CN107329563A - A kind of recognition methods of type of action, device and equipment - Google Patents
A kind of recognition methods of type of action, device and equipment Download PDFInfo
- Publication number
- CN107329563A CN107329563A CN201710364549.2A CN201710364549A CN107329563A CN 107329563 A CN107329563 A CN 107329563A CN 201710364549 A CN201710364549 A CN 201710364549A CN 107329563 A CN107329563 A CN 107329563A
- Authority
- CN
- China
- Prior art keywords
- characteristic
- action
- action data
- type
- information
- 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
-
- 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
- G06F3/011—Arrangements for interaction with the human body, e.g. for user immersion in virtual reality
-
- 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
- G06F18/24147—Distances to closest patterns, e.g. nearest neighbour classification
Abstract
The embodiments of the invention provide a kind of recognition methods of type of action, device and equipment, methods described includes：Gather the action data of user；Extract multiple characteristic elements in the action data；Using the multiple characteristic element, the characteristic information of the action data is generated；According to the characteristic information, recognize the corresponding type of action of the action data, solve the problem of type of action of user can not being recognized by common Intelligent worn device in the prior art, characteristic information of the embodiment of the present invention based on different type of action is different, the identification of type of action is carried out by the characteristic information of identification maneuver data, the efficiency and accuracy of identification is improved.
Description
Technical field
The present invention relates to areas of information technology, more particularly to a kind of recognition methods of type of action, a kind of type of action
Identifying device and a kind of identification equipment of type of action.
Background technology
Intelligent worn device is also referred to as wearable device, is that a kind of application wearable technology carries out intellectuality to daily wearing
The general name for the equipment that can be dressed design, developed, such as intelligent glasses, intelligent watch, Intelligent bracelet.
Generally, after user is wearing Intelligent worn device, Intelligent worn device can be entered to the walking or running of user
Row is simple to be counted, to the step number and distance of the current walking of user's output or running.But, existing Intelligent worn device is but
The action of user can not be identified, for example, dribble in basketball movement, shooting, passing, receiving, be detained in badminton
Kill, toe lift, the action such as batting, the basic None- identified of current Intelligent worn device.
The content of the invention
In view of the above problems, it is proposed that the embodiment of the present invention overcomes above mentioned problem or at least in part to provide one kind
Recognition methods, a kind of identifying device of type of action and a kind of corresponding action class of a kind of type of action solved the above problems
The identification equipment of type.
In order to solve the above problems, the embodiment of the invention discloses a kind of recognition methods of type of action, including：
Gather the action data of user；
Extract multiple characteristic elements in the action data；
Using the multiple characteristic element, the characteristic information of the action data is generated；
According to the characteristic information, the corresponding type of action of the action data is recognized.
Alternatively, the step of action data of the collection user includes：
Gather acceleration signal and angular velocity signal of the user in action process.
Alternatively, it is described extract in the action data multiple characteristic elements the step of include：
The maximum of the acceleration signal and angular velocity signal, minimum value, mean variance, slope, rising are extracted respectively
Time, and/or, fall time.
Alternatively, the multiple characteristic element has corresponding character numerical value respectively, described to use the multiple characteristic element
The step of element, characteristic information for generating the action data, includes：
The character numerical value of multiple characteristic elements is normalized, to obtain the normalization characteristic of multiple characteristic elements
Numerical value；
Dimension-reduction treatment is carried out to the normalization characteristic numerical value, to obtain target signature numerical value；
Using the target signature numerical value, the characteristic vector of the action data is generated.
Alternatively, described according to the characteristic information, the step of recognizing the action data corresponding type of action includes：
Multiple sample action data in preset motion characteristic storehouse and the similarity of the characteristic information are calculated, it is described many
Individual sample action data have corresponding label information respectively；
Multiple sample action data of the similarity more than predetermined threshold value are extracted as target sample action data；
Type of action indicated by the label information of the target sample action data is identified as the action data
Type of action.
Alternatively, described according to the characteristic information, the step of recognizing the action data corresponding type of action includes：
Calculate the similarity of the multiple characteristic sets and the characteristic information in preset motion characteristic storehouse, the multiple spy
Collection is closed has corresponding label information respectively；
The corresponding characteristic set of the similarity maximum is extracted for target signature set；
Type of action indicated by the label information of the target signature set is identified as to the action of the action data
Type.
Alternatively, the preset motion characteristic storehouse is generated in the following way：
Multiple sample action data are gathered, the multiple sample action data have corresponding label information respectively；
Extract multiple characteristic elements in the multiple sample action data；
Using the multiple characteristic element, the characteristic information of the multiple sample action data is generated；
According to the characteristic information and its corresponding label information, generation motion characteristic storehouse.
Alternatively, in addition to：
Multiple characteristic informations with same label information are combined as characteristic set respectively.
In order to solve the above problems, the embodiment of the invention discloses a kind of identifying device of type of action, including：
Acquisition module, the action data for gathering user；
Extraction module, for extracting multiple characteristic elements in the action data；
Generation module, for using the multiple characteristic element, generates the characteristic information of the action data；
Identification module, for according to the characteristic information, recognizing the corresponding type of action of the action data.
Alternatively, the acquisition module includes：
Submodule is gathered, for gathering acceleration signal and angular velocity signal of the user in action process.
Alternatively, the extraction module includes：
Extracting sub-module, maximum, minimum value, average for extracting the acceleration signal and angular velocity signal respectively
Variance, slope, rise time, and/or, fall time.
Alternatively, the multiple characteristic element has corresponding character numerical value respectively, and the generation module includes：
Submodule is normalized, is normalized for the character numerical value to multiple characteristic elements, to obtain multiple spies
Levy the normalization characteristic numerical value of element；
Dimensionality reduction submodule, for carrying out dimension-reduction treatment to the normalization characteristic numerical value, to obtain target signature numerical value；
Submodule is generated, for using the target signature numerical value, the characteristic vector of the action data is generated.
Alternatively, the identification module includes：
Multiple sample action data and the feature in first calculating sub module, the motion characteristic storehouse preset for calculating
The similarity of information, the multiple sample action data have corresponding label information respectively；
First extracting sub-module, for extracting multiple sample action data of the similarity more than predetermined threshold value as mesh
This action data of standard specimen；
First identification submodule, for the type of action indicated by the label information of the target sample action data to be known
Not Wei the action data type of action.
Alternatively, the identification module includes：
Multiple characteristic sets and the characteristic information in second calculating sub module, the motion characteristic storehouse preset for calculating
Similarity, the multiple characteristic set respectively have corresponding label information；
Second extracting sub-module, for extracting the corresponding characteristic set of the similarity maximum for target signature set；
Second identification submodule, for the type of action indicated by the label information of the target signature set to be identified as
The type of action of the action data.
Alternatively, the preset motion characteristic storehouse is by calling following module to generate：
Sample action data acquisition module, for gathering multiple sample action data, the multiple sample action data point
Ju You not corresponding label information；
Characteristic element extraction module, for extracting multiple characteristic elements in the multiple sample action data；
Characteristic information generation module, for using the multiple characteristic element, generates the multiple sample action data
Characteristic information；
Motion characteristic storehouse generation module, for according to the characteristic information and its corresponding label information, generation action to be special
Levy storehouse.
Alternatively, in addition to：
Characteristic information composite module, for multiple characteristic informations with same label information to be combined as into feature set respectively
Close.
In order to solve the above problems, the embodiment of the invention discloses a kind of identification equipment of type of action, including：Sensing
Device, processor, memory, wireless connection module, power supply module and display module；
The sensor is used for the action data for gathering user；
The processor is used to extract multiple characteristic elements in the action data；Using the multiple characteristic element,
Generate the characteristic information of the action data；According to the characteristic information, the corresponding type of action of the action data is recognized.
Compared with background technology, the embodiment of the present invention includes advantages below：
The embodiment of the present invention, by gathering the action data of user, and extracts multiple characteristic elements in the action data
Element, then using the multiple characteristic element, generates the characteristic information of the action data, and then can believe according to the feature
Breath, recognizes the corresponding type of action of the action data, solving in the prior art can not be by common Intelligent worn device
The problem of recognizing the type of action of user, characteristic information of the embodiment of the present invention based on different type of action is different, by knowing
The characteristic information of other action data carries out the identification of type of action, improves the efficiency and accuracy of identification.
Brief description of the drawings
Fig. 1 is a kind of step flow chart of the recognition methods embodiment one of type of action of the present invention；
Fig. 2 is a kind of step flow chart of the recognition methods embodiment two of type of action of the present invention；
Fig. 3 is a kind of schematic diagram of the recognition methods of type of action of the present invention；
Fig. 4 is a kind of structured flowchart of the identifying device embodiment of type of action of the present invention；
Fig. 5 is a kind of structured flowchart of the identification equipment embodiment of type of action of the present invention.
Embodiment
In order to facilitate the understanding of the purposes, features and advantages of the present invention, it is below in conjunction with the accompanying drawings and specific real
Applying mode, the present invention is further detailed explanation.
Reference picture 1, shows a kind of step flow chart of the recognition methods embodiment one of type of action of the present invention, specifically
It may include steps of：
Step 101, the action data of user is gathered；
In embodiments of the present invention, methods described can apply in the Intelligent worn devices such as Intelligent bracelet, intelligent watch,
The embodiment of the present invention is not construed as limiting to the particular type of Intelligent worn device.The action data can be user in motion process
In produced related data when making some action, for example, it may be user's dribble made in basketball movement, throwing
Basket, the action such as pass, receive, or the correlation that smash, toe lift, batting made in badminton etc. is produced when acting
The exercise data such as acceleration signal, angular velocity signal.
Produced in the specific implementation, user can be gathered by sensors such as gyroscope, accelerometers in action process
Acceleration signal and angular velocity signal, certainly, those skilled in the art can also using other modes gather user action
Data, the embodiment of the present invention is not construed as limiting to this.
Step 102, multiple characteristic elements in the action data are extracted；
Usually, the signals such as the acceleration and angular speed that collection is obtained can form corresponding waveform configuration respectively, from this
Multiple different characteristic elements can be extracted in waveform configuration.For example, can be extracted from the waveform configuration of acceleration signal
Go out the maximum of acceleration, minimum value, mean variance, slope, rise time, fall time；Correspondingly, can also be from angular speed
Maximum, minimum value, mean variance, slope, rise time, fall time of angular velocity etc. are extracted in the waveform configuration of signal
Deng.Certainly, those skilled in the art according to actual needs, can also extract other from acceleration signal and angular velocity signal
The characteristic element of type, for example, the characteristic element such as energy, zero crossing, the embodiment of the present invention is not construed as limiting to this.
Step 103, using the multiple characteristic element, the characteristic information of the action data is generated；
Generally, each characteristic element can be represented with corresponding character numerical value, but the amount of different types of characteristic element
Guiding principle is different so that the character numerical value tool of the characteristic element directly gathered is very different, for example, the feature of some characteristic elements
Numerical value may be very big, and the character numerical value of some characteristic elements then may very little, it is impossible to is directly compared.
Therefore, for unified metric, the influence that different numerical value differences are brought is eliminated, in the action data of user is extracted
Characteristic element after, the character numerical value of each characteristic element can be normalized, for example, a logarithm can be selected
The character numerical value of whole characteristic elements is normalized in numerical intervals [- 1,1] by function, obtains the normalization of each characteristic element
Character numerical value.Certainly, those skilled in the art can according to actual needs, the number after specifically chosen normalized function and normalization
It is worth interval scope, for example, the character numerical value of whole characteristic elements can be normalized in interval [0,1] or [0,10],
The embodiment of the present invention is not construed as limiting to this.
On the other hand, calculate for convenience, reduce amount of calculation, after normalization characteristic numerical value is obtained, can also continue to pair
The normalization characteristic numerical value obtained carries out dimension-reduction treatment, to reduce the dimension of characteristic element.For example, characteristic element can be dropped
It is low to be tieed up to 10, and according to the character numerical value of the characteristic element obtained after dimensionality reduction, generate a characteristic vector.Certainly, above-mentioned 10 Wei Te
It is only an example to levy element, and those skilled in the art can select the quantity of the dimension after dimensionality reduction according to actual needs, for example,
Can be 15 dimensions etc., the embodiment of the present invention is not construed as limiting to this.
Step 104, according to the characteristic information, the corresponding type of action of the action data is recognized.
In embodiments of the present invention, some sample action data can be gathered in advance, and for example collection experimenter is making respectively
Go out act of shooting, passing, action of receiving, dribbling movements, rob backboard action or three-step basket action when acceleration signal
And angular velocity signal, then extracted from the acceleration signal and angular velocity signal maximum, minimum value, mean variance, tiltedly
Rate, rise time, and/or, multiple characteristic elements such as fall time, and above-mentioned multiple characteristic elements are used, generation corresponds to upper
State act of shooting, passing, action of receiving, dribbling movements, rob backboard action or three-step basket action characteristic information, and according to
According to this feature information and it is marked with the label informations of corresponding actions, generation motion characteristic storehouse.Action is only a kind of example above,
Those skilled in the art can select other kinds of action as sample action according to actual needs, be moved so as to gather the sample
The action data of work is to generate motion characteristic storehouse, it is for instance possible to use the various methods of machine learning, such as KNN (k-Nearest
Neighbor, nearest neighbor algorithm), SVM (Support Vector Machine, SVMs), random forest, neutral net,
Deep learning method etc. is modeled, so as to generate motion characteristic storehouse, the embodiment of the present invention is not construed as limiting to this.
By taking KNN algorithms as an example, the core concept of KNN algorithms is if the k in feature space, a sample is most adjacent
Sample in it is most of belong to some classification, then the sample falls within this classification, and with sample in this classification
Characteristic.This method according to the classification of one or several closest samples on categorised decision it is determined that only determine sample to be divided
Affiliated classification.KNN algorithms are only relevant with minimal amount of adjacent sample in classification decision-making, mainly by limited neighbouring around
Sample, rather than by differentiating that the method for class field determines generic, intersection or overlapping more sample to be divided for class field
For this collection, KNN algorithms are more adapted to.
And then, the current action data of user is being collected, and generate after the characteristic information of action data, just can be by spy
Reference breath is compared with the characteristic information in preset motion characteristic storehouse, identifies the similarity highest with this feature information
Several motion characteristics, the corresponding action class of the action data is used as using the corresponding type of action of several above-mentioned motion characteristics
Type.
The embodiment of the present invention, by gathering the action data of user, and extracts multiple characteristic elements in the action data
Element, then using the multiple characteristic element, generates the characteristic information of the action data, and then can believe according to the feature
Breath, recognizes the corresponding type of action of the action data, solving in the prior art can not be by common Intelligent worn device
The problem of recognizing the type of action of user, characteristic information of the embodiment of the present invention based on different type of action is different, by knowing
The characteristic information of other action data carries out the identification of type of action, improves the efficiency and accuracy of identification.
Reference picture 2, shows a kind of step flow chart of the recognition methods embodiment two of type of action of the present invention, specifically
It may include steps of：
Step 201, generation motion characteristic storehouse；
In embodiments of the present invention, in order to realize the identification to the type of action of user, an action can be firstly generated special
Storehouse is levied, and the motion characteristic storehouse is preset in the Intelligent worn devices such as Intelligent bracelet, intelligent watch, so as to be set in intelligently wearing
After the standby action data for collecting user, the motion characteristic in preset motion characteristic storehouse can be used to carry out the action data
Identification.The embodiment of the present invention is not construed as limiting to the particular type of Intelligent worn device.
In embodiments of the present invention, in order to generate motion characteristic storehouse, multiple sample action data can be gathered first.The sample
This action data can the data such as associated acceleration and angular speed of the experimenter produced by when making a certain action.Example
Such as, the acceleration of the experimenter can be gathered by sensors such as gyroscope, accelerometers when experimenter makes act of shooting
Signal and angular velocity signal, so as to form the action data of act of shooting；Or, can when experimenter makes passing,
By gathering the acceleration signal and angular velocity signal of the experimenter, so as to form the action data of passing.Certainly, the above
Only a kind of example, those skilled in the art can need the action of identification according to follow-up, it is determined that specific sample action, so that
The action data of collecting sample action, the embodiment of the present invention is not construed as limiting to this.
In the specific implementation, can using six axle MEMS (Micro Electro Mechanical Systems, it is micro electronmechanical
System) sensor collecting sample action data, so as to export the acceleration signal of three axial directions and the angular speed letter of three axial directions
Number, sample frequency can be set as 100hz (hertz).
Generally, after multiple sample action data are collected, subsequent treatment, can be moved to each sample respectively for convenience
Make the label information of data setting one, for identifying the type of action corresponding to the sample action data.
It is then possible to extract multiple characteristic elements in the multiple sample action data according to certain time interval.
For example, when being sampled according to 100hz frequency, a characteristic element can be carried out using the data of each second as one section
Extraction, so as to obtain six waveform configurations, and extract from the waveform configuration maximum, minimum value, mean variance, tiltedly
The characteristic elements such as rate, rise time, fall time.Certainly, those skilled in the art according to actual needs, can also be from acceleration
Other kinds of characteristic element is extracted in signal and angular velocity signal, the embodiment of the present invention is not construed as limiting to this.
After multiple characteristic elements are extracted, the multiple characteristic element can be used, the multiple sample action is generated
The characteristic information of data.
Generally, each characteristic element can be represented with corresponding character numerical value, but the amount of different types of characteristic element
Guiding principle is different so that the character numerical value tool of the characteristic element directly gathered is very different, for example, the feature of some characteristic elements
Numerical value may be very big, and the character numerical value of some characteristic elements then may very little, it is impossible to is directly compared.
Therefore, for unified metric, the influence that different numerical value differences are brought is eliminated, in the action data of user is extracted
Characteristic element after, the character numerical value of each characteristic element can be normalized, for example, a logarithm can be selected
The character numerical value of whole characteristic elements is normalized in numerical intervals [- 1,1] by function, obtains the normalization of each characteristic element
Character numerical value.Certainly, those skilled in the art can according to actual needs, the number after specifically chosen normalized function and normalization
It is worth interval scope, for example, the character numerical value of whole characteristic elements can be normalized in interval [0,1] or [0,10],
The embodiment of the present invention is not construed as limiting to this.
On the other hand, calculate for convenience, reduce amount of calculation, after normalization characteristic numerical value is obtained, can also continue to pair
The normalization characteristic numerical value obtained carries out dimension-reduction treatment, to reduce the dimension of characteristic element.For example, characteristic element can be dropped
It is low to be tieed up to 10, and according to the character numerical value of the characteristic element obtained after dimensionality reduction, generate the characteristic information of sample action data, the spy
Reference breath can be a characteristic vector.Certainly, above-mentioned 10 dimensional feature element is only an example, and those skilled in the art can root
According to the quantity for the dimension being actually needed after selection dimensionality reduction, for example, it may be 15 dimensions etc., the embodiment of the present invention is not limited this
It is fixed.
So as to according to the characteristic information and its corresponding label information, generate motion characteristic storehouse, and this is acted
Feature database is preset in Intelligent worn device.In the motion characteristic storehouse, the corresponding spy of each sample action data can be included
Reference ceases, and marks the corresponding type of action of this feature information by label information.
As a kind of example of the present invention, in the characteristic information according to sample action data and its life of corresponding label information
Into after motion characteristic storehouse, multiple characteristic informations with same label information can also be combined as characteristic set respectively.For example,
Multiple characteristic informations labeled as " act of shooting " are combined as a characteristic set, and this feature set is used as using " act of shooting "
Label information.
Step 202, acceleration signal and angular velocity signal of the collection user in action process；
In embodiments of the present invention, user is wearing or dressed the intelligence in the preset motion characteristic storehouse just like in step 201
After the energy Intelligent worn device such as bracelet or intelligent watch, when user is making a certain action, the Intelligent worn device can lead to
Cross sensor and collect acceleration signal and angular velocity signal of the user in action process, generally, above-mentioned signal can be with phase
The waveform configuration answered is presented.
Step 203, the maximum of the acceleration signal and angular velocity signal, minimum value, mean variance, tiltedly are extracted respectively
Rate, rise time, and/or, fall time；
In the specific implementation, the processor of Intelligent worn device can extract acceleration from the waveform configuration of acceleration
The characteristic element such as maximum, minimum value, mean variance, slope, rise time, fall time, from the waveform of angular velocity signal
The characteristic elements such as maximum, minimum value, mean variance, slope, rise time, the fall time of angular velocity are extracted in structure.
Certainly, those skilled in the art according to actual needs, can also extract other classes from acceleration signal and angular velocity signal
The characteristic element of type, the embodiment of the present invention is not construed as limiting to this.
Step 204, the character numerical value of multiple characteristic elements is normalized, to obtain returning for multiple characteristic elements
One changes character numerical value；
In the specific implementation, for unified metric, eliminating the influence that different numerical value differences are brought, extracting such as step
The maximum of acceleration signal and angular velocity signal in 203, minimum value, mean variance, slope, rise time, and/or, under
After the characteristic elements such as drop time, the character numerical value of each characteristic element can be normalized respectively, so that will be all
The character numerical value of characteristic element is normalized in a certain specific numerical intervals, for example, it may be interval [- 1,1].
Step 205, dimension-reduction treatment is carried out to the normalization characteristic numerical value, to obtain target signature numerical value；
On the other hand, calculate for convenience, reduce amount of calculation, after normalization characteristic numerical value is obtained, can also continue to pair
The normalization characteristic numerical value obtained carries out dimension-reduction treatment, to reduce the dimension of characteristic element.For example, characteristic element can be dropped
It is low to be tieed up to 10, and target signature numerical value is used as using the corresponding character numerical value of above-mentioned 10 dimensional feature element.Certainly, above-mentioned 10 dimensional feature
Element is only an example, and those skilled in the art can select the quantity of the dimension after dimensionality reduction according to actual needs, for example, can
To be 15 dimensions etc., the embodiment of the present invention is not construed as limiting to this.
Step 206, using the target signature numerical value, the characteristic vector of the action data is generated；
It is then possible to according to the target signature numerical value obtained after dimensionality reduction, generate a characteristic vector, with this feature to amount instruction
The current action data of user.
Step 207, according to the characteristic information, the corresponding type of action of the action data is recognized.
In embodiments of the present invention, handle, and obtain corresponding in the action data of the user obtained to collection
After characteristic information, the corresponding type of action of characteristic information can be treated using preset motion characteristic storehouse according to this feature information
It is identified.
As a kind of example of the present invention, multiple sample action data in preset motion characteristic storehouse can be calculated first
With the similarity of the characteristic information, the multiple sample action data can have corresponding label information, Ran Houti respectively
The similarity is taken to exceed multiple sample action data of predetermined threshold value as target sample action data, so that by the target
Type of action indicated by the label information of sample action data is identified as the type of action of the action data.
In the specific implementation, can calculate respectively the features of each sample action data in preset motion characteristic storehouse to
Measure the cosine similarity between the characteristic vector current action data, using the cosine similarity as sample action data with
Similarity between the current action data of user, so that it is determined that it is most close to go out a number of action data current with user
Sample action data, and user is used as using the type of action indicated by the label information of above-mentioned most close sample action data
The corresponding type of action of current action data.
Certainly, in the similarity between calculating characteristic vector, except using cosine similarity, other classes can also be used
The similarity calculating method of type, for example, Euclidean distance (Euclidean Distance), manhatton distance (Manhattan
Distance), Chebyshev's distance (Chebyshev Distance), and, comentropy (Information Entropy) etc.
Method, the concrete mode used when the embodiment of the present invention is to calculating similarity is not construed as limiting.
As another example of the present invention, multiple characteristic sets in preset motion characteristic storehouse can also be calculated first
With the similarity of the characteristic information, the multiple characteristic set can have corresponding label information respectively, then extract institute
The corresponding characteristic set of similarity maximum is stated for target signature set, so that by the label information institute of the target signature set
The type of action of instruction is identified as the type of action of the action data.
In the specific implementation, can be handled and be generated corresponding in the current action data of the user to collecting
After characteristic information, it can be determined which characteristic set that this feature information belongs in preset motion characteristic storehouse, so that should
Type of action indicated by the label information of characteristic set is identified as the type of action of the current action data of user.
The embodiment of the present invention is by previously generating a motion characteristic storehouse, so that the current action data of user is being collected,
And generated according to the action data after corresponding characteristic information, can be by this feature information and preset motion characteristic storehouse
The characteristic information of sample action data such as is compared at the mode, identifies the type of action of the action data, solves existing skill
The problem of can not recognizing the type of action of user by common Intelligent worn device in art, improves the knowledge of type of action identification
Other efficiency and accuracy.
In order to make it easy to understand, below with a specific example, Jie is made in the recognition methods to the type of action of the present invention
Continue.
Reference picture 3, shows a kind of schematic diagram of the recognition methods of type of action of the present invention.In figure 3, methods described
It can include learning and two big processes of identification, the step of corresponding respectively to the generation motion characteristic storehouse in step 201, and step
The step of identification maneuver type in 202- steps 207.It is below above-mentioned study and recognizes illustrating for two big processes.
First, learning process
1st, data acquisition：Six axle MEMS sensor collecting sample action datas can be used, the acceleration of three axial directions is exported
Signal and the angle rate signal of three axial directions are spent, sample frequency is 100hz, and using the data of each second as one section, carry out one
The extraction of secondary characteristic element, obtains six waveform configurations.Sample action selection shooting, dribble and three kinds of actions of running, pass through 10
Name experimenter, each experimenter is repeated 10 times to each action.Thus there are three classes, common 3*10*10=300 training sample.
2nd, pre-process：The characteristic element of training sample on last stage is extracted, characteristic element includes each waveform configuration
Maximum, slope and rise time, thus obtain 6*3=18 characteristic element.Due to each characteristic element dimension not
Together, it is normalized, it is interval that the character numerical value of each characteristic element is normalized into [- 1,1].Then in order to reduce calculating
Amount, it may be considered that simplify dimension, 10 dimensions are dropped to by dimension, and generate corresponding characteristic vector, so that generating one includes 300
Individual sample, each sample is the motion characteristic storehouse of 10 dimensional feature vectors, and each characteristic vector correspond to the sample action of collection
When label information.
2nd, identification process
1st, data acquisition and pretreatment：The data acquisition and pretreatment of identification process and the data acquisition of learning process and pre-
Processing is similar, and the current action data of user is gathered by sensor, and extracts the characteristic element in the action data, and successively
It is normalized and dimension-reduction treatment, the characteristic vector of the current action data of generation user.
2nd, type of action is recognized：In specific identification, by the characteristic vector of obtained action data and motion characteristic storehouse
In sample be compared, pick out K most close sample, check this corresponding label information of K sample, you can learn this
The type of action of action.Or, can be with if the sample of same type is divided into same characteristic set in motion characteristic storehouse
After the characteristic vector of action data is obtained, judge which characteristic set this feature vector belongs to, so that with this feature set
Corresponding type of action is used as the current type of action of user.
It should be noted that for embodiment of the method, in order to be briefly described, therefore it to be all expressed as to a series of action group
Close, but those skilled in the art should know, the embodiment of the present invention is not limited by described sequence of movement, because according to
According to the embodiment of the present invention, some steps can be carried out sequentially or simultaneously using other.Secondly, those skilled in the art also should
Know, embodiment described in this description belongs to preferred embodiment, the involved action not necessarily present invention is implemented
Necessary to example.
Reference picture 4, shows a kind of structured flowchart of the identifying device embodiment of type of action of the present invention, specifically can be with
Including following module：
Acquisition module 401, the action data for gathering user；
Extraction module 402, for extracting multiple characteristic elements in the action data；
Generation module 403, for using the multiple characteristic element, generates the characteristic information of the action data；
Identification module 404, for according to the characteristic information, recognizing the corresponding type of action of the action data.
In embodiments of the present invention, the acquisition module 401 can specifically include following submodule：
Submodule is gathered, for gathering acceleration signal and angular velocity signal of the user in action process.
In embodiments of the present invention, the extraction module 402 can specifically include following submodule：
Extracting sub-module, maximum, minimum value, average for extracting the acceleration signal and angular velocity signal respectively
Variance, slope, rise time, and/or, fall time.
In embodiments of the present invention, the multiple characteristic element can have corresponding character numerical value, the generation respectively
Module 403 can specifically include following submodule：
Submodule is normalized, is normalized for the character numerical value to multiple characteristic elements, to obtain multiple spies
Levy the normalization characteristic numerical value of element；
Dimensionality reduction submodule, for carrying out dimension-reduction treatment to the normalization characteristic numerical value, to obtain target signature numerical value；
Submodule is generated, for using the target signature numerical value, the characteristic vector of the action data is generated.
In embodiments of the present invention, the identification module 404 can specifically include following submodule：
Multiple sample action data and the feature in first calculating sub module, the motion characteristic storehouse preset for calculating
The similarity of information, the multiple sample action data can have corresponding label information respectively；
First extracting sub-module, for extracting multiple sample action data of the similarity more than predetermined threshold value as mesh
This action data of standard specimen；
First identification submodule, for the type of action indicated by the label information of the target sample action data to be known
Not Wei the action data type of action.
In embodiments of the present invention, the identification module 404 can also include following submodule：
Multiple characteristic sets and the characteristic information in second calculating sub module, the motion characteristic storehouse preset for calculating
Similarity, the multiple characteristic set can have corresponding label information respectively；
Second extracting sub-module, for extracting the corresponding characteristic set of the similarity maximum for target signature set；
Second identification submodule, for the type of action indicated by the label information of the target signature set to be identified as
The type of action of the action data.
In embodiments of the present invention, the preset motion characteristic storehouse can be by calling following module to generate：
Sample action data acquisition module, for gathering multiple sample action data, the multiple sample action data can
To have corresponding label information respectively；
Characteristic element extraction module, for extracting multiple characteristic elements in the multiple sample action data；
Characteristic information generation module, for using the multiple characteristic element, generates the multiple sample action data
Characteristic information；
Motion characteristic storehouse generation module, for according to the characteristic information and its corresponding label information, generation action to be special
Levy storehouse.
In embodiments of the present invention, the preset motion characteristic storehouse can also be by calling following module to generate：
Characteristic information composite module, for multiple characteristic informations with same label information to be combined as into feature set respectively
Close.
For device embodiment, because it is substantially similar to embodiment of the method, so description is fairly simple, it is related
Part illustrates referring to the part of embodiment of the method.
Reference picture 5, shows a kind of structured flowchart of the identification equipment embodiment of type of action of the present invention, the equipment
Sensor 501, processor 502, memory 503, wireless connection module 504, power supply module 505 and display module can be included
506；
The sensor 501 can be used for the action data for gathering user；
The processor 502 can be used for extracting multiple characteristic elements in the action data；Using the multiple spy
Element is levied, the characteristic information of the action data is generated；According to the characteristic information, the corresponding action of the action data is recognized
Type.
Each embodiment in this specification is described by the way of progressive, what each embodiment was stressed be with
Between the difference of other embodiment, each embodiment identical similar part mutually referring to.
It should be understood by those skilled in the art that, the embodiment of the embodiment of the present invention can be provided as method, device or calculate
Machine program product.Therefore, the embodiment of the present invention can using complete hardware embodiment, complete software embodiment or combine software and
The form of the embodiment of hardware aspect.Moreover, the embodiment of the present invention can use it is one or more wherein include computer can
With in the computer-usable storage medium (including but is not limited to magnetic disk storage, CD-ROM, optical memory etc.) of program code
The form of the computer program product of implementation.
The embodiment of the present invention is with reference to method according to embodiments of the present invention, terminal device (system) and computer program
The flow chart and/or block diagram of product is described.It should be understood that can be by computer program instructions implementation process figure and/or block diagram
In each flow and/or square frame and the flow in flow chart and/or block diagram and/or the combination of square frame.These can be provided
Computer program instructions are set to all-purpose computer, special-purpose computer, Embedded Processor or other programmable data processing terminals
Standby processor is to produce a machine so that held by the processor of computer or other programmable data processing terminal equipments
Capable instruction is produced for realizing in one flow of flow chart or multiple flows and/or one square frame of block diagram or multiple square frames
The device for the function of specifying.
These computer program instructions, which may be alternatively stored in, can guide computer or other programmable data processing terminal equipments
In the computer-readable memory worked in a specific way so that the instruction being stored in the computer-readable memory produces bag
The manufacture of command device is included, the command device is realized in one flow of flow chart or multiple flows and/or one side of block diagram
The function of being specified in frame or multiple square frames.
These computer program instructions can be also loaded into computer or other programmable data processing terminal equipments so that
Series of operation steps is performed on computer or other programmable terminal equipments to produce computer implemented processing, so that
The instruction performed on computer or other programmable terminal equipments is provided for realizing in one flow of flow chart or multiple flows
And/or specified in one square frame of block diagram or multiple square frames function the step of.
Although having been described for the preferred embodiment of the embodiment of the present invention, those skilled in the art once know base
This creative concept, then can make other change and modification to these embodiments.So, appended claims are intended to be construed to
Including preferred embodiment and fall into having altered and changing for range of embodiment of the invention.
Finally, in addition it is also necessary to explanation, herein, such as first and second or the like relational terms be used merely to by
One entity or operation make a distinction with another entity or operation, and not necessarily require or imply these entities or operation
Between there is any this actual relation or order.Moreover, term " comprising ", "comprising" or its any other variant meaning
Covering including for nonexcludability, so that process, method, article or terminal device including a series of key elements are not only wrapped
Those key elements, but also other key elements including being not expressly set out are included, or also include being this process, method, article
Or the intrinsic key element of terminal device.In the absence of more restrictions, by wanting that sentence "including a ..." is limited
Element, it is not excluded that also there is other identical element in the process including the key element, method, article or terminal device.
Above to a kind of recognition methods of type of action provided by the present invention, the identifying device of a kind of type of action and one
The identification equipment of kind of type of action, is described in detail, used herein principle and implementation of the specific case to the present invention
Mode is set forth, and the explanation of above example is only intended to the method and its core concept for helping to understand the present invention；Meanwhile,
For those of ordinary skill in the art, according to the thought of the present invention, have change in specific embodiments and applications
Become part, in summary, this specification content should not be construed as limiting the invention.
Claims (17)
1. a kind of recognition methods of type of action, it is characterised in that including：
Gather the action data of user；
Extract multiple characteristic elements in the action data；
Using the multiple characteristic element, the characteristic information of the action data is generated；
According to the characteristic information, the corresponding type of action of the action data is recognized.
2. according to the method described in claim 1, it is characterised in that include the step of the action data of the collection user：
Gather acceleration signal and angular velocity signal of the user in action process.
3. method according to claim 2, it is characterised in that multiple characteristic elements in the extraction action data
The step of include：
The maximum of the acceleration signal and angular velocity signal, minimum value, mean variance, slope, rise time are extracted respectively,
And/or, fall time.
4. according to any described methods of claim 1-3, it is characterised in that the multiple characteristic element has corresponding respectively
The step of character numerical value, the multiple characteristic element of the use, characteristic information for generating the action data, includes：
The character numerical value of multiple characteristic elements is normalized, to obtain the normalization characteristic number of multiple characteristic elements
Value；
Dimension-reduction treatment is carried out to the normalization characteristic numerical value, to obtain target signature numerical value；
Using the target signature numerical value, the characteristic vector of the action data is generated.
5. method according to claim 4, it is characterised in that described according to the characteristic information, recognizes the action number
The step of according to corresponding type of action, includes：
Calculate multiple sample action data in preset motion characteristic storehouse and the similarity of the characteristic information, the multiple sample
This action data has corresponding label information respectively；
Multiple sample action data of the similarity more than predetermined threshold value are extracted as target sample action data；
Type of action indicated by the label information of the target sample action data is identified as to the action of the action data
Type.
6. method according to claim 4, it is characterised in that described according to the characteristic information, recognizes the action number
The step of according to corresponding type of action, includes：
Calculate the similarity of the multiple characteristic sets and the characteristic information in preset motion characteristic storehouse, the multiple feature set
Close has corresponding label information respectively；
The corresponding characteristic set of the similarity maximum is extracted for target signature set；
Type of action indicated by the label information of the target signature set is identified as to the type of action of the action data.
7. the method according to claim 5 or 6, it is characterised in that the preset motion characteristic storehouse is in the following way
Generation：
Multiple sample action data are gathered, the multiple sample action data have corresponding label information respectively；
Extract multiple characteristic elements in the multiple sample action data；
Using the multiple characteristic element, the characteristic information of the multiple sample action data is generated；
According to the characteristic information and its corresponding label information, generation motion characteristic storehouse.
8. method according to claim 7, it is characterised in that also include：
Multiple characteristic informations with same label information are combined as characteristic set respectively.
9. a kind of identifying device of type of action, it is characterised in that including：
Acquisition module, the action data for gathering user；
Extraction module, for extracting multiple characteristic elements in the action data；
Generation module, for using the multiple characteristic element, generates the characteristic information of the action data；
Identification module, for according to the characteristic information, recognizing the corresponding type of action of the action data.
10. device according to claim 9, it is characterised in that the acquisition module includes：
Submodule is gathered, for gathering acceleration signal and angular velocity signal of the user in action process.
11. device according to claim 10, it is characterised in that the extraction module includes：
Extracting sub-module, maximum, minimum value, average side for extracting the acceleration signal and angular velocity signal respectively
Difference, slope, rise time, and/or, fall time.
12. according to any described devices of claim 9-11, it is characterised in that the multiple characteristic element has corresponding respectively
Character numerical value, the generation module includes：
Submodule is normalized, is normalized for the character numerical value to multiple characteristic elements, to obtain multiple characteristic elements
The normalization characteristic numerical value of element；
Dimensionality reduction submodule, for carrying out dimension-reduction treatment to the normalization characteristic numerical value, to obtain target signature numerical value；
Submodule is generated, for using the target signature numerical value, the characteristic vector of the action data is generated.
13. device according to claim 12, it is characterised in that the identification module includes：
Multiple sample action data and the characteristic information in first calculating sub module, the motion characteristic storehouse preset for calculating
Similarity, the multiple sample action data respectively have corresponding label information；
First extracting sub-module, for extracting multiple sample action data of the similarity more than predetermined threshold value as target sample
This action data；
First identification submodule, for the type of action indicated by the label information of the target sample action data to be identified as
The type of action of the action data.
14. device according to claim 12, it is characterised in that the identification module includes：
The phase of multiple characteristic sets and the characteristic information in second calculating sub module, the motion characteristic storehouse preset for calculating
Like spending, the multiple characteristic set has corresponding label information respectively；
Second extracting sub-module, for extracting the corresponding characteristic set of the similarity maximum for target signature set；
Second identification submodule, it is described for the type of action indicated by the label information of the target signature set to be identified as
The type of action of action data.
15. the device according to claim 13 or 14, it is characterised in that the preset motion characteristic storehouse by call as
Lower module is generated：
Sample action data acquisition module, for gathering multiple sample action data, the multiple sample action data have respectively
There is corresponding label information；
Characteristic element extraction module, for extracting multiple characteristic elements in the multiple sample action data；
Characteristic information generation module, for using the multiple characteristic element, generates the feature of the multiple sample action data
Information；
Motion characteristic storehouse generation module, for according to the characteristic information and its corresponding label information, generation motion characteristic storehouse.
16. device according to claim 15, it is characterised in that also include：
Characteristic information composite module, for multiple characteristic informations with same label information to be combined as into characteristic set respectively.
17. a kind of identification equipment of type of action, it is characterised in that including：Sensor, processor, memory, wireless connection mould
Group, power supply module and display module；
The sensor is used for the action data for gathering user；
The processor is used to extract multiple characteristic elements in the action data；Using the multiple characteristic element, generation
The characteristic information of the action data；According to the characteristic information, the corresponding type of action of the action data is recognized.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710364549.2A CN107329563A (en) | 2017-05-22 | 2017-05-22 | A kind of recognition methods of type of action, device and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710364549.2A CN107329563A (en) | 2017-05-22 | 2017-05-22 | A kind of recognition methods of type of action, device and equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107329563A true CN107329563A (en) | 2017-11-07 |
Family
ID=60192790
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710364549.2A Pending CN107329563A (en) | 2017-05-22 | 2017-05-22 | A kind of recognition methods of type of action, device and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107329563A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108670119A (en) * | 2018-05-11 | 2018-10-19 | 莱克电气股份有限公司 | A kind of method and intellective dust collector of operation intellective dust collector |
| CN110163086A (en) * | 2019-04-09 | 2019-08-23 | 缤刻普达（北京）科技有限责任公司 | Body-building action identification method, device, equipment and medium neural network based |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104483519A (en) * | 2014-12-17 | 2015-04-01 | 中国气象科学研究院 | Trigger-based thunder and lightning signal triggering system and method |
| CN105184325A (en) * | 2015-09-23 | 2015-12-23 | 歌尔声学股份有限公司 | Human body action recognition method and mobile intelligent terminal |
| CN106203484A (en) * | 2016-06-29 | 2016-12-07 | 北京工业大学 | A kind of human motion state sorting technique based on classification layering |
| US20170065230A1 (en) * | 2015-06-15 | 2017-03-09 | Vital Labs, Inc. | Method and system for acquiring data for assessment of cardiovascular disease |
-
2017
- 2017-05-22 CN CN201710364549.2A patent/CN107329563A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104483519A (en) * | 2014-12-17 | 2015-04-01 | 中国气象科学研究院 | Trigger-based thunder and lightning signal triggering system and method |
| US20170065230A1 (en) * | 2015-06-15 | 2017-03-09 | Vital Labs, Inc. | Method and system for acquiring data for assessment of cardiovascular disease |
| CN105184325A (en) * | 2015-09-23 | 2015-12-23 | 歌尔声学股份有限公司 | Human body action recognition method and mobile intelligent terminal |
| CN106203484A (en) * | 2016-06-29 | 2016-12-07 | 北京工业大学 | A kind of human motion state sorting technique based on classification layering |
Non-Patent Citations (1)
| Title |
|---|
| 王见等: "基于改进SVM分类器的动作识别方法", 《重庆大学学报》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108670119A (en) * | 2018-05-11 | 2018-10-19 | 莱克电气股份有限公司 | A kind of method and intellective dust collector of operation intellective dust collector |
| CN110163086A (en) * | 2019-04-09 | 2019-08-23 | 缤刻普达（北京）科技有限责任公司 | Body-building action identification method, device, equipment and medium neural network based |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Quaid et al. | Wearable sensors based human behavioral pattern recognition using statistical features and reweighted genetic algorithm | |
| CN110765967B (en) | Action recognition method based on artificial intelligence and related device | |
| Zhu et al. | Efficient human activity recognition solving the confusing activities via deep ensemble learning | |
| CN105242779B (en) | A kind of method and mobile intelligent terminal of identification user action | |
| CN108703824B (en) | Bionic hand control system and control method based on myoelectricity bracelet | |
| KR102089002B1 (en) | Method and wearable device for providing feedback on action | |
| CN110472506A (en) | A kind of gesture identification method based on support vector machines and Neural Network Optimization | |
| Zinnen et al. | An analysis of sensor-oriented vs. model-based activity recognition | |
| CN110188654A (en) | A kind of video behavior recognition methods not cutting network based on movement | |
| CN110866468A (en) | Gesture recognition system and method based on passive RFID | |
| Fan et al. | Sensor fusion basketball shooting posture recognition system based on CNN | |
| CN109663323A (en) | A kind of motion state monitoring method and wearable device based on wearable device | |
| Ahmad et al. | Towards improved human action recognition using convolutional neural networks and multimodal fusion of depth and inertial sensor data | |
| Kumar et al. | A hybrid gesture recognition method for American sign language | |
| CN107329563A (en) | A kind of recognition methods of type of action, device and equipment | |
| CN110163264B (en) | Walking pattern recognition method based on machine learning | |
| CN113856132B (en) | Rope skipping action recognition system and method and rope skipping | |
| Zheng et al. | L-sign: Large-vocabulary sign gestures recognition system | |
| CN104850225B (en) | A kind of activity recognition method based on multi-level Fusion | |
| Abdullahi et al. | IDF-Sign: addressing inconsistent depth features for dynamic sign word recognition | |
| CN106569590A (en) | Object selection method and device | |
| Sarbishei | A platform and methodology enabling real-time motion pattern recognition on low-power smart devices | |
| Dennemont et al. | A discrete Hidden Markov models recognition module for temporal series: Application to real-time 3D hand gestures | |
| KR20210054349A (en) | Method for predicting clinical functional assessment scale using feature values derived by upper limb movement of patients | |
| CN116301328A (en) | Virtual reality equipment interaction identification system and control equipment based on arm knocking |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171107 |