CN104765453B - A kind of identity identifying method of the handheld device based on embedded three-dimensional accelerometer - Google Patents
A kind of identity identifying method of the handheld device based on embedded three-dimensional accelerometer Download PDFInfo
- Publication number
- CN104765453B CN104765453B CN201510153477.8A CN201510153477A CN104765453B CN 104765453 B CN104765453 B CN 104765453B CN 201510153477 A CN201510153477 A CN 201510153477A CN 104765453 B CN104765453 B CN 104765453B
- Authority
- CN
- China
- Prior art keywords
- authentication
- queues
- model
- matrixes
- features
- 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
Landscapes
- Collating Specific Patterns (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
The identity identifying method of the handheld device based on embedded three-dimensional accelerometer, including the hand-held equipment for being embedded with three-dimensional acceleration flowmeter sensor of user continuously does same action at least 40 times with gathered data;Singular Value Decomposition Using is carried out to gathered data;Feature enumeration is carried out to the data after decomposition, 20 features are enumerated;20 features are analyzed based on maximal correlation minimal redundancy method, authenticating user identification model is trained with one-class support vector machine using selected best features;Model first judges whether expired after setting up before each certification, and authentication is not carried out then;It is expired, it is updated and is authenticated again.Obviously the present invention is high compared with the certification accuracy rate of prior art, and False Rate is low, anti-to peep and impersonation attack, the safer privacy for protecting user;The feature calculation expense extracted is small, and calculating speed is fast;It is easy to operate, singlehanded can complete, mode, position and the gesture duration of user's handheld mobile phone are not relied on, and be not required to additional hardware equipment.
Description
Technical field
The present invention relates to a kind of identity identifying method of Behavior-based control feature, and in particular to one kind is based on embedded three-dimensional acceleration
The identity identifying method of the handheld device of meter, belongs to human-computer interaction technique field.
Background technology
At present, the embedded three-dimensional accelerometer of increasing handheld device, such as smart mobile phone, Intelligent bracelet, intelligent watch, just
Formula GPS is taken, wherein be representative with more and more universal smart mobile phone, the application on smart machine is also more and more, such as electronics postal
Part, E-Payment and various social networks etc. are applied.The substantial amounts of private sensitive information of these application memories, so intelligent hand
Machine increasingly privatization, the authentication of cellphone subscriber is also just more and more important.
There are many identity identifying methods on current smart mobile phone, these methods can be divided three classes:Knowledge based engineering certification
Technology, the authentication techniques of authentication techniques and Behavior-based control feature based on biological characteristic.
Knowledge based engineering authentication techniques depend on the information or secret that only user oneself knows, such as password, PIN or figure
Case, these information are easily attacked by person of peeping, and some passwords or pattern it is more complicated be not easy memory.
Authentication techniques based on biological characteristic depend on the exclusive biological characteristic of user, such as fingerprint, face feature, iris
Deng.Finger print identifying needs extra hardware device, costly;Face and iris authentication depend on light condition and user with
Angle between mobile phone, it has not been convenient to the use of user.
A kind of authentication techniques of Behavior-based control feature, there is provided mode of natural man-machine interaction, such as whipping or rock hand
Machine.These behavior acts are general all fairly simple, it is easy to which user operates and remembers.The authentication techniques of Behavior-based control feature are extracted hidden
Feature in plant behavior, these characteristic presents user is how to complete this behavior act.And existing Behavior-based control
The method of feature verification is mostly based on dynamic time programming (DTW) algorithm, but the performance of correlation technique depends on user's row
For similitude, the time requirement of mode, position and process performing for user's handheld mobile phone is strict.But user is hand-held
The time of the mode, position and process performing of mobile phone is change in actual life, so it is not conform to that it is strict with
Reason, only there is theoretic feasibility, but in actual applications and without reliability very high.
In a word, existing mobile phone identity authentication technology is to need extra hardware device a bit, costly;Some need spy
Different environmental condition and other conditions, it has not been convenient to the use of user;And the mobile phone identity authentication technology pair of Behavior-based control feature
The behavior requirement of user's operating handset is strict, such as the mode of user's handheld mobile phone, position and shake mobile phone duration.Hand
Mode, position and the shake mobile phone duration for holding mobile phone are change in actual life, so it is strict with being
It is irrational, it is in actual applications infeasible.And existing method can not be resisted to be peeped or impersonation attack, used to mobile phone
The secret protection at family brings very big threat.
The content of the invention
It is an object of the present invention to provide a kind of identity identifying method of the handheld device based on embedded three-dimensional accelerometer, the method
On the basis of the 20 behavioral statisticses features enumerated, with one-class support vector machine (one-class SVM) as core, based on not
The behavioral statisticses feature that the acceleration change of same action is done with people is different, realizes authenticating identity of mobile phone user (also referred to as mobile phone
Unblock), obtain authentication precision higher and the function of attack is imitated and peeped with resistance.
Technology design of the invention is the 3-axis acceleration sensor that is embedded in using intelligent handheld device records user
Whipping mobile phone unlocks the acceleration change process of gesture, original acceleration information is input into feature enumeration part, the part
SVD calculating is carried out to raw acceleration data first, feature enumeration is then carried out by physical analysis, finally supported using a class
Vector machine (one-class SVM) was carried out once according to the characteristic set training Model of Identity Authentication System selected every three days
Model modification, so as to carry out authentication using the model of training, obtains high-precision authentication result and is imitated with resistance
With the function of peeping attack.The method is selected to enumerate from raw acceleration data and has taken out 20 that can authenticate user statistics spies
Levy, with authentication precision higher, efficiently solve the problems, such as that conventional mobile phone unlocking method is easily peeped, imitated, these gestures
It is fairly simple, it is user-friendly to and remembers, and extra hardware device is not needed, it is less expensive.
A kind of identity identifying method of the handheld device based on embedded three-dimensional accelerometer, it is characterised in that including following step
Suddenly:
Step 1:The hand-held equipment for being embedded with three-dimensional acceleration flowmeter sensor of user, continuously does same action at least 40
Secondary, the three-dimensional acceleration flowmeter sensor embedded using handheld device gathers sensing data, and sample frequency is 80H;
Step 2:40 sensing datas to gathering carry out Singular Value Decomposition Using (SVD), A=U ∑s V*;
A is one of 40 sensing datas, to obtaining three sub-matrix U after matrix A singular value decomposition, ∑, and V*,
U is a matrix of n*3, is made up of three row, and U1, U2, U3 are designated as respectively;
∑ is a matrix of 3*3, and ∑ matrix has three characteristic values, and δ 1, δ 2, δ 3 are designated as respectively;
V* is a conjugate matrices for the unitary matrix of 3*3;
Step 3:Feature enumeration is carried out to the data after Singular Value Decomposition Using (SVD);
I.e. to U, ∑, tri- matrixes of V* extract five characteristic sets:
1. the ratio of ∑ matrix exgenvalue is calculated,<f1,f2,f3>
2. the frequency and energy of U matrixes are calculated,<f4,f5,f6,f7,f8,f9>
3. the zero-crossing rate of the row of U ∑s three is calculated respectively,<f10,f11,f12>
4. the average value and standard deviation of the row of U ∑s three are calculated respectively,<f13,f14,f15,f16,f17,f18>
5. average value and standard deviation that the row of U ∑s three are combined together are calculated,<f19,f20>
Five characteristic sets have 20 features, constitute a big characteristic set
<f1,f2,f3,f4,f5,f6,f7,f8,f9,f10,f11,f12,f13,f14,f15,f16,f17,f18,f19,
f20>;
Step 4:Based on maximal correlation minimal redundancy method (mRMR), 20 features are analyzed to more than, and then by most
Big correlation minimal redundancy method is therefrom selected to the feature for being used, i.e. best features, and all of best features composition is optimal
Characteristic set, then random subset is selected from best features set and is instructed using one-class support vector machine (one-class SVM)
Practice authenticating user identification model;
Step 5:After Model of Identity Authentication System foundation, before each authentication, the authentication mould is first determined whether
Type whether the term of validity more than 3 days, there is no expired, then carry out authentication;
It is expired, then existing model is updated, authentication is carried out after renewal again;Described renewal refers to continuous doing together
One kind action 40 times is simultaneously set up after Model of Identity Authentication System, and user has the new same authentication made to act, then act this
Data add queue, 40 times nearest action datas are stored in queue all the time, according to step 2-4 method using a class support to
Amount machine (one-class SVM) sets up new Model of Identity Authentication System, and carries out authentication;
Step 6:The model of the new training set up based on step 5, carries out authentication, i.e. handheld device user
Whether identity is that legitimate user is authenticated.
Features described above f1, f2, f3 calculating process is specific as follows:
(1) ∑ matrix is read, the value on the diagonal of matrix is designated as:δ 1, δ 2, δ 3, are three characteristic values of A matrixes
(2) relative ratios are calculated to δ 1, δ 2, δ 3, computing formula is as follows:
(3) f1 is obtained, tri- features of f2, f3 constitute a character subset:
f1:First ratio of characteristic value δ 1,
f2:Second ratio of characteristic value δ 2,
f3:3rd ratio of characteristic value δ 3.
Features described above f4, f5, f6, f7, f8, f9 calculating process are specific as follows:
(1) U matrixes are read in,
(2) each row U1, U2, the U3 to U matrixes carry out Fourier transformation (FFT) respectively,
(3) to the data of each row Fourier transformation, the ceiling capacity in addition to fundamental frequency and the first big energy extreme value is calculated
Frequency and energy corresponding to extreme value,
(4) frequency that will be obtained is multiplied by sample frequency 80H, obtains the corresponding cycle,
(5) six features of f4, f5, f6, f7, f8, f9 are obtained, a character subset is constituted, i.e.,
f4:After U1 Fourier transformations, the frequency for removing the ceiling capacity extreme value after fundamental frequency and the first big energy extreme value is multiplied by
Result after sample frequency 80H,
f5:After U1 Fourier transformations, the energy of the ceiling capacity extreme value after fundamental frequency and the first big energy extreme value is removed,
f6:After U2 Fourier transformations, the frequency for removing the ceiling capacity extreme value after fundamental frequency and the first big energy extreme value is multiplied by
Result after sample frequency 80H,
f7:After U2 Fourier transformations, the energy of the ceiling capacity extreme value after fundamental frequency and the first big energy extreme value is removed,
f8:After U3 Fourier transformations, the frequency for removing the ceiling capacity extreme value after fundamental frequency and the first big energy extreme value is multiplied by
Result after sample frequency 80H,
f9:After U3 Fourier transformations, the energy of the ceiling capacity extreme value after fundamental frequency and the first big energy extreme value is removed.
Features described above f10, f11, f12 calculating process are specific as follows:
(1) U ∑ matrixes are read in, this matrix is tieed up for n*3,
(2) to each column count zero-crossing rate of U ∑ matrixes, i.e., how many time passes through null value in sequence,
(3) tri- features of f10, f11, f12 are obtained, a character subset is constituted, i.e.,
f10:U ∑ matrixes, the zero-crossing rate of first row,
f11:U ∑ matrixes, the zero-crossing rate of secondary series,
f12:U ∑ matrixes, tertial zero-crossing rate.
Features described above f13, f14, f15, f16, f17, f18 calculating process are specific as follows:
(1) U ∑ matrixes are read in, this matrix is tieed up for n*3,
(2) each row to U ∑ matrixes calculate average value and standard deviation respectively,
(3) f13, f14, f15, f16, f17, f18 are obtained, six features constitute a character subset, i.e.,
f1:3:U ∑ matrixes, the average value of first row,
f14:U ∑ matrixes, the standard deviation of first row,
f15:U ∑ matrixes, the average value of secondary series,
f16:U ∑ matrixes, the standard deviation of secondary series,
f17:U ∑ matrixes, tertial average value,
f18:U ∑ matrixes, tertial standard deviation.
Features described above f19, f20 calculating process is specific as follows:
(1) U ∑ matrixes are read in, this matrix is tieed up for n*3,
(2) three row to U ∑ matrixes are calculated as below:
(3) its average and standard deviation are calculated C (i),
Obtain f19, f20, two characteristic values, composition characteristic a subset, i.e.,
f19:U ∑ matrixes, three column count quadratic sums, then evolution, finally calculates average value,
f20:U ∑ matrixes, three column count quadratic sums, then evolution, finally calculates standard deviation.
Above-mentioned steps (4) and (5) one-class support vector machine (one-class SVM) model training and model modification process tool
Body is as follows:
Two queues are used in the model training stage, one is TQ queues, for storing training sample data;Another
It is VQ queues, for authentication storage success and the sample data of authentification failure, in one-class support vector machine model training
(nu, γ) two searching processes of parameter, two length of queue of TQ and VQ all be 40;
(1) judge whether user uses this identity authentication function for the first time, if user uses for the first time, perform (2) (3)
(4), if user is not to use for the first time, perform (5),
(2) user uses for the first time, and TQ queues are sky, so need user to shake the acts of authentication 40 times of his setting, initially
Change TQ queues,
(3) using one-class support vector machine (one-class SVM), extracted respectively by 40 sample datas in TQ
20 features carry out model training, form Model of Identity Authentication System, in the process, one-class support vector machine (one-class
SVM several best features) are selected from 20 features, these best features constitute the subclass of above-mentioned 20 features, the son
Gather for building model,
(4) authentication is carried out using the model for training, TQ and VQ queues are safeguarded according to authentication result,
(5) training pattern has been present, and judges whether existing model is expired, if without expired, performing (6), if expired
, then (7) (8) are performed,
(6) authentication is carried out using existing model, TQ and VQ queues are safeguarded according to authentication result,
(7) model is out of date, carries out model modification,
(8) authentication is carried out using the model after renewal, TQ and VQ queues are safeguarded according to authentication result.
Above-mentioned (4) are as follows to the detailed process that TQ and VQ queues are safeguarded according to authentication result:
(1) judge identity authentication result, if certification success, performs (2), if authentification failure, perform (3),
(2) authentication success, TQ queues will remove date oldest sample data, and adopting now authentication
All notebook datas are stored in TQ queues, and that oldest sample data of the date removed in TQ queues finally is stored in into VQ queues,
(3) authentication failure, the sample data of now authentication are stored in VQ queues.
In above-mentioned (6), model is not out of date, as follows to the detailed process that TQ and VQ queues are safeguarded according to authentication result:
(1) judge identity authentication result, if certification success, performs (2), if authentification failure, perform (3),
(2) authentication success, TQ queues will remove date oldest sample data, and adopting now authentication
All notebook datas are stored in TQ queues, and that oldest sample data of the date removed in TQ queues finally is stored in into VQ queues,
(3) authentication failure, if VQ is full queue, deletes date oldest sample data in VQ queues, then will
The sample data of now authentication are stored in VQ queues;If VQ is not full queue, direct adopting now authentication
All notebook datas are stored in VQ queues.
In above-mentioned (7), model modification detailed process is as follows:
(1) existing model is alreadyd exceed three days, carries out model modification,
(2) 40 sample datas in present newest TQ queues are used to gather as training, with the sample in present VQ queues
Notebook data carries out parameter optimization as cross validation set, finally trains Model of Identity Authentication System, as the model after renewal,
(3) authentication is carried out using the model after renewal;
It is as follows to the detailed process that TQ and VQ queues are safeguarded according to authentication result after model modification in above-mentioned (8):
(1) judge identity authentication result, if certification success, performs (2), if authentification failure, perform (3),
(2) authentication success, TQ queues will remove date oldest sample data, and adopting now authentication
All notebook datas are stored in TQ queues, and that oldest sample data of the date removed in TQ queues finally is stored in into VQ queues,
(3) authentication failure, if VQ is full queue, deletes date oldest sample data in VQ queues, then will
The sample data of now authentication are stored in VQ queues;If VQ is not full queue, direct adopting now authentication
All notebook datas are stored in VQ queues.
The present invention on the basis of the above technical background further to lying in human behavior during feature divided
Analysis, has enumerated the physical features lain in during user behavior, and these features are due to different people because their age, property
Not, the difference such as dynamics, behavioural habits, feature can all differences when they do identical behavior gesture.So, we enumerate can
20 statistical natures of certification user are used for carrying out authentication.Finally use one-class support vector machine (one-class SVM)
Model of Identity Authentication System is trained according to the best features set selected, lightweight, high-precision hand are realized on smart mobile phone
Gesture identity identifying method, and a model modification was carried out every three days, so as to carry out authentication using the model of training, obtain
High-precision authentication result and imitated with resistance and peep the function of attack.Its remarkable advantage has:Certification accuracy rate is high, by mistake
Sentence rate low, anti-to peep and impersonation attack, the privacy for protecting user of better and safer;It is extracted and lies in differentiation in acceleration
20 physical statistics features of cellphone subscriber, these feature calculation expenses are small, and calculating speed is fast, easy to operate, it is only necessary to one
Hand can just be completed, and simply be easy to memory, not rely on mode, position and the gesture duration of user's handheld mobile phone, be had
Feasibility very high, it is not necessary to additional hardware equipment, low cost.
Brief description of the drawings
The identity identifying method flow chart of handheld device of Fig. 1 present invention based on embedded three-dimensional accelerometer.
Singular Value Decomposition Using (SVD) flow chart in Fig. 2 the inventive method.
Feature enumeration flow chart in Fig. 3 the inventive method.
One-class support vector machine (one-class SVM) model training flow chart in Fig. 4 the inventive method.
Specific implementation
As Figure 1-4, following examples are provided with reference to the content of the inventive method:
A kind of identity identifying method of the handheld device based on embedded three-dimensional accelerometer, it is characterised in that including following step
Suddenly:
Step 1:The hand-held equipment for being embedded with three-dimensional acceleration flowmeter sensor of user, continuously does same action at least 40
Secondary, the three-dimensional acceleration flowmeter sensor embedded using handheld device gathers sensing data, and sample frequency is 80H;
Step 2:40 sensing datas to gathering carry out Singular Value Decomposition Using (SVD), A=U ∑ V*, such as Fig. 2;
A is one of 40 sensing datas, to obtaining three sub-matrix U after matrix A singular value decomposition, ∑, and V*,
U is a matrix of n*3, is made up of three row, and U1, U2, U3 are designated as respectively;
∑ is a matrix of 3*3, and ∑ matrix has three characteristic values, and δ 1, δ 2, δ 3 are designated as respectively;
V* is a conjugate matrices for the unitary matrix of 3*3;
Step 3:Feature enumeration, such as Fig. 3 are carried out to the data after Singular Value Decomposition Using (SVD);
I.e. to U, ∑, tri- matrixes of V* extract five characteristic sets:
1. the ratio of ∑ matrix exgenvalue is calculated,<f1,f2,f3>
2. the frequency and energy of U matrixes are calculated,<f4,f5,f6,f7,f8,f9>
3. the zero-crossing rate of the row of U ∑s three is calculated respectively,<f10,f11,f12>
4. the average value and standard deviation of the row of U ∑s three are calculated respectively,<f13,f14,f15,f16,f17,f18>
5. average value and standard deviation that the row of U ∑s three are combined together are calculated,<f19,f20>
Five characteristic sets have 20 features, constitute a big characteristic set
<f1,f2,f3,f4,f5,f6,f7,f8,f9,f10,f11,f12,f13,f14,f15,f16,f17,f18,f19,
f20>;
Step 4:Such as Fig. 4, based on maximal correlation minimal redundancy method (mRMR), 20 features are analyzed to more than, enter
And the feature for being therefrom selected to be used by maximal correlation minimal redundancy method, i.e. best features, all of best features group
Into best features set, then any nonvoid subset is selected from best features set and one-class support vector machine (one- is utilized
Class SVM) training authenticating user identification model;
Step 5:Such as Fig. 4, after Model of Identity Authentication System foundation, before each authentication, the identity is first determined whether
Authentication model whether the term of validity more than 3 days, there is no expired, then carry out authentication;
It is expired, then existing model is updated, authentication is carried out after renewal again;Described renewal refers to continuous doing together
One kind action 40 times is simultaneously set up after Model of Identity Authentication System, and user has the new same authentication made to act, then act this
Data add queue, 40 times nearest action datas are stored in queue all the time, according to step 2-4 method using a class support to
Amount machine (one-class SVM) sets up new Model of Identity Authentication System, and carries out authentication;
Step 6:The model of the new training set up based on step 5, carries out authentication, i.e. handheld device user
Whether identity is that legitimate user is authenticated.
Claims (10)
1. a kind of identity identifying method of the handheld device based on embedded three-dimensional accelerometer, it is characterised in that comprise the following steps:
Step 1:The hand-held equipment for being embedded with three-dimensional acceleration flowmeter sensor of user, continuously does same action at least 40 times, profit
The three-dimensional acceleration flowmeter sensor embedded with handheld device gathers sensing data, and sample frequency is 80Hz;
Step 2:40 sensing datas to gathering carry out Singular Value Decomposition Using, A=U ∑s V*;
A is one of 40 sensing datas, to obtaining three sub-matrix U after matrix A singular value decomposition, ∑, and V*,
U is a matrix of n*3, is made up of three row, and U1, U2, U3 are designated as respectively;
∑ is a matrix of 3*3, and ∑ matrix has three characteristic values, and δ 1, δ 2, δ 3 are designated as respectively;
V* is a conjugate matrices for the unitary matrix of 3*3;
Step 3:Feature enumeration is carried out to the data after Singular Value Decomposition Using;
I.e. to U, ∑, tri- matrixes of V* extract five characteristic sets:
1. the ratio of ∑ matrix exgenvalue is calculated,<f1,f2,f3>
2. the frequency and energy of U matrixes are calculated,<f4,f5,f6,f7,f8,f9>
3. the zero-crossing rate of the row of U ∑s three is calculated respectively,<f10,f11,f12>
4. the average value and standard deviation of the row of U ∑s three are calculated respectively,<f13,f14,f15,f16,f17,f18>
5. average value and standard deviation that the row of U ∑s three are combined together are calculated,<f19,f20>
Five characteristic sets have 20 features, constitute a big characteristic set
<f1,f2,f3,f4,f5,f6,f7,f8,f9,f10,f11,f12,f13,f14,f15,f16,f17,f18,f19,f20>;
Step 4:Based on maximal correlation minimal redundancy method, 20 features are analyzed to more than, and then minimum by maximal correlation
Redundancy approach is therefrom selected to the feature for being used, i.e. best features, and all of best features constitute best features set, then
Random subset is selected from best features set and authenticating user identification model is trained using one-class support vector machine;
Step 5:After Model of Identity Authentication System foundation, before each authentication, first determine whether that the Model of Identity Authentication System is
The no term of validity more than 3 days, does not have expired, then carry out authentication;
It is expired, then existing model is updated, authentication is carried out after renewal again;Described renewal refers to continuously to do same
Action 40 times is simultaneously set up after Model of Identity Authentication System, and user has the new same authentication made to act, then by the action data
Queue is added, 40 times nearest action datas are stored in queue all the time, one-class support vector machine is utilized according to the method for step 2-4
New Model of Identity Authentication System is set up, and carries out authentication;
Step 6:The model of the new training set up based on step 5, carries out authentication, the i.e. identity of handheld device user
Whether it is that legitimate user is authenticated.
2. a kind of identity identifying method of the handheld device based on embedded three-dimensional accelerometer as claimed in claim 1, its feature
It is features described above f1, f2, f3 calculating process are specific as follows:
(1) ∑ matrix is read, the value on the diagonal of matrix is designated as:δ 1, δ 2, δ 3, are three characteristic values of A matrixes
(2) relative ratios are calculated to δ 1, δ 2, δ 3, computing formula is as follows:
(3) f1 is obtained, tri- features of f2, f3 constitute a character subset:
f1:First ratio of characteristic value δ 1,
f2:Second ratio of characteristic value δ 2,
f3:3rd ratio of characteristic value δ 3.
3. a kind of identity identifying method of the handheld device based on embedded three-dimensional accelerometer as claimed in claim 1, its feature
It is features described above f4, f5, f6, f7, f8, f9 calculating process specific as follows:
(1) U matrixes are read in,
(2) each row U1, U2, the U3 to U matrixes carry out Fourier transformation respectively,
(3) to the data of each row Fourier transformation, the ceiling capacity extreme value in addition to fundamental frequency and the first big energy extreme value is calculated
Corresponding frequency and energy,
(4) frequency that will be obtained is multiplied by sample frequency 80Hz, obtains the corresponding cycle,
(5) six features of f4, f5, f6, f7, f8, f9 are obtained, a character subset is constituted, i.e.,
f4:After U1 Fourier transformations, the frequency for removing the ceiling capacity extreme value after fundamental frequency and the first big energy extreme value is multiplied by sampling
Result after frequency 80H,
f5:After U1 Fourier transformations, the energy of the ceiling capacity extreme value after fundamental frequency and the first big energy extreme value is removed,
f6:After U2 Fourier transformations, the frequency for removing the ceiling capacity extreme value after fundamental frequency and the first big energy extreme value is multiplied by sampling
Result after frequency 80H,
f7:After U2 Fourier transformations, the energy of the ceiling capacity extreme value after fundamental frequency and the first big energy extreme value is removed,
f8:After U3 Fourier transformations, the frequency for removing the ceiling capacity extreme value after fundamental frequency and the first big energy extreme value is multiplied by sampling
Result after frequency 80H,
f9:After U3 Fourier transformations, the energy of the ceiling capacity extreme value after fundamental frequency and the first big energy extreme value is removed.
4. a kind of identity identifying method of the handheld device based on embedded three-dimensional accelerometer as claimed in claim 1, its feature
It is features described above f10, f11, f12 calculating process specific as follows:
(1) U ∑ matrixes are read in, this matrix is tieed up for n*3,
(2) to each column count zero-crossing rate of U ∑ matrixes, i.e., how many time passes through null value in sequence,
(3) tri- features of f10, f11, f12 are obtained, a character subset is constituted, i.e.,
f10:U ∑ matrixes, the zero-crossing rate of first row,
f11:U ∑ matrixes, the zero-crossing rate of secondary series,
f12:U ∑ matrixes, tertial zero-crossing rate.
5. a kind of identity identifying method of the handheld device based on embedded three-dimensional accelerometer as claimed in claim 1, its feature
It is features described above f13, f14, f15, f16, f17, f18 calculating process specific as follows:
(1) U ∑ matrixes are read in, this matrix is tieed up for n*3,
(2) each row to U ∑ matrixes calculate average value and standard deviation respectively,
(3) f13, f14, f15, f16, f17, f18 are obtained, six features constitute a character subset, i.e.,
f1:3:U ∑ matrixes, the average value of first row,
f14:U ∑ matrixes, the standard deviation of first row,
f15:U ∑ matrixes, the average value of secondary series,
f16:U ∑ matrixes, the standard deviation of secondary series,
f17:U ∑ matrixes, tertial average value,
f18:U Σ matrixes, tertial standard deviation.
6. a kind of identity identifying method of the handheld device based on embedded three-dimensional accelerometer as claimed in claim 1, its feature
It is that features described above f19, f20 calculating process is specific as follows:
(1) U ∑ matrixes are read in, this matrix is tieed up for n*3,
(2) three row to U ∑ matrixes are calculated as below:
(3) its average and standard deviation are calculated C (i),
Obtain f19, f20, two characteristic values, composition characteristic a subset, i.e.,
f19:U ∑ matrixes, three column count quadratic sums, then evolution, finally calculates average value,
f20:U ∑ matrixes, three column count quadratic sums, then evolution, finally calculates standard deviation.
7. a kind of identity identifying method of the handheld device based on embedded three-dimensional accelerometer as claimed in claim 1, its feature
It is that step 4 and 5 one-class support vector machine model trainings and model modification process are specific as follows:
Two queues are used in the model training stage, one is TQ queues, for storing training sample data;Another is VQ
Queue, for authentication storage success and the sample data of authentification failure, in one-class support vector machine model training
(nu, γ) two searching processes of parameter, two length of queue of TQ and VQ are all 40;
(1) judge whether user uses this identity authentication function for the first time, if user uses for the first time, perform (2) (3) (4), if
User is not to use for the first time, then perform (5),
(2) user uses for the first time, TQ queues for sky, so need user shake he setting authentication act 40 times, initially
Change TQ queues,
(3) one-class support vector machine is used, 20 features extracted respectively by 40 sample datas in TQ carry out model
Training, forms Model of Identity Authentication System, and in the process, one-class support vector machine selects several optimal spies from 20 features
Levy, for building model,
(4) authentication is carried out using the model for training, TQ and VQ queues are safeguarded according to authentication result,
(5) training pattern has been present, and judges whether existing model is expired, if without expired, performing (6), if expired,
Perform (7) (8),
(6) authentication is carried out using existing model, TQ and VQ queues are safeguarded according to authentication result,
(7) model is out of date, carries out model modification,
(8) authentication is carried out using the model after renewal, TQ and VQ queues are safeguarded according to authentication result.
8. a kind of identity identifying method of the handheld device based on embedded three-dimensional accelerometer as claimed in claim 7, its feature
It is above-mentioned (4) as follows to the detailed process that TQ and VQ queues are safeguarded according to authentication result:
(1) judge identity authentication result, if certification success, performs (2), if authentification failure, perform (3),
(2) authentication success, TQ queues will remove date oldest sample data, and by the sampling sample of now authentication
Notebook data is stored in TQ queues, and that oldest sample data of the date removed in TQ queues finally is stored in into VQ queues,
(3) authentication failure, the sample data of now authentication are stored in VQ queues.
9. a kind of identity identifying method of the handheld device based on embedded three-dimensional accelerometer as claimed in claim 7, its feature
It is that model is not out of date in above-mentioned (6), it is as follows to the detailed process that TQ and VQ queues are safeguarded according to authentication result:
(1) judge identity authentication result, if certification success, performs (2), if authentification failure, perform (3),
(2) authentication success, TQ queues will remove date oldest sample data, and by the sampling sample of now authentication
Notebook data is stored in TQ queues, and that oldest sample data of the date removed in TQ queues finally is stored in into VQ queues,
(3) authentication failure, if VQ is full queue, deletes date oldest sample data in VQ queues, then will now
The sample data of authentication are stored in VQ queues;If VQ is not full queue, directly by the sampling sample of now authentication
Notebook data is stored in VQ queues.
10. a kind of identity identifying method of the handheld device based on embedded three-dimensional accelerometer as claimed in claim 7, its feature
It is that model modification detailed process is as follows in above-mentioned (7):
(1) existing model is alreadyd exceed three days, carries out model modification,
(2) 40 sample datas in present newest TQ queues are used to gather as training, with the sample number in present VQ queues
Parameter optimization is carried out according to as cross validation set, Model of Identity Authentication System is finally trained, as the model after renewal,
(3) authentication is carried out using the model after renewal;
It is as follows to the detailed process that TQ and VQ queues are safeguarded according to authentication result after template renewal in above-mentioned (8):
(1) judge identity authentication result, if certification success, performs (2), if authentification failure, perform (3),
(2) authentication success, TQ queues will remove date oldest sample data, and by the sampling sample of now authentication
Notebook data is stored in TQ queues, and that oldest sample data of the date removed in TQ queues finally is stored in into VQ queues,
(3) authentication failure, if VQ is full queue, deletes date oldest sample data in VQ queues, then will now
The sample data of authentication are stored in VQ queues;If VQ is not full queue, directly by the sampling sample of now authentication
Notebook data is stored in VQ queues.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510153477.8A CN104765453B (en) | 2015-03-29 | 2015-03-29 | A kind of identity identifying method of the handheld device based on embedded three-dimensional accelerometer |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510153477.8A CN104765453B (en) | 2015-03-29 | 2015-03-29 | A kind of identity identifying method of the handheld device based on embedded three-dimensional accelerometer |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104765453A CN104765453A (en) | 2015-07-08 |
CN104765453B true CN104765453B (en) | 2017-06-27 |
Family
ID=53647339
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510153477.8A Active CN104765453B (en) | 2015-03-29 | 2015-03-29 | A kind of identity identifying method of the handheld device based on embedded three-dimensional accelerometer |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104765453B (en) |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105208561A (en) * | 2015-10-13 | 2015-12-30 | 哈尔滨工程大学 | Mobile phone user authentication method based on mobile phone sensor |
CN106686234B (en) * | 2016-12-28 | 2020-09-08 | 西北工业大学 | User age identification method based on mobile phone sensor data |
CN107392106B (en) * | 2017-06-26 | 2021-03-02 | 辽宁大学 | Human activity endpoint detection method based on double thresholds |
CN109740418B (en) * | 2018-11-21 | 2022-10-14 | 中山大学 | Yoga action identification method based on multiple acceleration sensors |
CN110287664A (en) * | 2019-07-01 | 2019-09-27 | 贵州大学 | A kind of identity identifying method being characterized selection based on multirow |
CN110324350B (en) * | 2019-07-09 | 2021-12-07 | 中国工商银行股份有限公司 | Identity authentication method and server based on mobile terminal non-sensitive sensor data |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103595538A (en) * | 2013-11-25 | 2014-02-19 | 中南大学 | Identity verification method based on mobile phone acceleration sensor |
CN104408341A (en) * | 2014-11-13 | 2015-03-11 | 西安交通大学 | Smart phone user identity authentication method based on gyroscope behavior characteristics |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
GB0201232D0 (en) * | 2002-01-19 | 2002-03-06 | Queen Mary & Westfield College | Authentication systems |
-
2015
- 2015-03-29 CN CN201510153477.8A patent/CN104765453B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103595538A (en) * | 2013-11-25 | 2014-02-19 | 中南大学 | Identity verification method based on mobile phone acceleration sensor |
CN104408341A (en) * | 2014-11-13 | 2015-03-11 | 西安交通大学 | Smart phone user identity authentication method based on gyroscope behavior characteristics |
Also Published As
Publication number | Publication date |
---|---|
CN104765453A (en) | 2015-07-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104765453B (en) | A kind of identity identifying method of the handheld device based on embedded three-dimensional accelerometer | |
Yang et al. | BehaveSense: Continuous authentication for security-sensitive mobile apps using behavioral biometrics | |
CN106022030B (en) | A kind of identity authorization system and method based on user's acquired behavior feature | |
CN108229130B (en) | Verification method and device | |
CN104408341B (en) | Smart phone user identity identifying method based on gyroscope behavioural characteristic | |
Serwadda et al. | When kids' toys breach mobile phone security | |
Lu et al. | Safeguard: User reauthentication on smartphones via behavioral biometrics | |
Khan et al. | A comparative evaluation of implicit authentication schemes | |
US20090328200A1 (en) | Hidden Markov Model ("HMM")-Based User Authentication Using Keystroke Dynamics | |
WO2007075982A1 (en) | Method and apparatus for multi-model hybrid comparison system | |
CN107153780B (en) | Writing behavior feature authentication method based on wrist wearable electronic equipment | |
Wang et al. | Improving reliability: User authentication on smartphones using keystroke biometrics | |
Anwar et al. | A Comparative Study of Graphical and Alphanumeric Passwords for Mobile Device Authentication. | |
CN110990811A (en) | Identity authentication method and device | |
CN104281795A (en) | Mouse action based password fault tolerance method | |
CN109684812A (en) | A kind of lasting identity identifying method of mobile device acquisition user movement behavior | |
CN106921500B (en) | Identity authentication method and device for mobile equipment | |
CN103297237A (en) | Identity registration method, identity authentication method, identity registration system, identity authentication system, personal authentication equipment and authentication server | |
CN107426397A (en) | Model training method and auth method based on user behavior feature | |
Shrestha et al. | Theft-resilient mobile wallets: Transparently authenticating NFC users with tapping gesture biometrics | |
Lin et al. | Crossbehaauth: Cross-scenario behavioral biometrics authentication using keystroke dynamics | |
Wangsuk et al. | Trajectory mining for keystroke dynamics authentication | |
CN110472389A (en) | Identity authorization system and its working method based on touch control gesture | |
CN109446780A (en) | A kind of identity identifying method, device and its storage medium | |
Saini et al. | Authenticating mobile phone user using keystroke dynamics |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
EXSB | Decision made by sipo to initiate substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |