CN106339071A - Method and device for identifying behaviors - Google Patents
Method and device for identifying behaviors Download PDFInfo
- Publication number
- CN106339071A CN106339071A CN201510400436.4A CN201510400436A CN106339071A CN 106339071 A CN106339071 A CN 106339071A CN 201510400436 A CN201510400436 A CN 201510400436A CN 106339071 A CN106339071 A CN 106339071A
- Authority
- CN
- China
- Prior art keywords
- layer
- approximation coefficient
- behavior
- wavelet
- behavioral 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
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
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01P—MEASURING LINEAR OR ANGULAR SPEED, ACCELERATION, DECELERATION, OR SHOCK; INDICATING PRESENCE, ABSENCE, OR DIRECTION, OF MOVEMENT
- G01P13/00—Indicating or recording presence, absence, or direction, of movement
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
- G06F3/03—Arrangements for converting the position or the displacement of a member into a coded form
- G06F3/033—Pointing devices displaced or positioned by the user, e.g. mice, trackballs, pens or joysticks; Accessories therefor
- G06F3/0346—Pointing devices displaced or positioned by the user, e.g. mice, trackballs, pens or joysticks; Accessories therefor with detection of the device orientation or free movement in a 3D space, e.g. 3D mice, 6-DOF [six degrees of freedom] pointers using gyroscopes, accelerometers or tilt-sensors
Abstract
The invention discloses a method for identifying behaviors comprising following steps: obtaining behavior data corresponding to behaviors to be identified; obtaining wavelet characteristic numbers of the behavior data; determining the behaviors to be identified according to the wavelet characteristic numbers. The invention also discloses a device for identifying behaviors.
Description
Technical field
The present invention relates to Activity recognition technology is and in particular to a kind of Activity recognition method and apparatus.
Background technology
In recent years, with the popularization of the mobile devices such as mobile phone, intelligent watch, Intelligent bracelet, set based on movement
Standby Human bodys' response technology becomes study hotspot.Wherein, usual Activity recognition method is as described below:
The behavioral data gathering human body by the sensor of configuration in mobile device is sample data, to gathered
Sample data carries out feature preferably, and identifies human body behavior by the behavior model set up.Described human body row
For being static, the human body everyday actions such as walk, run, jumping, sitting, standing.These human bodies are daily dynamic
Make due to action between differing greatly, so be easier identification and recognition accuracy is higher.It is contemplated that
In actual applications, human body also occurs such as upstairs with downstairs, is careful, hurries up and remains where one is etc. so
The action of subdivision, current Activity recognition method is not high, by mistake for the accuracy identifying this subdivision action
Identification probability is larger.
Content of the invention
For solving existing technical problem, the embodiment of the present invention provides a kind of Activity recognition method and apparatus,
The accuracy to subdivision action recognition can be improved, reduce probability of misrecognition.
The technical scheme of the embodiment of the present invention is achieved in that
Embodiments provide a kind of Activity recognition method, methods described includes:
Obtain the corresponding behavioral data of behavior to be identified；
Obtain the wavelet character value of described behavioral data；
According to described wavelet character value, determine described behavior to be identified.
In aforementioned schemes, the described acquisition corresponding behavioral data of behavior to be identified, comprising:
Human body is gathered when there is described behavior to be identified in predetermined collection period with predetermined sample frequency
The acceleration signal producing；
Determine that any acceleration signal being collected is described behavioral data.
In aforementioned schemes, the described wavelet character value obtaining described behavioral data, comprising:
Described behavioral data is carried out n-layer wavelet decomposition, obtains every layer of approximation coefficient, described approximation coefficient
Including high-frequency approximation coefficient and low-frequency approximation coefficient, n is the wavelet decomposition number of plies, is just whole more than or equal to 1
Number；
Retain the low-frequency approximation coefficient of at least one predetermined layer；
According to the low-frequency approximation coefficient of at least one predetermined layer being retained, determine at least one predetermined layer described
In every layer of wavelet character value.
In aforementioned schemes, methods described includes:
Input and become to small echo set in advance to behavioral data, predetermined wavelet decomposition number of plies n=6 described in major general
Carry out 6 layers of wavelet decomposition in exchange the letters number, obtain every layer of high-frequency approximation coefficient and low-frequency approximation coefficient；
Retain the low-frequency approximation coefficient of the 2nd layer, the 3rd layer and the 4th layer；
Using the low-frequency approximation coefficient of the 2nd layer, the 3rd layer and the 4th layer, calculate following wherein at least one special
Levy parameter: every layer of wavelet energy value, the mean size of little crest value, small echo in all layers being retained
Peak value number；
The characteristic parameter determining calculated every layer is the described wavelet character value of equivalent layer.
In aforementioned schemes, described according to described wavelet character value, determine described behavior to be identified, comprising:
Determine all layers characteristic parameter be training in advance sorter model input signal；
Described input signal is inputted to described sorter model；
By the computing that described sorter model carries out preset algorithm to described input signal identify described in wait to know
Other behavior.
The embodiment of the present invention additionally provides a kind of Activity recognition equipment, and described equipment includes:
First acquisition unit, for obtaining the corresponding behavioral data of behavior to be identified；
Second acquisition unit, for obtaining the wavelet character value of described behavioral data；
First determining unit, for according to described wavelet character value, determining described behavior to be identified.
In aforementioned schemes, described first acquisition unit, it is additionally operable to:
Human body is gathered when there is described behavior to be identified in predetermined collection period with predetermined sample frequency
The acceleration signal producing；
Determine that any acceleration signal being collected is described behavioral data.
In aforementioned schemes, described second acquisition unit, it is additionally operable to:
Described behavioral data is carried out n-layer wavelet decomposition, obtains every layer of approximation coefficient, described approximation coefficient
Including high-frequency approximation coefficient and low-frequency approximation coefficient, n is the wavelet decomposition number of plies, is just whole more than or equal to 1
Number；
Retain the low-frequency approximation coefficient of at least one predetermined layer；
According to the low-frequency approximation coefficient of at least one predetermined layer being retained, determine at least one predetermined layer described
In every layer of wavelet character value.
In aforementioned schemes, described second acquisition unit, it is additionally operable to:
Input and become to small echo set in advance to behavioral data, predetermined wavelet decomposition number of plies n=6 described in major general
Carry out 6 layers of wavelet decomposition in exchange the letters number, obtain every layer of high-frequency approximation coefficient and low-frequency approximation coefficient；
Retain the low-frequency approximation coefficient of the 2nd layer, the 3rd layer and the 4th layer；
Using the low-frequency approximation coefficient of the 2nd layer, the 3rd layer and the 4th layer, calculate following wherein at least one special
Levy parameter: every layer of wavelet energy value, the mean size of little crest value, small echo in all layers being retained
Peak value number；
The characteristic parameter determining calculated every layer is the described wavelet character value of equivalent layer.
In aforementioned schemes, described first determining unit, it is additionally operable to:
Determine all layers characteristic parameter be training in advance sorter model input signal；
Described input signal is inputted to described sorter model；
By the computing that described sorter model carries out preset algorithm to described input signal identify described in wait to know
Other behavior.
Activity recognition method and apparatus provided in an embodiment of the present invention, obtains behavior to be identified corresponding behavior number
According to；Obtain the wavelet character value of described behavioral data；According to described wavelet character value, determine described to be identified
Behavior.Wavelet character value is the characteristic parameter on frequency domain, carries out human body from frequency domain angle and segments action behavior
Identification, can improve identification accuracy, reduce probability of misrecognition.
Brief description
Fig. 1 realizes schematic flow sheet for the Activity recognition method of the embodiment of the present invention；
Fig. 2 realizes schematic flow sheet for the wavelet character value of the acquisition behavioral data of the embodiment of the present invention；
Fig. 3 is the coordinate schematic diagram with sample amplitude for the sample sequence number of the embodiment of the present invention；
Fig. 4 is the composition structural representation of the Activity recognition equipment of the embodiment of the present invention.
Specific embodiment
Below in conjunction with accompanying drawing to a preferred embodiment of the present invention will be described in detail it will be appreciated that following described
Bright preferred embodiment is merely to illustrate and explains the present invention, is not intended to limit the present invention.
The Activity recognition method of the embodiment of the present invention, is applied in a mobile device, this mobile device can be with handss
Machine, intelligent watch, Intelligent bracelet, intelligent glasses or panel computer pad, personal digital assistant
Pda, electronic reader etc., are not specifically limited herein.
Fig. 1 realizes schematic flow sheet for the Activity recognition method of the embodiment of the present invention；As shown in figure 1, institute
The method of stating includes:
Step 11: obtain the corresponding behavioral data of behavior to be identified；
Described behavior to be identified is at least two subdivision action behaviors, for example, move with both subdivisions downstairs upstairs
Make, these three subdivision actions of being also for example careful, hurry up and remain where one is, also for example upstairs, go downstairs and walking
These three subdivision actions.The technical scheme of the embodiment of the present invention is that the identification of how high accuracy is every kind of thin
Divide action behavior.
In the present embodiment, included as a example upstairs and downstairs by behavior to be identified, described behavioral data exists for human body
The acceleration signal producing under certain behavior, such as the acceleration producing when upstairs, downstairs when produce plus
Speed, the preferably synthetic acceleration signal of this acceleration signal.This acceleration signal can be by being built in movement
Acceleration transducer in equipment and sense.When this acceleration transducer is 3-axis acceleration sensor,
Sensed by this 3-axis acceleration sensor is in x direction, y direction and z under xyz three-dimensional system of coordinate
Acceleration on direction, and the acceleration on these three directions is carried out synthesizing the described resultant acceleration obtaining
Signal.
Generally, acceleration transducer with predetermined sample frequency using the acceleration signal being under certain behavior,
As being as a example 50 times/s by frequency acquisition, behavioral data 1s under there is current behavior for the collection human body, at this
The behavioral data collecting 50 times/s*1s=50 in 1s altogether is acceleration signal, regards this 1s as predetermined one
The individual sampling period.In the present embodiment, the described acquisition corresponding behavioral data of behavior to be identified can be with predetermined
Sample frequency gather human body acceleration of producing when there is described behavior to be identified in predetermined collection period
Degree signal；Determine that the arbitrarily individual acceleration signal being collected is described behavioral data.And by follow-up skill
Which kind of behavior the art scheme identification behavior is specially as being the still behavior downstairs of behavior upstairs.Wherein, described adopt
Sample frequency flexibly can be set according to practical situation with the sampling period.
Step 12: obtain the wavelet character value of described behavioral data；
Here, described wavelet character value includes following a kind of at least within characteristic parameter: behavioral data is through small echo
Wavelet energy after decomposition, the mean size of little crest value, little crest value number.
As shown in Fig. 2 described step 12 further includes:
Step 121: described behavioral data is carried out n-layer wavelet decomposition, obtains every layer of approximation coefficient, institute
State approximation coefficient and include high-frequency approximation coefficient and low-frequency approximation coefficient；
Here, n is the positive integer more than or equal to 1, can rule of thumb and flexibly set, the maximum of usual n
Value needs the signal meeting in this maximum layer when behavioral data is decomposed into maximum layer can be expressed as one
Complete behavior act is such as expressed as a complete action upstairs.In the present embodiment, preferably wavelet decomposition layer
Number n=6.In pre-configured built-in function, call wavelet transform function, and will be by acceleration sensing
Current behavior data, the wavelet basis function chosen in advance and wavelet decomposition number of plies n conduct that device collects
Three input signals of wavelet transform function input to wavelet transform function, and the output of wavelet transform function is at once
For data through wavelet decomposition for every layer of n-layer approximation coefficient.It is near that every layer of approximation coefficient at least includes low frequency
Like coefficient and high-frequency approximation coefficient；Wherein, low-frequency approximation coefficient be broken down into this layer signal intensity slow
Behavioral data can be presented as the outline portion of behavior signal, high-frequency approximation coefficient is the letter being broken down into this layer
Number changing rapid behavioral data can be presented as the detail section of behavior signal.In this step, behavioral data is entered
Row n-layer wavelet decomposition will behavioral data through on wavelet transformation to frequency domain.
Preferably employ the wavelet transformation based on multiresolution analysis in the present embodiment and little wavelength-division is carried out to behavioral data
Solution, selected wavelet basis function is Haar wavelet transform.With regard to aforesaid high frequency, low-frequency approximation coefficient and little
The concept of ripple basic function and purposes refer to existing related description, and this repeats no more.
Step 122: retain the low-frequency approximation coefficient of at least one predetermined layer；
Here, because high-frequency approximation coefficient is the detail section of signal, detail section is generally viewed as noise,
All layers of high-frequency approximation coefficient is given up in the embodiment of the present invention；Retain the low-frequency approximation of at least one predetermined layer
Coefficient., after behavioral data carries out the wavelet decomposition of n=6 layer, the obtained 1st～6 taking n=6 as a example
In the low-frequency approximation coefficient of layer, choose the 2nd, 3,4 layers of low-frequency approximation coefficient as being retained at least
The low-frequency approximation coefficient of one predetermined layer.Retain the low-frequency approximation coefficient of which or which layer generally according to
Experience and obtain.
Step 123: according to the low-frequency approximation coefficient of at least one predetermined layer being retained, described in determination at least
Every layer of wavelet character value in one predetermined layer；
Described wavelet character value includes following a kind of at least within: each after wavelet decomposition of behavioral data is pre-
The wavelet energy of given layer, the mean size of little crest value, little crest value number.In the present embodiment, with small echo
As a example eigenvalue includes aforementioned three parameters.Generally, acceleration transducer with predetermined sample frequency using at
Acceleration signal under certain behavior, is as a example 50 times/s such as by frequency acquisition, collection human body is occurring to work as
Move ahead under behavioral data 1s, the behavioral data collecting 50 times/s*1s=50 in this 1s altogether adds
Rate signal, this 50 acceleration signal correspondences are designated the 1st sample signal, the 2nd sample signal ...
50th sample signal.Xoy axis coordinate system as shown in Figure 3, the serial number of x-axis representative sample signal
From 1 to 50, vertical coordinate represents the amplitude of this 50 sample signal one of predetermined layer after wavelet decomposition
I.e. the amplitude of low-frequency approximation coefficient of this layer, be centrifugal pump, represented with black circle.In figure 3, first
Find out and have multiple amplitude peaks in so many corresponding amplitudes of sample signal, preset the first amplitude
Threshold value, as shown in Fig. 3 parallel to the straight line of x-axis, the previous sample point of certain sample point with latter one
When each corresponding amplitude of sample point is all more than the first amplitude threshold, regard the corresponding amplitude of this sample point as being somebody's turn to do
The little crest value of layer, can find out all little crest in this 50 sample signals of this layer according to the method
The number of value, and the size of all little crest values.The mean size of the little crest value of this layer is equal to all little
Crest value sum is divided by the total number of the little crest value of this layer.The wavelet energy of this layer is equal to behavioral data through little
The quadratic sum of the wavelet coefficient that wave conversion obtains.For wavelet energy, the mean size of little crest value and small echo
The computational methods of peak value number, the concept of wavelet coefficient specifically refers to existing related description, this time no longer superfluous
State.
Wherein, the mean size of little crest value and little crest value number are called little crest feature.Wavelet energy divides
Cloth feature can reflect upstairs, downstairs Energy distribution situation on different wavelet decomposition layers for both behaviors, little
Crest feature can reflect the amplitude of the acceleration signal that human body produces under both behaviors, both features
In conjunction with than only identifying that human body behavior compares by time domain specification such as mean square deviation, variance etc. in correlation technique,
The recognition accuracy of behavior can be significantly improved.Additionally, wavelet energy, the mean size of little crest value and little
Crest value number is the characteristic parameter on frequency domain, and the embodiment of the present invention passes through a characteristic parameter on frequency domain
Or at least two the combination of characteristic parameter realize identification to human body behavior.
Step 13: according to described wavelet character value, determine described behavior to be identified.
In the present embodiment, using the decision tree in data mining, support vector machine (svm, support vector
Machine) or very fast learning machine scheduling algorithm, what a sorter model of training in advance, by described at least one
In predetermined layer, every layer of wavelet energy, the mean size of little crest value, little crest value number are as this grader
The input signal of model inputs to sorter model such as by every layer calculated in the 2nd, 3,4 layers of small echo
Energy, the mean size of little crest value, little crest value number input to sorter model as input signal,
Described row to be identified is identified by the computing that described sorter model carries out preset algorithm to described input signal
For；Described preset algorithm can be decision tree, svm or very fast learning machine algorithm.Pass through this grader mould
Type may recognize which kind of behavior of behavior human body to be identified.
Wherein, the concrete formula of described sorter model and its parameters refer to existing related description, this
Place does not repeat.So that wavelet decomposition number of plies n=6, predetermined layer are the 2nd～4 layer as a example, in the instruction of sorter model
During white silk, the input of sorter model includes behavioral data with behavior label, behavior data through n
2nd～4 layer every layer of wavelet energy, the mean size of little crest value and little crest value after layer wavelet decomposition
Number, using these input signals, the parameters of sorter model are adjusted correspondingly, when grader mould
The parameters of type adjust consistent with the behavior that behavior label indicates to the output result making sorter model
When, illustrate that now sorter model trains.Specifically, the process of a sorter model is trained to include:
With predetermined frequency acquisition, behavioral data in behavior upstairs and repeatedly is in by sensor multi collect
Collection is in the behavioral data of behavior downstairs, and will be characterized as the behavioral data record of same behavior same
In individual file.Give an example, be that as a example 50 times/s, continuous collecting human body is in behavior upstairs by frequency acquisition
Behavioral data 10s, collect the behavior of the 50*10=500 of the behavior upstairs that may be characterized as in this 10s altogether
Data, and this 500 behavioral datas are saved in the first file, this first file record is that human body exists
It is in the behavioral data upstairs being collected during behavior upstairs.Behavior in behavior downstairs for the continuous collecting human body
Data 5s, collects the behavioral data of 50*5=250 of the behavior downstairs that may be characterized as in this 5s altogether, and
This 250 behavioral datas are saved in the second file, this second file record be in human body under being in
The behavioral data downstairs being collected during building behavior.It should be noted that those skilled in the art should and know,
If being considered as the variety classes of behavior upstairs and downstairs, then the number of the file of the different behavioral datas that are stored with
Amount should be identical with the species of behavior to be identified.From the foregoing, it will be observed that the row storing in the first file and the second file
It is (the corresponding behavior of certain behavioral data can be distinguished by behavior label with behavior label for data
For which kind of behavior), that is, in the first file, the behavioral data of storage corresponds to behavior upstairs, stores in the second file
Behavioral data correspond to behavior downstairs.The all behavioral datas storing in both of these documents are carried out n=6 layer
Wavelet decomposition, obtain each behavioral data in the 2nd～4 layer every layer of wavelet energy, little crest value flat
All size and little crest value numbers, by some or all behavioral datas with behavior label, this part or
All behavioral datas after n=6 layer wavelet decomposition the 2nd～4 layer every layer of wavelet character value as disaggregated model
Training dataset, and training dataset is substituting to sorter model to train the parameters of this model, when point
The parameters of class device model are adjusted to the behavior phase making the output result of sorter model be indicated with behavior label
When consistent, illustrate that now sorter model trains.After sorter model trains, when by accelerating
When degree sensor acquisition is to certain behavioral data, will carry out according to abovementioned steps 11～13, with by training
Sorter model identifying the corresponding behavior of behavior data.Concrete to the parameters of sorter model
The concept of training process and training dataset refers to existing related description, does not repeat herein.
In aforementioned schemes, it is that considering in addition should in reality to identify the explanation carrying out upstairs and as a example going downstairs
With in also there are level land between stair and stair, the present embodiment can also identify upstairs, downstairs and walking
It is also possible to identification is careful, hurries up and is remained where one is, these three segment action behavior to these three subdivision action behaviors.
The advantage of the technical scheme of the embodiment of the present invention is:
1) by the behavioral data collecting through, on wavelet decomposition transform to frequency domain, carrying out human body row from frequency domain angle
For identification, with simple compared with time domain angle carries out Human bodys' response, identification accuracy can be improved, subtract
Few probability of misrecognition；
2) by inputting the combination of a characteristic parameter on frequency domain or at least two characteristic parameters to instructing in advance
In the sorter model perfected, human body behavior is identified with the sorter model good by training in advance；Wherein
The training dataset that this sorter model adopts in training process in advance is at least one feature ginseng on frequency domain
Number, makes the accuracy of the sorter model training improve, to be identified by the higher sorter model of accuracy
The human body behavior of subdivision, can improve recognition accuracy, reduce probability of misrecognition.
Based on aforesaid Activity recognition method, the embodiment of the present invention additionally provides a kind of Activity recognition equipment, such as
Shown in Fig. 4, described equipment includes: first acquisition unit 401, second acquisition unit 402, first determine single
Unit 403；Wherein,
First acquisition unit 401, for obtaining the corresponding behavioral data of behavior to be identified；
Second acquisition unit 402, for obtaining the wavelet character value of described behavioral data；
First determining unit 403, for according to described wavelet character value, determining described behavior to be identified.
Wherein, described first acquisition unit 401, is additionally operable to: with predetermined sample frequency in predetermined collection
The acceleration signal that in cycle, collection human body produces in the described behavior to be identified of generation；Determination is collected
Arbitrarily acceleration signal is described behavioral data.
Wherein, described second acquisition unit 402, is additionally operable to: described behavioral data is carried out the little wavelength-division of n-layer
Solution, obtains every layer of approximation coefficient, described approximation coefficient includes high-frequency approximation coefficient and low-frequency approximation coefficient,
N is the wavelet decomposition number of plies, is positive integer more than or equal to 1；Retain the low-frequency approximation of at least one predetermined layer
Coefficient；According to the low-frequency approximation coefficient of at least one predetermined layer being retained, determine described at least one make a reservation for
Every layer of wavelet character value in layer.Further, described second acquisition unit 402 is to behavior number described in major general
Input and carry out 6 layers of small echo to wavelet transform function set in advance according to, predetermined wavelet decomposition number of plies n=6
Decompose, obtain every layer of high-frequency approximation coefficient and low-frequency approximation coefficient；Retain the 2nd layer, the 3rd layer and the 4th
The low-frequency approximation coefficient of layer；Using the low-frequency approximation coefficient of the 2nd layer, the 3rd layer and the 4th layer, calculate following
Wherein at least one characteristic parameter: every layer of wavelet energy value in all layers being retained, little crest value
Mean size, little crest value number；Determine that every layer calculated of characteristic parameter is described little for equivalent layer
Baud value indicative.
Wherein, described first determining unit 403, is additionally operable to: determines all layers of characteristic parameter for instructing in advance
The input signal of the sorter model practiced；Described input signal is inputted to described sorter model；By institute
State the computing that sorter model carries out preset algorithm to described input signal and identify described behavior to be identified.
For realizing said method, the embodiment of the present invention additionally provides a kind of Activity recognition equipment, due to this equipment
The principle of solve problem is similar to method, therefore, the implementation process of Activity recognition equipment and implementation principle
To describe referring to the implementation process of preceding method and implementation principle, repeat no more in place of repetition.
In actual applications, described first acquisition unit 401, second acquisition unit 402, the first determining unit
403 all can by CPU (cpu, central processing unit) or Digital Signal Processing (dsp,
Digital signal processor) or microprocessor (mpu, micro processor unit) or scene can
Program gate array (fpga, field programmable gate array) etc. to realize.
It will be appreciated by those skilled in the art that each processing module in the Activity recognition equipment shown in Fig. 4
Realize function to can refer to the associated description of aforementioned Activity recognition method and understand.Those skilled in the art should manage
Solution, in the Activity recognition equipment shown in Fig. 4, the function of each processing unit can be by running on the journey on processor
Sequence and realize, also can be realized by specific logic circuit.
Those skilled in the art are it should be appreciated that embodiments of the invention can be provided as method, system or meter
Calculation machine program product.Therefore, the present invention can using hardware embodiment, software implementation or combine software and
The form of the embodiment of hardware aspect.And, the present invention can adopt and wherein include calculating one or more
Computer-usable storage medium (including but not limited to disk memory and the optical storage of machine usable program code
Device etc.) the upper computer program implemented form.
The present invention is with reference to method according to embodiments of the present invention, equipment (system) and computer program
Flow chart and/or block diagram describing.It should be understood that can be by computer program instructions flowchart and/or side
Each flow process in block diagram and/or the knot of the flow process in square frame and flow chart and/or block diagram and/or square frame
Close.Can provide these computer program instructions to general purpose computer, special-purpose computer, Embedded Processor or
The processor of other programmable data processing device with produce a machine so that by computer or other can
The instruction of the computing device of programming data processing equipment produces for realizing in one flow process or multiple of flow chart
The device of the function of specifying in flow process and/or one square frame of block diagram or multiple square frame.
These computer program instructions may be alternatively stored in and can guide computer or other programmable data processing device
So that being stored in this computer-readable memory in the computer-readable memory working in a specific way
Instruction produces the manufacture including command device, and this command device is realized in one flow process of flow chart or multiple stream
The function of specifying in journey and/or one square frame of block diagram or multiple square frame.
These computer program instructions also can be loaded in computer or other programmable data processing device, makes
Obtain and series of operation steps is executed on computer or other programmable devices to produce computer implemented place
Reason, thus the instruction of execution is provided for realizing in flow chart one on computer or other programmable devices
The step of the function of specifying in flow process or multiple flow process and/or one square frame of block diagram or multiple square frame.
The above, only presently preferred embodiments of the present invention, it is not intended to limit the protection model of the present invention
Enclose.
Claims (10)
1. a kind of Activity recognition method is it is characterised in that methods described includes:
Obtain the corresponding behavioral data of behavior to be identified；
Obtain the wavelet character value of described behavioral data；
According to described wavelet character value, determine described behavior to be identified.
2. method according to claim 1 is it is characterised in that described acquisition behavior to be identified is corresponding
Behavioral data, comprising:
Human body is gathered when there is described behavior to be identified in predetermined collection period with predetermined sample frequency
The acceleration signal producing；
Determine that any acceleration signal being collected is described behavioral data.
3. method according to claim 1 and 2 is it is characterised in that the described behavioral data of described acquisition
Wavelet character value, comprising:
Described behavioral data is carried out n-layer wavelet decomposition, obtains every layer of approximation coefficient, described approximation coefficient
Including high-frequency approximation coefficient and low-frequency approximation coefficient, n is the wavelet decomposition number of plies, is just whole more than or equal to 1
Number；
Retain the low-frequency approximation coefficient of at least one predetermined layer；
According to the low-frequency approximation coefficient of at least one predetermined layer being retained, determine at least one predetermined layer described
In every layer of wavelet character value.
4. method according to claim 3 is it is characterised in that methods described includes:
Input and become to small echo set in advance to behavioral data, predetermined wavelet decomposition number of plies n=6 described in major general
Carry out 6 layers of wavelet decomposition in exchange the letters number, obtain every layer of high-frequency approximation coefficient and low-frequency approximation coefficient；
Retain the low-frequency approximation coefficient of the 2nd layer, the 3rd layer and the 4th layer；
Using the low-frequency approximation coefficient of the 2nd layer, the 3rd layer and the 4th layer, calculate following wherein at least one special
Levy parameter: every layer of wavelet energy value, the mean size of little crest value, small echo in all layers being retained
Peak value number；
The characteristic parameter determining calculated every layer is the described wavelet character value of equivalent layer.
5. method according to claim 4 it is characterised in that described according to described wavelet character value,
Determine described behavior to be identified, comprising:
Determine all layers characteristic parameter be training in advance sorter model input signal；
Described input signal is inputted to described sorter model；
By the computing that described sorter model carries out preset algorithm to described input signal identify described in wait to know
Other behavior.
6. a kind of Activity recognition equipment is it is characterised in that described equipment includes:
First acquisition unit, for obtaining the corresponding behavioral data of behavior to be identified；
Second acquisition unit, for obtaining the wavelet character value of described behavioral data；
First determining unit, for according to described wavelet character value, determining described behavior to be identified.
7. equipment according to claim 6, it is characterised in that described first acquisition unit, is additionally operable to:
Human body is gathered when there is described behavior to be identified in predetermined collection period with predetermined sample frequency
The acceleration signal producing；
Determine that any acceleration signal being collected is described behavioral data.
8. the equipment according to claim 6 or 7, it is characterised in that described second acquisition unit, is gone back
For:
Described behavioral data is carried out n-layer wavelet decomposition, obtains every layer of approximation coefficient, described approximation coefficient
Including high-frequency approximation coefficient and low-frequency approximation coefficient, n is the wavelet decomposition number of plies, is just whole more than or equal to 1
Number；
Retain the low-frequency approximation coefficient of at least one predetermined layer；
According to the low-frequency approximation coefficient of at least one predetermined layer being retained, determine at least one predetermined layer described
In every layer of wavelet character value.
9. equipment according to claim 8, it is characterised in that described second acquisition unit, is additionally operable to:
Input and become to small echo set in advance to behavioral data, predetermined wavelet decomposition number of plies n=6 described in major general
Carry out 6 layers of wavelet decomposition in exchange the letters number, obtain every layer of high-frequency approximation coefficient and low-frequency approximation coefficient；
Retain the low-frequency approximation coefficient of the 2nd layer, the 3rd layer and the 4th layer；
Using the low-frequency approximation coefficient of the 2nd layer, the 3rd layer and the 4th layer, calculate following wherein at least one special
Levy parameter: every layer of wavelet energy value, the mean size of little crest value, small echo in all layers being retained
Peak value number；
The characteristic parameter determining calculated every layer is the described wavelet character value of equivalent layer.
10. equipment according to claim 9, it is characterised in that described first determining unit, is also used
In:
Determine all layers characteristic parameter be training in advance sorter model input signal；
Described input signal is inputted to described sorter model；
By the computing that described sorter model carries out preset algorithm to described input signal identify described in wait to know
Other behavior.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510400436.4A CN106339071A (en) | 2015-07-08 | 2015-07-08 | Method and device for identifying behaviors |
| PCT/CN2015/089058 WO2017004880A1 (en) | 2015-07-08 | 2015-09-07 | Method, device for behavior recognition and computer storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510400436.4A CN106339071A (en) | 2015-07-08 | 2015-07-08 | Method and device for identifying behaviors |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106339071A true CN106339071A (en) | 2017-01-18 |
Family
ID=57684611
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510400436.4A Pending CN106339071A (en) | 2015-07-08 | 2015-07-08 | Method and device for identifying behaviors |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN106339071A (en) |
| WO (1) | WO2017004880A1 (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107862276A (en) * | 2017-11-01 | 2018-03-30 | 北京旋极信息技术股份有限公司 | A kind of Activity recognition method and terminal |
| WO2018232557A1 (en) * | 2017-06-19 | 2018-12-27 | 深圳市酷浪云计算有限公司 | Exercise movement recognition method and apparatus, and electronic device |
| CN109241926A (en) * | 2018-09-19 | 2019-01-18 | 重庆信络威科技有限公司 | A kind of driving behavior recognition methods based on wavelet analysis |
| CN110017834A (en) * | 2019-04-15 | 2019-07-16 | 歌尔科技有限公司 | Method, equipment and storage medium are determined using object |
| CN113030890A (en) * | 2019-12-24 | 2021-06-25 | 深圳市大富科技股份有限公司 | Target identification method and device based on vehicle-mounted radar |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112383488B (en) * | 2020-10-26 | 2022-06-17 | 中山大学 | Content identification method suitable for encrypted and non-encrypted data streams |
| CN113887812B (en) * | 2021-10-14 | 2023-07-07 | 广东电网有限责任公司 | Clustering-based small sample load prediction method, device, equipment and storage medium |
| CN114081504B (en) * | 2021-11-23 | 2024-03-01 | 青岛理工大学 | Driving intention recognition method and system based on electroencephalogram signals |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TW200929088A (en) * | 2007-12-24 | 2009-07-01 | Univ Nat Chiao Tung | Multi-person pose estimation using a zigbee sensor network |
| CN103886323A (en) * | 2013-09-24 | 2014-06-25 | 清华大学 | Behavior identification method based on mobile terminal and mobile terminal |
| US20140278208A1 (en) * | 2013-03-15 | 2014-09-18 | Aliphcom | Feature extraction and classification to determine one or more activities from sensed motion signals |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103699795B (en) * | 2013-12-20 | 2018-01-23 | 东软熙康健康科技有限公司 | A kind of motor behavior recognition methods, device and exercise intensity monitoring system |
| CN104598880A (en) * | 2015-03-06 | 2015-05-06 | 中山大学 | Behavior identification method based on fuzzy support vector machine |
-
2015
- 2015-07-08 CN CN201510400436.4A patent/CN106339071A/en active Pending
- 2015-09-07 WO PCT/CN2015/089058 patent/WO2017004880A1/en active Application Filing
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TW200929088A (en) * | 2007-12-24 | 2009-07-01 | Univ Nat Chiao Tung | Multi-person pose estimation using a zigbee sensor network |
| US20140278208A1 (en) * | 2013-03-15 | 2014-09-18 | Aliphcom | Feature extraction and classification to determine one or more activities from sensed motion signals |
| CN103886323A (en) * | 2013-09-24 | 2014-06-25 | 清华大学 | Behavior identification method based on mobile terminal and mobile terminal |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018232557A1 (en) * | 2017-06-19 | 2018-12-27 | 深圳市酷浪云计算有限公司 | Exercise movement recognition method and apparatus, and electronic device |
| CN107862276A (en) * | 2017-11-01 | 2018-03-30 | 北京旋极信息技术股份有限公司 | A kind of Activity recognition method and terminal |
| CN107862276B (en) * | 2017-11-01 | 2021-05-18 | 北京旋极信息技术股份有限公司 | Behavior identification method and terminal |
| CN109241926A (en) * | 2018-09-19 | 2019-01-18 | 重庆信络威科技有限公司 | A kind of driving behavior recognition methods based on wavelet analysis |
| CN110017834A (en) * | 2019-04-15 | 2019-07-16 | 歌尔科技有限公司 | Method, equipment and storage medium are determined using object |
| CN110017834B (en) * | 2019-04-15 | 2021-12-24 | 歌尔科技有限公司 | Usage object determination method, usage object determination apparatus, and storage medium |
| CN113030890A (en) * | 2019-12-24 | 2021-06-25 | 深圳市大富科技股份有限公司 | Target identification method and device based on vehicle-mounted radar |
| CN113030890B (en) * | 2019-12-24 | 2023-11-21 | 大富科技(安徽)股份有限公司 | Target identification method and device based on vehicle-mounted radar |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2017004880A1 (en) | 2017-01-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106339071A (en) | Method and device for identifying behaviors | |
| Yang et al. | Dynamic gesture recognition using surface EMG signals based on multi-stream residual network | |
| CN111027487B (en) | Behavior recognition system, method, medium and equipment based on multi-convolution kernel residual error network | |
| CN109948647B (en) | Electrocardiogram classification method and system based on depth residual error network | |
| CN104523266B (en) | A kind of electrocardiosignal automatic classification method | |
| CN105902257B (en) | Sleep state analysis method and device, intelligent wearable device | |
| CN107024685A (en) | A kind of gesture identification method based on apart from velocity characteristic | |
| CN106919958A (en) | A kind of human finger action identification method based on intelligent watch | |
| CN106503642B (en) | A kind of model of vibration method for building up applied to optical fiber sensing system | |
| CN108647614A (en) | The recognition methods of electrocardiogram beat classification and system | |
| CN105426842A (en) | Support vector machine based surface electromyogram signal multi-hand action identification method | |
| CN105117003A (en) | Intelligent wearing equipment and working method thereof | |
| CN101859377A (en) | Electromyographic signal classification method based on multi-kernel support vector machine | |
| CN106200984A (en) | Mental imagery brain-computer interface model modelling approach | |
| CN111723662B (en) | Human body posture recognition method based on convolutional neural network | |
| CN109993131A (en) | A kind of design idea judgement system and method based on multi-modal signal fused | |
| CN110866468A (en) | Gesture recognition system and method based on passive RFID | |
| CN106874959A (en) | A kind of multiple dimensioned scanning cascade forestry practises the training method of machine | |
| CN110443113A (en) | A kind of virtual reality Writing method, system and storage medium | |
| CN105046044A (en) | Non-stationary wind speed prediction method based on optimum wavelet-packet transformation | |
| CN103908361A (en) | Method for acquiring and operating artificial limb joint movement coupling drive signals | |
| CN104021372A (en) | Face recognition method and device thereof | |
| CN116400831B (en) | Handwriting pressure induction compensation algorithm of passive electromagnetic touch screen | |
| CN113642432A (en) | Method for identifying human body posture by convolutional neural network based on covariance matrix transformation | |
| Prasad et al. | A wireless dynamic gesture user interface for HCI using hand data glove |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C41 | Transfer of patent application or patent right or utility model | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20170303 Address after: 710121 Xi'an, Shaanxi, Changan District West Chang'an Avenue Applicant after: Xi'an University of Posts & Telecommunications Applicant after: ZTE Corporation Address before: 518057 Nanshan District Guangdong high tech Industrial Park, South Road, science and technology, ZTE building, Ministry of Justice Applicant before: ZTE Corporation |
|
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170118 |