acquired in paces data capture unit 602 eliminate noise by the pretreatment of pretreatment unit 603.Paces are known Pretreated current pace data are inputted the initial step recognition model by other unit 604, thus identify the step of user Cut down classification.Although the database is not necessary to step recognition referring herein to sample database 606.Sample number It is a conventional memory unit according to library 606, is used to store the paces data for the user that step recognition unit 604 successfully identifies And its paces classification, and stored as training sample data.Know although sample database 606 can store from storage paces The paces data for the user that other unit 604 successfully identifies and its paces classification, but can also directly pass through I/O interface 607 obtain various existing paces training sample data from network.
In addition, step recognition device 600 may also include parameter updating unit 605, storage will be accumulated whithin a period of time The paces data that user in sample database 607 is successfully identified as training sample data input least square support to The step recognition model of amount machine is trained, and each parameter of the step recognition model of user is updated by cross validation, is thus obtained It must be exclusively used in the step recognition model of the user.The parameter updating unit 605, can be with although showing as independent component Only a update trigger element, parameter renewal process can be completed directly by parameter set unit 601, thus be simplified The structure and configuration of step recognition device 600.
Using the step recognition method and step recognition installation method and device of the disclosure, by collecting six pedestrian's hands The acceleration and angular velocity data when smart phone continuous walking are held or carry, the data that every pedestrian obtains include 20 groups, often One group of data is the acceleration acquired when walking on the straight line of one section of 30m, and test shows that the step recognition method of the disclosure exists Step recognition in these data has reached 98.8% accuracy.As it can be seen that method of disclosure can be improved the standard of step recognition True property simultaneously carries out personalized identification for different user.
The basic principle of the disclosure is described in conjunction with specific embodiments above, however, it is desirable to, it is noted that this field For those of ordinary skill, it is to be understood that the whole or any steps or component of disclosed method and device, Ke Yi Any computing device (including processor, storage medium etc.) perhaps in the network of computing device with hardware, firmware, software or Their combination is realized that this is that those of ordinary skill in the art use them in the case where having read the explanation of the disclosure Basic programming skill can be achieved with.
Therefore, the purpose of the disclosure can also by run on any computing device a program or batch processing come It realizes.The computing device can be well known fexible unit.Therefore, the purpose of the disclosure can also include only by offer The program product of the program code of the method or device is realized to realize.That is, such program product is also constituted The disclosure, and the storage medium for being stored with such program product also constitutes the disclosure.Obviously, the storage medium can be Any well known storage medium or any storage medium that developed in the future.
It may also be noted that in the device and method of the disclosure, it is clear that each component or each step are can to decompose And/or reconfigure.These decompose and/or reconfigure the equivalent scheme that should be regarded as the disclosure.Also, execute above-mentioned series The step of processing, can execute according to the sequence of explanation in chronological order naturally, but not need centainly sequentially in time It executes.Certain steps can execute parallel or independently of one another.
Above-mentioned specific embodiment does not constitute the limitation to disclosure protection scope.Those skilled in the art should be bright It is white, design requirement and other factors are depended on, various modifications, combination, sub-portfolio and substitution can occur.It is any Made modifications, equivalent substitutions and improvements etc., should be included in disclosure protection scope within the spirit and principle of the disclosure Within.

Claims (13)

1. a kind of step recognition method, comprising:
The step recognition model of multistage paces training sample data input least square method supporting vector machine is trained, and is passed through Cross validation determines the initial value of each parameter of initial step recognition model, forms initial step recognition model；
Obtain the current pace data in user's certain time；
The current pace data are pre-processed；
Pretreated current pace data are inputted into the initial step recognition model, thus identify the paces class of user Not；
The paces data that user is successfully identified in accumulation a period of time；And
The step recognition model of the paces data accumulated input least square method supporting vector machine is trained, is tested by intersecting Card updates each parameter of the step recognition model of user, thus to obtain the step recognition model for being exclusively used in the user.
2. step recognition method according to claim 1, wherein the kernel function of the least square method supporting vector machine is diameter To base kernel function.
3. step recognition method according to claim 1 or 2, wherein the original step of the multistage paces training sample data Cut down data be by user carry equipment in 3-axis acceleration sensor and three axis angular rate sensors obtain in real time three Axle acceleration and three axis angular rates.
4. step recognition method according to claim 3, wherein the original paces data become described by pretreatment Paces training sample data.
5. step recognition method according to claim 4, wherein the pretreatment is low-pass filtering.
6. step recognition method according to claim 5, wherein the initial value of the parameter of the step recognition model includes The initial value of the width of the initial value and Radial basis kernel function of regularization parameter.
7. step recognition method according to claim 2, wherein each section of correspondence in the original paces data of the multistage One in preset multiple paces classifications.
8. a kind of step recognition device, comprising:
Parameter set unit instructs the step recognition model of paces training sample data input least square method supporting vector machine Practice, the initial value of each parameter of initial step recognition model is determined from there through cross validation；
Step of user data capture unit obtains the current pace data in user's certain time；
Pretreatment unit pre-processes the current pace data；
Pretreated current pace data are inputted the initial step recognition model, thus identified by step recognition unit The paces classification of user；And
Parameter updating unit accumulates user's paces data being successfully identified and the successful knowledge that will be accumulated in a period of time Other paces data are trained as the step recognition model of training sample data input least square method supporting vector machine, are passed through Cross validation updates each parameter of the step recognition model of user, thus to obtain the step recognition model for being exclusively used in the user.
9. step recognition device according to claim 8, wherein the kernel function of the least square method supporting vector machine is diameter To base kernel function.
10. step recognition device according to claim 8 or claim 9, wherein the paces data are by the step of user The 3-axis acceleration and three axis that 3-axis acceleration sensor and three axis angular rate sensors in data capture unit obtain in real time Angular speed.
11. step recognition device according to claim 8, wherein the pretreatment includes carrying out low pass filtered to paces data Wave.
12. step recognition device according to claim 8, wherein each parameter of the initial step recognition model is first Initial value is the initial value of the initial value of regularization parameter and the width of Radial basis kernel function.
13. step recognition device according to claim 8, wherein multistage instruction corresponding with completely original paces data Each sample in white silk sample data corresponds to one in preset multiple paces classifications.
CN201610639204.9A 2016-08-05 2016-08-05 Step recognition method and device Active CN106264545B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610639204.9A CN106264545B (en) 2016-08-05 2016-08-05 Step recognition method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610639204.9A CN106264545B (en) 2016-08-05 2016-08-05 Step recognition method and device

Publications (2)

Publication Number Publication Date
CN106264545A CN106264545A (en) 2017-01-04
CN106264545B true CN106264545B (en) 2019-06-25

Family

ID=57665662

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610639204.9A Active CN106264545B (en) 2016-08-05 2016-08-05 Step recognition method and device

Country Status (1)

Country Link
CN (1) CN106264545B (en)

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106725512B (en) * 2017-02-22 2020-09-08 安徽华米信息科技有限公司 Motion monitoring method and device and wearable equipment
CN109426829A (en) * 2017-08-28 2019-03-05 中兴通讯股份有限公司 A kind of estimation method, device and the computer readable storage medium of user's height
CN107998635B (en) * 2018-01-25 2023-01-03 桂林格物电子科技有限责任公司 Intelligent football pace training system and method
CN109453501A (en) * 2018-10-25 2019-03-12 惠州市德赛工业研究院有限公司 Ball training data processing method, equipment
CN112990869B (en) * 2021-03-12 2023-08-22 盐城师范学院 Intelligent student attendance management device

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104268577A (en) * 2014-06-27 2015-01-07 大连理工大学 Human body behavior identification method based on inertial sensor
CN104287703A (en) * 2013-06-03 2015-01-21 飞比特公司 Use of gyroscopes in personal fitness tracking devices
CN104729507A (en) * 2015-04-13 2015-06-24 大连理工大学 Gait recognition method based on inertial sensor
CN104913772A (en) * 2015-06-05 2015-09-16 北京信息科技大学 Pedestrian motion detecting method based on leg pose information
CN104323780B (en) * 2014-10-30 2016-07-06 上海交通大学 Pedestrian's gait classification system and method based on support vector machine

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104287703A (en) * 2013-06-03 2015-01-21 飞比特公司 Use of gyroscopes in personal fitness tracking devices
CN104268577A (en) * 2014-06-27 2015-01-07 大连理工大学 Human body behavior identification method based on inertial sensor
CN104323780B (en) * 2014-10-30 2016-07-06 上海交通大学 Pedestrian's gait classification system and method based on support vector machine
CN104729507A (en) * 2015-04-13 2015-06-24 大连理工大学 Gait recognition method based on inertial sensor
CN104913772A (en) * 2015-06-05 2015-09-16 北京信息科技大学 Pedestrian motion detecting method based on leg pose information

Also Published As

Publication number Publication date
CN106264545A (en) 2017-01-04

Similar Documents

Publication Publication Date Title
CN106264545B (en) Step recognition method and device
Kim Interpretable and accurate convolutional neural networks for human activity recognition
CN110309706A (en) Face critical point detection method, apparatus, computer equipment and storage medium
Rau et al. An empirical Bayesian method for estimating biological networks from temporal microarray data
US20170105683A1 (en) System and method of adaptive interpretation of ecg waveforms
Zhang et al. Human activity recognition based on time series analysis using U-Net
CN104484644B (en) A kind of gesture identification method and device
CN116994709B (en) Personalized diet and exercise recommendation method and system and electronic equipment
CN109101938A (en) A kind of multi-tag age estimation method based on convolutional neural networks
CN107958230A (en) Facial expression recognizing method and device
JP2013041323A (en) Action identification device, action identification method, and program
CN106537387B (en) Retrieval/storage image associated with event
WO2015018780A4 (en) Method, device and system for annotated capture of sensor data and crowd modelling of activities
CN109770912A (en) A kind of abnormal gait classification method based on depth convolutional neural networks
CN108960269A (en) Characteristic-acquisition method, device and the calculating equipment of data set
CN106528655A (en) Text subject recognition method and device
CN109614904A (en) A kind of activity recognition method of the Multi-sensor Fusion based on Shapelet
Shi et al. Smartphone-aided human activity recognition method using residual multi-layer perceptron
CN103839047A (en) Human motion track recognition method and device
CN109740418A (en) A kind of Yoga action identification method based on multiple acceleration transducers
Permatasari et al. Inertial sensor fusion for gait recognition with symmetric positive definite Gaussian kernels analysis
CN107463689A (en) Generation method, moving state identification method and the terminal in motion characteristic data storehouse
CN112370058A (en) Method for identifying and monitoring emotion of user based on mobile terminal
Priyadharshini et al. Classification and analysis of human activities
Wilson et al. Domain Adaptation Under Behavioral and Temporal Shifts for Natural Time Series Mobile Activity Recognition

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant