CN104331431B - A kind of Mobile solution sort method of context aware - Google Patents
A kind of Mobile solution sort method of context aware Download PDFInfo
- Publication number
- CN104331431B CN104331431B CN201410567472.5A CN201410567472A CN104331431B CN 104331431 B CN104331431 B CN 104331431B CN 201410567472 A CN201410567472 A CN 201410567472A CN 104331431 B CN104331431 B CN 104331431B
- Authority
- CN
- China
- Prior art keywords
- axis
- place
- gps
- semantization
- value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9537—Spatial or temporal dependent retrieval, e.g. spatiotemporal queries
-
- 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
- G06F3/048—Interaction techniques based on graphical user interfaces [GUI]
- G06F3/0481—Interaction techniques based on graphical user interfaces [GUI] based on specific properties of the displayed interaction object or a metaphor-based environment, e.g. interaction with desktop elements like windows or icons, or assisted by a cursor's changing behaviour or appearance
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Human Computer Interaction (AREA)
- Navigation (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
The present invention relates to a kind of Mobile solution sort method of context aware, this method comprises the following steps：First, semantization place is obtained：2nd, acquisition activity：3rd, using sequence：The problem of present invention does not account for situation residing for user for tradition using sequence, and nowadays application sequence context perception ability is weaker, proposes a kind of Mobile solution sort method of new context aware, compared with the existing methods, the advantage is that：1) a variety of situations have been used in context aware；2) the stronger advanced situation of ability to express is established by rudimentary situation；3) context perception ability is enhanced when application is sorted so that more meet the individual demand of user using sequence.
Description
Technical field
The present invention relates to application to sort, and in particular to a kind of Mobile solution sort method of context aware.
Background technology
With the continuous development of mobile Internet, mobile terminal has become indispensable one of user's daily life
Point, increasing user's custom completes daily various affairs on mobile terminals, occurs therewith in every field numerous
The different application of function is to facilitate user to complete these affairs.Compared to traditional interconnected network mode, by colourful application,
User, which handles affairs, to be become more to follow one's bent, but increasing application can also cause to perplex to user, i.e., can not move
Required application is found in time in numerous applications in dynamic terminal.
With the rapid growth of number of applications, seem very necessary using recommendation or sequence.Traditional application recommends one
As only carried out by using number or frequency, record user use some application number, then according to frequency of use
Height is recommended to apply, and this mode obviously can not adapt to the different situations residing for user, because user is under different situations
Different applications can typically be used.With developing rapidly for sensing equipment technology, context aware can make full use of as one kind
Mobile terminal sensing data, the situation residing for intelligent analysis user, there is provided the technology of personalized service, increasingly obtain user
Favor.Most common context aware application is position navigation Service, using sensing equipments such as the GPS and Wi-Fi of mobile terminal,
Speculate the geographical position residing for user, there is provided related service.Application such as based on geographical position similitude is recommended, and passes through statistics
A certain geographical position different user is recommended user using the frequency of various applications, and this is pushed away for user to a certain extent
Recommend than convenient application.But this way of recommendation can only reflect in general rule, the personalization of user can not be obtained
Demand, while increasing for application species needs to use more type multiple-situation enhancing perception to carry out more intelligent recommendation.Now such as
The context perception ability that the modern application based on context aware technology is recommended is weaker, general only to use a kind of situation, and the situation
Ability to express it is weaker, can not solve the otherness between different user.
The content of the invention
The problem to be solved in the present invention is how using a variety of rudimentary situations to establish the stronger advanced situation of ability to express, from
And strengthening context perception ability, the more intelligent application on customer mobile terminal is ranked up.In order to solve the above problems, this
Invention proposes a kind of Mobile solution sort method of context aware.The present invention passes through the various low of acquisition for mobile terminal user first
Level contextual information (such as position, acceleration), feature extraction is then carried out on the basis of rudimentary contextual information, establishes advanced situation
Information (such as semantization place, activity) disaggregated model, finally using preset rules under different advanced contextual informations to movement
Application in terminal is ranked up.
In order to realize above-mentioned purpose, present invention employs following technical scheme：
A kind of Mobile solution sort method of context aware, this method comprise the following steps：
First, semantization place is obtained：
Step 1-1, the GPS position information of user is gathered by the GPS sensing equipments on mobile terminal；
Step 1-2, GPS position information is clustered into by GPS access locations according to time and distance；
Step 1-3, four kinds of week, time, duration and responsiveness temporal mode spies are extracted to all GPS access locations
Sign；
Step 1-4, to all GPS access locations its corresponding GPS characteristic vector of four kinds of temporal mode feature constructions；
A series of step 1-5, based on correct GPS characteristic vectors training semantization place classification for being labelled with semantization place
Model；
Step 1-6, it would be desirable to the GPS position information of identification by step 1-4 processing after obtain its corresponding GPS feature to
Amount,
But its corresponding semantization place is obtained according to the semantization place disaggregated model trained in step 1-5；
2nd, acquisition activity：
Step 2-1, the acceleration information of user is gathered by the acceleration sensing equipment on mobile terminal；
Step 2-2, acceleration information is cut by framing according to time window；
Step 2-3, statistical nature and frequency domain character are extracted to all obtained accelerometer frames；
Step 2-4, its corresponding acceleration signature vector of the feature construction obtained to the extraction of all accelerometer frames；
A series of step 2-5, based on correct acceleration signature vector training activity disaggregated models for being labelled with activity；
Step 2-6, it would be desirable to which the acceleration information of identification obtains its corresponding acceleration signature after step 2-4 processing
Vector,
It is but movable according to corresponding to the activity classification model trained in step 2-5 obtains it；
3rd, using sequence：
Step 3-1, ordering rule is set, including establish rank data structure, renewal sorting data and design sort algorithm；
Step 3-2, obtain the current semantization place of user and activity；
Step 3-3, the application on customer mobile terminal is ranked up based on ordering rule.
Comprise the following steps as a further improvement, obtaining semantization place：
Step 1, GPS position information is gathered：GPS position information is gathered from the GPS sensing equipments on mobile terminal；It is each
For individual GPS position information shaped like L=(lng, lat, t), wherein lng, lat is the GPS location longitude and latitude value, t be the position when
Between；GPS track data are shaped like LSeq=(L0,…,Ln), wherein LkFor k-th of GPS position information；
Step 2, GPS access locations are clustered：Each GPS position information L in GPS track data LSeq is compared its with
The distance of current cluster centre, if the distance is less than threshold value δ cluster_distance, the GPS position information is added and worked as
In preceding cluster, the duration currently clustered is otherwise calculated, i.e. the GPS position information time and first GPS in current cluster
The difference of positional information time, if the difference is more than threshold value δ time, will currently it cluster as a GPS access locations；
Step 3, extraction time pattern feature：To each GPS access locations extraction week, time, duration and sound
Should four kinds of temporal mode features of rate；Wherein represent that still day off occurs on weekdays for access behavior week；Time represents to access
Behavior occur interlude, its value by it is discrete be 24 values, represent 24 hours one day；Duration represents that access behavior occurs
Duration, and by it is discrete for it is longer, medium and it is shorter three value；Responsiveness represents gps signal pot life during the visit
Ratio, and represent to represent outdoor site, small indoor place and large-scale indoor ground respectively for high, medium and low three values by discrete
Point；
Step 4, GPS characteristic vectors are built：Above-mentioned four kinds of temporal mode features constitutive characteristic is vectorial, shaped like VL=(V0,
V1,V2,V3), wherein V0Week feature is represented, its value was 0 expression working day, and its value was 1 expression day off；Wherein V1Represent the time
Feature, its value are one of 0-23；Wherein V2Duration features are represented, its value is that 0 expression is longer, and its value is in 1 expression
It is 2 to represent shorter Deng, its value；Wherein V3Responsiveness feature is represented, its value is that 0 expression is high, and during its value represents for 1, its value is 2 tables
Show low；
Step 5, semantization place disaggregated model is trained：GPS characteristic vectors VL for being correctly labelled with semantization place
Semantization place disaggregated model is obtained using Bayesian network training grader, wherein semantization place include family, workplace,
Dining room, supermarket, shop, Condom, premise and the major class of scenic spot eight；
Step 6, semantization place is obtained：The GPS position information for needing to identify is obtained into GPS after step 1-4 processing
Characteristic vector VL, but its corresponding semantization place is obtained according to the semantization place disaggregated model trained in step 5.
As a further improvement, acquisition activity comprises the following steps：
Step 1, acceleration information is gathered：Acceleration information is gathered from the acceleration sensing equipment on mobile terminal；Often
One acceleration information is the acceleration information X-axis, Y-axis, the value of Z axis shaped like A=(x, y, z, t), wherein x, y, z, and t adds to be somebody's turn to do
The time that speed occurs；Acceleration time series data is shaped like ASeq=(A0,…,An), wherein AkFor k-th of acceleration information；
Step 2, accelerometer frame is cut into：By acceleration time series data according to sliding time window (such as total time be 6 seconds,
Step-length is 3 seconds) cut into accelerometer frame AF；If any acceleration time series data ASeq=(A0,…,A3n), it is set to 2n total time
The time of acceleration information, step-length are set to the time of n acceleration information, then the corresponding accelerometer frame cut out is followed successively by AF0
=(A0,…,A2n), AF1=(An,…,A3n)；
Step 3, extraction statistics and frequency domain character：Statistical nature and frequency domain character, statistics are extracted to each accelerometer frame AF
Feature includes average, variance, maximum, minimum value, energy and coefficient correlation, wherein average, variance, maximum, minimum value, energy
Amount needs to ask for acceleration X-axis, Y-axis, Z axis respectively, and coefficient correlation includes X-axis and three kinds of Y-axis, X-axis and Z axis, Y-axis and Z axis；
Frequency domain character is mainly Fourier Transform Coefficients；
Step 4, acceleration signature vector is built：Above-mentioned statistical nature is formed into statistical nature vector, shaped like VS=
(S0,…,S17), wherein S0, S1, S2X-axis average, Y-axis average, Z axis average, S are represented respectively3, S4, S5Represent respectively X-axis variance,
Y-axis variance, Z axis variance, S6, S7, S8X-axis maximum, Y-axis maximum, Z axis maximum, S are represented respectively9, S10, S11Represent respectively
X-axis minimum value, Y-axis minimum value, Z axis minimum value, S12, S13, S14X-axis energy, Y-axis energy, Z axis energy, S are represented respectively15,
S16, S17X-axis and Y-axis coefficient correlation, X-axis and Z axis coefficient correlation, Y-axis and Z axis coefficient correlation are represented respectively；Above-mentioned frequency domain is special
Sign forms frequency domain character vector, shaped like VF=(F0,…,Fn), wherein FkRepresent the Fourier Transform Coefficients of k-th of component of frequency domain；
Statistical nature vector VS and frequency domain character vector VF is formed into acceleration signature vector VA=(VS, VF)；
Step 5, training activity disaggregated model：Instructed for the acceleration signature vector VA for being correctly labelled with activity using SVM
Practice grader obtain activity classification model, wherein activity include meet, work, walk, upstairs, downstairs, diet, sleep, housework,
Do shopping, run, cycling, seeing 12 kinds of TV；
Step 6, acquisition activity：The acceleration information for needing to identify is obtained into acceleration signature after step 1-4 processing
Vectorial VA, however it is movable according to corresponding to the activity classification model trained in step 5 obtains it.
As a further improvement, comprise the following steps using sequence：
Step 1, ordering rule is set：
1) build semantization place access times vector P respectively to each Mobile solution that may be used and activity use it is secondary
Number vector A；Wherein semantization place access times vector is shaped like Pi=(p0,…,p7), wherein i represents i-th of application, pkRepresent
This application i number is used in k-th of semantization place；Movable access times vector is shaped like Ai=(a0,…,a11), wherein i
Represent i-th of application, akExpression uses this application i number when k-th movable；
2) user every time using it is a kind of using during i just by the vectorial P of the applicationiThe middle dimension for representing current semantics place
Value pkAdd 1, the same vectorial A by the applicationiThe middle dimension value a for representing current activekAdd 1；
3) when the application on customer mobile terminal needs rearrangement, i is applied to each, by the vectorial P of the applicationiIn
Represent the dimension values p in current semantics placekWith vectorial AiThe middle dimension values a for representing current activekAddition obtains weighted value Di,
Then according to DiValue size is ranked up to application from big to small；
Step 2, semantization place and the activity of user is obtained：From acquisition semantization place part and obtain in movable part
Obtain the current semantization place of user and activity；
Step 3, rule-based sequence：Just advised when the semantization place of user or activity change according to the sequence of setting
Sequence then is re-started to the application on customer mobile terminal.
The present invention does not account for situation residing for user for tradition using sequence, nowadays application sequence context perception ability
The problem of weaker, a kind of Mobile solution sort method of new context aware is proposed, compared with the existing methods, its advantage exists
In：
1) a variety of situations have been used in context aware；
2) the stronger advanced situation of ability to express is established by rudimentary situation；
3) context perception ability is enhanced when application is sorted so that more meet the personalized need of user using sequence
Ask.
Brief description of the drawings
Fig. 1 is overall flow figure of the present invention.
Fig. 2 obtains semantization place flow chart for the present invention.
Fig. 3 obtains activity flow chart for the present invention.
Fig. 4 is present invention application sequence flow chart.
Embodiment
The present invention proposes a kind of Mobile solution sort method of context aware, and overall flow figure is as shown in figure 1, this method
It is divided into and obtains three semantization place, acquisition activity and application sequence parts.Wherein obtain semantization place part and first gather GPS
Positional information, GPS access locations are then clustered into, it is special then to form GPS to GPS access locations extraction time pattern feature
Sign vector, finally train semantization place disaggregated model with GPS characteristic vectors；Obtain movable part and first gather acceleration information,
Then be cut to accelerometer frame, then accelerometer frame is extracted statistical nature and frequency domain character form acceleration signature to
Amount, finally with acceleration signature vector training activity disaggregated model；Ordering rule is first set using sort sections, is then obtained
The current semantization place of user and activity, it is finally based on ordering rule and user is moved with reference to different semantization places and activity
Application in dynamic terminal is ranked up.
The flow chart in semantization place is obtained as shown in Fig. 2 comprising the following steps that：
Step 1, GPS position information is gathered：GPS position information is gathered from the GPS sensing equipments on mobile terminal.It is each
For individual GPS position information shaped like L=(lng, lat, t), wherein lng, lat is the GPS location longitude and latitude value, t be the position when
Between.GPS track data are shaped like LSeq=(L0,…,Ln), wherein LkFor k-th of GPS position information.
Step 2, GPS access locations are clustered：Each GPS position information L in GPS track data LSeq is compared its with
The distance of current cluster centre, if the distance is less than threshold value δ cluster_distance, the GPS position information is added and worked as
In preceding cluster, the duration currently clustered is otherwise calculated, i.e. the GPS position information time and first GPS in current cluster
The difference of positional information time, if the difference is more than threshold value δ time, will currently it cluster as a GPS access locations.
Step 3, extraction time pattern feature：To each GPS access locations extraction week, time, duration and sound
Should four kinds of temporal mode features of rate.Wherein represent that still day off occurs on weekdays for access behavior week；Time represents to access
Behavior occur interlude, its value by it is discrete be 24 values, represent 24 hours one day；Duration represents that access behavior occurs
Duration, and by it is discrete for it is longer, medium and it is shorter three value；Responsiveness represents gps signal pot life during the visit
Ratio, and represent to represent outdoor site, small indoor place and large-scale indoor ground respectively for high, medium and low three values by discrete
Point.
Step 4, GPS characteristic vectors are built：Above-mentioned four kinds of temporal mode features constitutive characteristic is vectorial, shaped like VL=(V0,
V1,V2,V3), wherein V0Week feature is represented, its value was 0 expression working day, and its value was 1 expression day off；Wherein V1Represent the time
Feature, its value are one of 0-23；Wherein V2Duration features are represented, its value is that 0 expression is longer, and its value is in 1 expression
It is 2 to represent shorter Deng, its value；Wherein V3Responsiveness feature is represented, its value is that 0 expression is high, and during its value represents for 1, its value is 2 tables
Show low.
Step 5, semantization place disaggregated model is trained：GPS characteristic vectors VL for being correctly labelled with semantization place
Semantization place disaggregated model is obtained using Bayesian network training grader, wherein semantization place include family, workplace,
Dining room, supermarket, shop (representing small-sized shopping place, such as grocery store, clothes shop), Condom (represent to stop for interior
The place of spare time amusement, such as cinema, KTV), premise (represent Public place, such as hospital, bank) and scenic spot
(representing outdoor tourist quarters, such as park, seabeach) eight major classes.
Step 6, semantization place is obtained：The GPS position information for needing to identify is obtained into GPS after step 1-4 processing
Characteristic vector VL, but its corresponding semantization place is obtained according to the semantization place disaggregated model trained in step 5.
The flow chart of acquisition activity is as shown in figure 3, comprise the following steps that：
Step 1, acceleration information is gathered：Acceleration information is gathered from the acceleration sensing equipment on mobile terminal.Often
One acceleration information is the acceleration information X-axis, Y-axis, the value of Z axis shaped like A=(x, y, z, t), wherein x, y, z, and t adds to be somebody's turn to do
The time that speed occurs.Acceleration time series data is shaped like ASeq=(A0,…,An), wherein AkFor k-th of acceleration information.
Step 2, accelerometer frame is cut into：By acceleration time series data according to sliding time window (such as total time be 6 seconds,
Step-length is 3 seconds) cut into accelerometer frame AF.If any acceleration time series data ASeq=(A0,…,A3n), it is set to 2n total time
The time of acceleration information, step-length are set to the time of n acceleration information, then the corresponding accelerometer frame cut out is followed successively by AF0
=(A0,…,A2n), AF1=(An,…,A3n)。
Step 3, extraction statistics and frequency domain character：Statistical nature and frequency domain character, statistics are extracted to each accelerometer frame AF
Feature includes average, variance, maximum, minimum value, energy and coefficient correlation, wherein average, variance, maximum, minimum value, energy
Amount needs to ask for acceleration X-axis, Y-axis, Z axis respectively, and coefficient correlation includes X-axis and three kinds of Y-axis, X-axis and Z axis, Y-axis and Z axis；
Frequency domain character is mainly Fourier Transform Coefficients.
Step 4, acceleration signature vector is built：Above-mentioned statistical nature is formed into statistical nature vector, shaped like VS=
(S0,…,S17), wherein S0, S1, S2X-axis average, Y-axis average, Z axis average, S are represented respectively3, S4, S5Represent respectively X-axis variance,
Y-axis variance, Z axis variance, S6, S7, S8X-axis maximum, Y-axis maximum, Z axis maximum, S are represented respectively9, S10, S11Represent respectively
X-axis minimum value, Y-axis minimum value, Z axis minimum value, S12, S13, S14X-axis energy, Y-axis energy, Z axis energy, S are represented respectively15,
S16, S17X-axis and Y-axis coefficient correlation, X-axis and Z axis coefficient correlation, Y-axis and Z axis coefficient correlation are represented respectively.Above-mentioned frequency domain is special
Sign forms frequency domain character vector, shaped like VF=(F0,…,Fn), wherein FkRepresent the Fourier Transform Coefficients of k-th of component of frequency domain.
Statistical nature vector VS and frequency domain character vector VF is formed into acceleration signature vector VA=(VS, VF).
Step 5, training activity disaggregated model：Instructed for the acceleration signature vector VA for being correctly labelled with activity using SVM
Practice grader obtain activity classification model, wherein activity include meet, work, walk, upstairs, downstairs, diet, sleep, housework,
Do shopping, run, cycling, seeing 12 kinds of TV.
Step 6, acquisition activity：The acceleration information for needing to identify is obtained into acceleration signature after step 1-4 processing
Vectorial VA, however it is movable according to corresponding to the activity classification model trained in step 5 obtains it.
Using the flow chart of sequence as shown in figure 4, comprising the following steps that：
Step 1, ordering rule is set：
1) build semantization place access times vector P respectively to each Mobile solution that may be used and activity use it is secondary
Number vector A.Wherein semantization place access times vector is shaped like Pi=(p0,…,p7), wherein i represents i-th of application, pkRepresent
This application i number is used in k-th of semantization place；Movable access times vector is shaped like Ai=(a0,…,a11), wherein i
Represent i-th of application, akExpression uses this application i number when k-th movable.
2) user every time using it is a kind of using during i just by the vectorial P of the applicationiThe middle dimension for representing current semantics place
Value pkAdd 1, the same vectorial A by the applicationiThe middle dimension value a for representing current activekAdd 1.
3) when the application on customer mobile terminal needs rearrangement, i is applied to each, by the vectorial P of the applicationiIn
Represent the dimension values p in current semantics placekWith vectorial AiThe middle dimension values a for representing current activekAddition obtains weighted value Di,
Then according to DiValue size is ranked up to application from big to small.
Step 2, semantization place and the activity of user is obtained：From acquisition semantization place part and obtain in movable part
Obtain the current semantization place of user and activity.
Step 3, rule-based sequence：Just advised when the semantization place of user or activity change according to the sequence of setting
Sequence then is re-started to the application on customer mobile terminal.
Claims (4)
1. the Mobile solution sort method of a kind of context aware, it is characterised in that this method comprises the following steps：
First, semantization place is obtained：
Step 1-1, the GPS position information of user is gathered by the GPS sensing equipments on mobile terminal；
Step 1-2, GPS position information is clustered into by GPS access locations according to time and distance；
Step 1-3, four kinds of week, time, duration and responsiveness temporal mode features are extracted to all GPS access locations；
Step 1-4, to all GPS access locations its corresponding GPS characteristic vector of four kinds of temporal mode feature constructions；
A series of step 1-5, based on correct GPS characteristic vectors training semantization place classification moulds for being labelled with semantization place
Type；
Step 1-6, it would be desirable to which the GPS position information of identification obtains its corresponding GPS after step 1-1 to step 1-4 processing
Characteristic vector, but its corresponding semantization place is obtained according to the semantization place disaggregated model trained in step 1-5；
2nd, acquisition activity：
Step 2-1, the acceleration information of user is gathered by the acceleration sensing equipment on mobile terminal；
Step 2-2, acceleration information is cut by framing according to time window；
Step 2-3, statistical nature and frequency domain character are extracted to all obtained accelerometer frames；
Step 2-4, its corresponding acceleration signature vector of the feature construction obtained to the extraction of all accelerometer frames；
A series of step 2-5, based on correct acceleration signature vector training activity disaggregated models for being labelled with activity；
Step 2-6, it would be desirable to which the acceleration information of identification obtains accelerating corresponding to it after step 2-1 to step 2-4 processing
Characteristic vector is spent, however it is movable according to corresponding to the activity classification model trained in step 2-5 obtains it；
3rd, using sequence：
Step 3-1, ordering rule is set, including establish rank data structure, renewal sorting data and design sort algorithm；
Step 3-2, obtain the current semantization place of user and activity；
Step 3-3, the application on customer mobile terminal is ranked up based on ordering rule.
2. the Mobile solution sort method of a kind of context aware according to claim 1, it is characterised in that obtain semantization
Place comprises the following steps：
Step 1, GPS position information is gathered：GPS position information is gathered from the GPS sensing equipments on mobile terminal；Each
GPS position information shaped likeL=(lng, lat, t), whereinlng、latFor the GPS location longitude and latitude value,tFor the position when
Between；GPS track data shaped likeLSeq= (L 0 ,…,L n ), whereinL k ForkIndividual GPS position information；
Step 2, GPS access locations are clustered：To GPS track dataLSeqIn each GPS position informationLCompare its with it is current
The distance of cluster centre, if the distance is less than threshold valueδcluster_distance, then the GPS position information is added current poly-
In class, the duration currently clustered is otherwise calculated, i.e. the GPS position information time and first GPS location in current cluster
The difference of information time, if the difference is more than threshold valueδtime, then will currently cluster as a GPS access locations；
Step 3, extraction time pattern feature：To each GPS access locations extraction week, time, duration and responsiveness
Four kinds of temporal mode features；Wherein represent that still day off occurs on weekdays for access behavior week；Time represents access behavior
The interlude of generation, its value by it is discrete be 24 values, represent 24 hours one day；Duration represents holding for access behavior generation
The continuous time, and by discrete for longer, medium and shorter three values；Responsiveness represents the ratio of gps signal pot life during the visit
Rate, and represent outdoor site, small indoor place and large-scale indoor place by discrete for high, medium and low three values, respectively expression；
Step 4, GPS characteristic vectors are built：Above-mentioned four kinds of temporal mode features constitutive characteristic is vectorial, shaped likeVL= (V 0 ,V 1 ,V 2 ,V 3 ), whereinV 0 Week feature is represented, its value was 0 expression working day, and its value was 1 expression day off；WhereinV 1 Represent that the time is special
Sign, its value are one of 0-23；WhereinV 2 Duration features are represented, its value is that 0 expression is longer, and its value is that 1 expression is medium,
Its value is that 2 expressions are shorter；WhereinV 3 Responsiveness feature is represented, its value is that 0 expression is high, and during its value represents for 1, its value is 2 expressions
It is low；
Step 5, semantization place disaggregated model is trained：GPS characteristic vectors for being correctly labelled with semantization placeVLUsing
Bayesian network training grader obtains semantization place disaggregated model, and wherein semantization place includes family, workplace, meal
The Room, supermarket, shop, Condom, premise and the major class of scenic spot eight；
Step 6, semantization place is obtained：The GPS position information for needing to identify is obtained into GPS features after step 1-4 processing
VectorVL, but its corresponding semantization place is obtained according to the semantization place disaggregated model trained in step 5.
A kind of 3. Mobile solution sort method of context aware according to claim 1, it is characterised in that acquisition activity bag
Include following step：
Step 1, acceleration information is gathered：Acceleration information is gathered from the acceleration sensing equipment on mobile terminal；Each
Acceleration information shaped likeA=(x, y, z, t), whereinx, y, zFor the acceleration information X-axis, Y-axis, the value of Z axis,tFor this plus
The time that speed occurs；Acceleration time series data shaped likeASeq= (A 0 ,…,A n ), whereinA k ForkIndividual acceleration information；
Step 2, accelerometer frame is cut into：Acceleration time series data is cut into accelerometer frame according to sliding time windowAF；Such as
There is acceleration time series dataASeq= (A 0 ,…,A 3n ), be set to time of 2n acceleration information total time, step-length be set to n plus
The time of velocity information, the then corresponding accelerometer frame cut out are followed successively byAF 0= (A 0 ,…,A 2n ),AF 1= (A n ,…,A 3n )；
Step 3, extraction statistics and frequency domain character：To each accelerometer frameAFExtract statistical nature and frequency domain character, statistical nature
Including average, variance, maximum, minimum value, energy and coefficient correlation, wherein average, variance, maximum, minimum value, energy needs
Acceleration X-axis, Y-axis, Z axis are asked for respectively, coefficient correlation includes X-axis and three kinds of Y-axis, X-axis and Z axis, Y-axis and Z axis；Frequency domain
Feature is mainly Fourier Transform Coefficients；
Step 4, acceleration signature vector is built：Above-mentioned statistical nature is formed into statistical nature vector, shaped likeVS=(S 0 ,…,S 17 ), whereinS 0, S 1, S 2 X-axis average, Y-axis average, Z axis average are represented respectively,S 3, S 4, S 5 Represent respectively X-axis variance, Y-axis variance,
Z axis variance,S 6, S 7, S 8 X-axis maximum, Y-axis maximum, Z axis maximum are represented respectively,S 9, S 10, S 11 Represent that X-axis is minimum respectively
Value, Y-axis minimum value, Z axis minimum value,S 12, S 13, S 14 X-axis energy, Y-axis energy, Z axis energy are represented respectively,S 15, S 16, S 17 Respectively
Represent X-axis and Y-axis coefficient correlation, X-axis and Z axis coefficient correlation, Y-axis and Z axis coefficient correlation；Above-mentioned frequency domain character is formed into frequency domain
Characteristic vector, shaped likeVF=(F 0 ,…,F n ), whereinF k Represent frequency domain thekThe Fourier Transform Coefficients of individual component；By statistical nature
VectorVSWith frequency domain character vectorVFForm acceleration signature vectorVA=(VS, VF)；
Step 5, training activity disaggregated model：Acceleration signature vector for being correctly labelled with activityVAUsing SVM training point
Class device obtains activity classification model, wherein activity include meet, work, walk, upstairs, downstairs, diet, sleep, housework, purchase
Thing, running, cycle, see 12 kinds of TV；
Step 6, acquisition activity：The acceleration information for needing to identify is obtained into acceleration signature vector after step 1-4 processingVA, but it is movable according to corresponding to the activity classification model trained in step 5 obtains it.
4. the Mobile solution sort method of a kind of context aware according to claim 1, it is characterised in that using ranked package
Include following step：
Step 1, ordering rule is set：
1）Build semantization place access times vector respectively to each Mobile solution that may be usedPWith movable access times to
AmountA；Wherein semantization place access times vector shaped likeP i =(p 0 ,…,p 7 ), whereiniRepresent theiIndividual application,p k Represent thek
This application is used in individual semantization placeiNumber；Movable access times vector shaped likeA i =(a 0 ,…,a 11 ), whereiniRepresent theiIndividual application,a k Represent thekThis application is used when individual movableiNumber；
2）User is every time using a kind of applicationiWhen just by the vector of the applicationP i The middle dimension values for representing current semantics placep k
Add 1, the same vector by the applicationA i The middle dimension value for representing current activea k Add 1；
3）When application on customer mobile terminal needs rearrangement, to each applicationi, by the vector of the applicationP i Middle representative
The dimension values in current semantics placep k With vectorA i The middle dimension values for representing current activea k Addition obtains weighted valueD i , then
According toD i Value size is ranked up to application from big to small；
Step 2, semantization place and the activity of user is obtained：Obtained from obtaining semantization place part and obtaining in movable part
The current semantization place of user and activity；
Step 3, rule-based sequence：When the semantization place of user or activity change just according to the ordering rule pair of setting
Application on customer mobile terminal re-starts sequence.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410567472.5A CN104331431B (en) | 2014-10-22 | 2014-10-22 | A kind of Mobile solution sort method of context aware |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410567472.5A CN104331431B (en) | 2014-10-22 | 2014-10-22 | A kind of Mobile solution sort method of context aware |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104331431A CN104331431A (en) | 2015-02-04 |
| CN104331431B true CN104331431B (en) | 2017-11-28 |
Family
ID=52406158
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410567472.5A Active CN104331431B (en) | 2014-10-22 | 2014-10-22 | A kind of Mobile solution sort method of context aware |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104331431B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105652304B (en) * | 2016-01-06 | 2017-10-20 | 成都小步创想畅联科技有限公司 | Merge the place recognition methods of multisensor |
| CN106502398B (en) * | 2016-10-21 | 2019-01-29 | 浙江工业大学 | A kind of semantization activity recognition method based on Multi-view Integration study |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103106259A (en) * | 2013-01-25 | 2013-05-15 | 西北工业大学 | Mobile webpage content recommending method based on situation |
| CN103929712A (en) * | 2013-01-11 | 2014-07-16 | 三星电子株式会社 | Method And Mobile Device For Providing Recommended Items Based On Context Awareness |
| CN104008204A (en) * | 2014-06-17 | 2014-08-27 | 南京大学 | Dynamic multi-dimensional context awareness film recommending system and achieving method thereof |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9330165B2 (en) * | 2009-02-13 | 2016-05-03 | Microsoft Technology Licensing, Llc | Context-aware query suggestion by mining log data |
-
2014
- 2014-10-22 CN CN201410567472.5A patent/CN104331431B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103929712A (en) * | 2013-01-11 | 2014-07-16 | 三星电子株式会社 | Method And Mobile Device For Providing Recommended Items Based On Context Awareness |
| CN103106259A (en) * | 2013-01-25 | 2013-05-15 | 西北工业大学 | Mobile webpage content recommending method based on situation |
| CN104008204A (en) * | 2014-06-17 | 2014-08-27 | 南京大学 | Dynamic multi-dimensional context awareness film recommending system and achieving method thereof |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104331431A (en) | 2015-02-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP2820616B1 (en) | Empirical expert determination and question routing system and method | |
| CN103942049B (en) | Implementation method, client terminal device and the server of augmented reality | |
| US8438127B2 (en) | Behaviour pattern analysis system, mobile terminal, behaviour pattern analysis method, and program | |
| JP5072183B2 (en) | Meeting location determination method using spatial semantic modeling | |
| CN105091890B (en) | Method and apparatus for recommended candidate path | |
| US8577962B2 (en) | Server apparatus, client apparatus, content recommendation method, and program | |
| CN107291888B (en) | Machine learning statistical model-based living recommendation system method near living hotel | |
| CN107742107A (en) | Facial image sorting technique, device and server | |
| CN103970861B (en) | Information demonstrating method and equipment | |
| CN106407519B (en) | A kind of modeling method of crowd's movement law | |
| CN106709235A (en) | Exercise training data processing method and device | |
| CN104915000A (en) | Multisensory biological recognition interaction method for naked eye 3D advertisement | |
| CN103500266A (en) | Method and device for pushing application information based on health recognition | |
| KR101445923B1 (en) | Method for providing interactive exhibit service | |
| CN104699245B (en) | Learning content and interactive learning scene pushing method, learning machine and system | |
| CN108876509B (en) | Method and system for analyzing user tag by using POI | |
| WO2022001122A1 (en) | Data processing method and apparatus, and device and storage medium | |
| CN107358471A (en) | A kind of tourist resources based on visit behavior recommends method and system | |
| CN104331431B (en) | A kind of Mobile solution sort method of context aware | |
| CN106570033A (en) | Individual recommendation method and device based on scene data | |
| EP3014493A1 (en) | Determining demographic data | |
| CN108629671A (en) | A kind of restaurant recommendation method of fusion user behavior information | |
| CN104636457B (en) | A kind of method and device of location finding cognition | |
| WO2014209571A1 (en) | Displaying demographic data | |
| CN104361023B (en) | A kind of mobile terminal Tobacco Reference method for pushing of context aware |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |