CN110298159B - Smart phone dynamic gesture identity authentication method - Google Patents
Smart phone dynamic gesture identity authentication method Download PDFInfo
- Publication number
- CN110298159B CN110298159B CN201910571581.7A CN201910571581A CN110298159B CN 110298159 B CN110298159 B CN 110298159B CN 201910571581 A CN201910571581 A CN 201910571581A CN 110298159 B CN110298159 B CN 110298159B
- Authority
- CN
- China
- Prior art keywords
- data
- gesture
- user
- template
- authentication
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/30—Authentication, i.e. establishing the identity or authorisation of security principals
- G06F21/31—User authentication
- G06F21/32—User authentication using biometric data, e.g. fingerprints, iris scans or voiceprints
-
- 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]
- G06F3/0487—Interaction techniques based on graphical user interfaces [GUI] using specific features provided by the input device, e.g. functions controlled by the rotation of a mouse with dual sensing arrangements, or of the nature of the input device, e.g. tap gestures based on pressure sensed by a digitiser
- G06F3/0488—Interaction techniques based on graphical user interfaces [GUI] using specific features provided by the input device, e.g. functions controlled by the rotation of a mouse with dual sensing arrangements, or of the nature of the input device, e.g. tap gestures based on pressure sensed by a digitiser using a touch-screen or digitiser, e.g. input of commands through traced gestures
- G06F3/04883—Interaction techniques based on graphical user interfaces [GUI] using specific features provided by the input device, e.g. functions controlled by the rotation of a mouse with dual sensing arrangements, or of the nature of the input device, e.g. tap gestures based on pressure sensed by a digitiser using a touch-screen or digitiser, e.g. input of commands through traced gestures for inputting data by handwriting, e.g. gesture or text
Abstract
The invention relates to a dynamic gesture identity authentication method for a smart phone, which improves the dynamic gesture authentication efficiency of the smart phone by global regular path constraint, and expands and contracts gesture data length to obtain a standardized gesture template by a normalization method, thereby eliminating the problem of inaccurate selection of the gesture template and ensuring the authentication precision of gesture identity authentication.
Description
Technical Field
The invention relates to the field of mobile equipment identity recognition, in particular to a dynamic gesture identity authentication method for a smart phone.
Background
With the popularization of intelligent mobile devices, intelligent mobile terminals face more and more security threats, and it is particularly necessary to design a secure and efficient mobile device identity authentication scheme. The traditional authentication method based on the cryptology mechanism has the problems of password embezzlement, password forgetting and the like, and the problems can be well solved through the biological characteristic identification technology.
In the prior art, researchers have proposed a biometric identification technology for identifying the identity of a person by collecting a biometric sample of the person. The biological characteristics can be divided into: physiological characteristics and behavior characteristics, wherein the physiological characteristics are determined by innate factors such as fingerprints, irises, human faces, DNA and the like, and the behavior characteristics are caused by acquired factors such as signatures, gesture postures, screen sliding unlocking postures and the like. Because the biological characteristics can distinguish different people identities and are difficult to imitate, the privacy and the safety of the individual can be effectively protected. The acceleration sensor is widely arranged in the smart phone with low price, high sensitivity and small volume, and the acceleration sensor arranged in the smart phone is used for acquiring an acceleration signal generated by a user authentication gesture and has operability as biological feature identification.
The popular methods of identity authentication based on dynamic gesture features include: a DTW-based method, HMM-based method, support vector machine-based method, Varga J, etc. propose a new authentication method for smart phones and similar devices, which authenticates the user based on the gestures made by the user to the device itself, in combination with password information. Liuwei et al have improved the Global Sequence Alignment (GSA) algorithm, carry on the interpolation operation to the sequence after matching, have improved the aerial signature identity authentication effect of Chinese. Liu Q et al propose a gesture authentication method capable of resisting user posture changes, which includes collecting gesture behavior data through a built-in sensor of a mobile phone, obtaining each gesture of a user by adopting a K-means algorithm, and training an authentication model for each gesture. And the Miao sensitivity and the like provide a difference-to-bottom-up linear segmentation method for automatically detecting the effective gesture action endpoint, and finally, a DTW algorithm is used for performing matching calculation on the test gesture and the template gesture. The method comprises the steps that data of a user when the user answers a call are collected by utilizing a mobile phone acceleration sensor, a gyroscope sensor, a distance sensor and the like in Liule music and the like, a probability distribution function is constructed by utilizing a recognition result of a DTW algorithm, and Dempster/Shafer evidence theory is adopted for fusion. And Pongyongchun and the like judge the authenticity of the user identity by extracting a first-order normalized derivative sequence and a second-order normalized derivative sequence of the gesture motion and a motion direction as identity verification feature sequences and comparing the registered template feature sequences with the test feature sequences by adopting a DTW algorithm. Meanwhile, a plurality of documents realize identity authentication of dynamic gesture features based on a DTW algorithm, but the DTW algorithm has higher computational complexity in computing similarity between time sequences. Liuxian plum and the like construct a distance matrix based on the window distance, and the motion similarity calculation is carried out by adopting a DTW (dynamic time warping) optimization algorithm based on global and local constraints, so that the efficiency of the algorithm is improved. Niennattrakul V et al limit the accumulated distance matrix in a small parallelogram and perform the calculation of the accumulated distance matrix in a table look-up manner by recording the value of the limited lower boundary, thereby reducing the calculation time and improving the algorithm efficiency. However, the algorithms can obtain the computational efficiency, and meanwhile, the error of matching precision identification is reduced and correspondingly improved.
Disclosure of Invention
In view of this, the present invention provides a method for authenticating a dynamic gesture of a smart phone, which not only improves the time efficiency of authentication, but also ensures the accuracy of user authentication.
The invention is realized by adopting the following scheme: a dynamic gesture identity authentication method for a smart phone specifically comprises a gesture template training stage and an identity verification stage;
the gesture template training stage comprises the following steps:
step S11: collecting user information for training, carrying out data standardization and smooth denoising treatment on a plurality of pieces of collected user information, and selecting a candidate template;
step S12: normalizing the candidate templates by adopting linear up-down sampling to generate a standard gesture template of each user; authenticating the training set through a standard gesture template, and obtaining an optimal identity authentication threshold value under the constraint of a weighting accuracy index;
the authentication phase comprises the following steps:
step S21: collecting test gesture data of a user, and carrying out data standardization and smooth denoising on the test gesture data of the user;
step S22: and (4) calculating a CM-DTW regular distance by using the data processed in the step (S21) and the user standard gesture template, wherein if the regular distance is smaller than the optimal authentication threshold value of the template, the test gesture data and the user standard gesture template data belong to the same user, otherwise, the test gesture data and the user standard gesture template data do not belong to the same user.
Further, in step S11 or step S21, when data collection is performed, a data value in the sliding unlocking process of the user is obtained by using an acceleration sensor and a pressure sensor built in the smart phone, and collection is started when the user slides the unlocked finger to touch the screen until the collection is finished when the finger leaves the screen; the frequency of the collected data is set to be 50Hz, the data sequence from the beginning to the end of the collection is called gesture data, and the definition of the gesture data is as follows:
R＝{r1,r2,...ri,...,rn}；
in the formula, ri＝{ri1,ri2,ri3,ri4Where 1,2, n is one4 dimensional data where ri1、ri2、ri3X, Y, Z triaxial parameter value r of built-in acceleration sensor of mobile phonei4Is the value of the built-in pressure sensor of the mobile phone.
Further, in step S11 and step S21, the data normalization specifically includes:
the gesture data is standardized by adopting a z-score standardization method, and the result after the standardization of the gesture data is shown as the following formula:
in the formula, R is a data sequence of the gesture data set R, μ is a mean value of R, σ is a standard deviation of R, and m represents the number of gesture data in the set R.
Further, step S11 is specifically: repeatedly collecting gesture data n times by a user, and collecting n groups of gesture data R1,R2,...,RnAcquiring m user data as interference gesture data in the same way; and carrying out standardization and smooth denoising processing on n groups of data collected by a user, and dividing training data and test data.
Further, step S12 specifically includes the following steps:
step S121: calculating regular distances between each group of training gesture data of the user and other training gesture data by using a DTW algorithm under the restriction of a Sakoe-Chiba window;
step S122: calculating the Mean value of regular distances between each group of gesture data and other gesture data of the same user as a standard threshold value of the gesture, wherein the Mean value of j groups of data of the ith user is Mean (i, j);
step S123: taking Mean (i, j) rate (i, j) as the optimal threshold value of the jth data of the ith user, and performing identity authentication on all other data; wherein rate (i, j) is the optimal threshold proportion of the jth data of the ith user;
step S124: based on the calculation result of step S121, under the optimal threshold classification of step S123, obtaining the identity authentication result of each group of gesture data, and evaluating it with a weighted accuracy index WA, calculating an identity authentication index value of each group of data, where WA (i, j) is a WA index value of a jth group of data of an ith user; calculating an index mean value AVG _ WA (i) of all data of the same user; when WA (i, j) ≧ AVG _ WA (i), the jth data is considered as the candidate template of the ith user; constructing a candidate template set by the gesture data meeting the requirement of the ith user candidate template, wherein the formula of the WA index is shown as the following formula:
in the formula, a is the legal gesture authentication as the legal gesture number, b is the illegal gesture authentication as the illegal gesture number, c is the illegal gesture authentication as the illegal gesture number, d is the illegal gesture authentication as the legal gesture number, and β is the corresponding weighted value, and the formula is expressed as follows:
step S125: averaging the lengths of all gesture data of the candidate template set to obtain an average value L;
step S126: in order to conveniently standardize the template, all gesture data in the candidate template set are subjected to stretching processing to be processed into a normalized template with the length L;
step S127: averaging the candidate templates processed in step S126, and using the average template as the standard template of the user
Wherein
A standard template of the ith user; will be provided with
Calculating regular distance with training gesture data at optimal threshold
To obtain an optimum WA (R)i) Index value, optimal threshold value
As a template
The optimal authentication threshold.
Compared with the prior art, the invention has the following beneficial effects: the invention provides a dynamic gesture identity authentication method for a smart phone, which improves the dynamic gesture authentication efficiency of the smart phone by global regular path constraint, and solves the problem of inaccurate selection of a gesture template by stretching and contracting the gesture data length to obtain a standardized gesture template by a normalization method, thereby ensuring the authentication precision of gesture identity authentication.
Drawings
FIG. 1 is a diagram of a Sakoe-Chiba confinement window of an embodiment of the present invention.
FIG. 2 is a Sakoe-Chiba constraint path reduction diagram according to an embodiment of the present invention.
FIG. 3 is a flow chart of a method according to an embodiment of the present invention.
Fig. 4 shows WA values for authentication of the standard template and the optimal template according to the embodiment of the present invention.
Fig. 5 is a schematic diagram illustrating the time consumption comparison between the CM-DTW algorithm and the classical DTW algorithm according to the embodiment of the present invention.
Fig. 6 is a WA value diagram of the CM-DTW algorithm and the classical DTW algorithm according to an embodiment of the present invention.
Detailed Description
The invention is further explained below with reference to the drawings and the embodiments.
It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the disclosure. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
As shown in fig. 3, the embodiment provides a method for authenticating an identity through a dynamic gesture of a smart phone, which specifically includes a gesture template training stage and an identity verification stage;
the gesture template training stage comprises the following steps:
step S11: collecting user information for training, carrying out data standardization and smooth denoising treatment on a plurality of pieces of collected user information, and selecting a candidate template;
step S12: normalizing the candidate templates by adopting linear up-down sampling to generate a standard gesture template of each user; authenticating the training set through a standard gesture template, and obtaining an optimal identity authentication threshold value under the constraint of a weighting accuracy index;
the authentication phase comprises the following steps:
step S21: collecting test gesture data of a user, and carrying out data standardization and smooth denoising on the test gesture data of the user;
step S22: and (4) calculating a CM-DTW regular distance by using the data processed in the step (S21) and the user standard gesture template, wherein if the regular distance is smaller than the optimal authentication threshold value of the template, the test gesture data and the user standard gesture template data belong to the same user, otherwise, the test gesture data and the user standard gesture template data do not belong to the same user.
In this embodiment, in step S11 or step S21, when data collection is performed, a data value in the sliding unlocking process of the user is obtained by using an acceleration sensor and a pressure sensor built in the smartphone, and collection is started when the user slides the unlocked finger to touch the screen until the collection is finished when the finger leaves the screen; the frequency of the collected data is set to be 50Hz, the data sequence from the beginning to the end of the collection is called gesture data, and the definition of the gesture data is as follows:
R＝{r1,r2,...ri,...,rn}；
in the formula, ri＝{ri1,ri2,ri3,ri4N is a 4-dimensional data, where r is 1,2i1、ri2、ri3X, Y, Z triaxial parameter value r of built-in acceleration sensor of mobile phonei4Is the value of the built-in pressure sensor of the mobile phone.
Preferably, in order to avoid the dimension inclination with a larger distance vector value caused by the inconsistency of the magnitude values of the dimensions, the collected gesture data needs to be standardized so as to reduce the influence that the distance cannot accurately reflect the relationship between the feature data due to the difference of the magnitude values of the dimensions. In this embodiment, in step S11 and step S21, the data normalization specifically includes:
the gesture data is standardized by adopting a z-score standardization method, and the result after the standardization of the gesture data is shown as the following formula:
in the formula, R is a data sequence of the gesture data set R, μ is a mean value of R, σ is a standard deviation of R, and m represents the number of gesture data in the set R.
Preferably, the collected data may generate noisy data because the user may shake with different amplitudes while sliding a finger across the screen when unlocking. By smoothing the gesture data, errors caused by environmental factors of the collected data can be effectively eliminated. Based on the idea of the weighted moving average method, the normalized result is further calculated by smoothing processing according to the following formula:
wherein k is more than or equal to N/2, N is the window size of the translation movement calculation by the weighted average method and is generally set as an odd number,
representing the smoothed result of the kth data point in a gesture sequence, wiIs the weight of the ith data.
In this embodiment, step S11 specifically includes: repeatedly collecting gesture data n times by a user, and collecting n groups of gesture data R1,R2,...,RnAcquiring m user data as interference gesture data in the same way; and carrying out standardization and smooth denoising processing on n groups of data collected by a user, and dividing training data and test data.
In this embodiment, step S12 specifically includes the following steps:
step S121: calculating regular distances between each group of training gesture data of the user and other training gesture data (other gesture data and interference gesture data under the same user) by using a DTW algorithm under the restriction of a Sakoe-Chiba window;
step S122: calculating the Mean value of regular distances between each group of gesture data and other gesture data of the same user as a standard threshold value of the gesture, wherein the Mean value of j groups of data of the ith user is Mean (i, j);
step S123: taking Mean (i, j) rate (i, j) as the optimal threshold value of the jth data of the ith user, and performing identity authentication on all other data; wherein rate (i, j) is the optimal threshold proportion of the jth data of the ith user;
step S124: based on the calculation result of step S121, under the optimal threshold classification of step S123, obtaining the identity authentication result of each group of gesture data, and evaluating it with a weighted accuracy index WA, calculating an identity authentication index value of each group of data, where WA (i, j) is a WA index value of a jth group of data of an ith user; calculating an index mean value AVG _ WA (i) of all data of the same user; when WA (i, j) ≧ AVG _ WA (i), the jth data is considered as the candidate template of the ith user; constructing a candidate template set by the gesture data meeting the requirement of the ith user candidate template, wherein the formula of the WA index is shown as the following formula:
in the formula, a is the legal gesture authentication as the legal gesture number, b is the illegal gesture authentication as the illegal gesture number, c is the illegal gesture authentication as the illegal gesture number, d is the illegal gesture authentication as the legal gesture number, and β is the corresponding weighted value, and the formula is expressed as follows:
step S125: averaging the lengths of all gesture data of the candidate template set to obtain an average value L;
step S126: in order to conveniently standardize the template, all gesture data in the candidate template set are subjected to stretching processing to be processed into a normalized template with the length L;
step S127: averaging the candidate templates processed in step S126, and using the average template as the standard template of the user
Wherein
A standard template of the ith user; will be provided with
Calculating regular distance with training gesture data at optimal threshold
To obtain an optimum WA (R)i) Index value, optimal threshold value
As a template
The optimal authentication threshold.
In particular, in the present embodiment, in step S121, the DTW algorithm solves the problem of template matching with different pronunciation lengths in speech recognition, and is widely used in isolated speech recognition. Since gesture data during slide unlocking is similar to voice signal data and has a time sequence relationship and time-space variability, many researchers apply the DTW algorithm to gesture authentication. The DTW algorithm is a non-linear warping technique that combines time warping and distance measures. And calculating the minimum distance between two time sequences with different lengths, wherein the smaller the distance is, the more similar the two time sequences are. When the DTW algorithm is used for identity authentication, the distance between the sequence and the user reference template sequence needs to be tested, and the legal identity and the illegal identity are determined according to the distance. The following two equations are shown for the test sequence and reference template sequence descriptions:
G(M)＝{g1,g2,...,gM}；
T(N)＝{t1,t2,...,tN}；
in the formula, G (M) and T (N) respectively represent a test sequence of M data points and a reference template sequence of N data points after data processing; calculating the distance between two sequence data points by using Euclidean distance, and testing ith data g of the sequenceiAnd j-th data t of the reference template sequencejThe distance between them is calculated as follows:
computing Euclidean distances from data points of the test sequence and the reference template sequence pairwise to create an M x N dimensional distance matrix D, elements (g)i,tj) Value d (g) ofi,tj) Represents the data point giAnd tjThe euclidean distance of; the DTW algorithm calculates the distance between the two sequences, essentially finding a suitable warping function j ═ f (i), and making this function satisfy the following equation:
in the formula, P (G, T) is the optimal matching distance between the test sequence G (M) and the reference template sequence T (N), and the principle of the dynamic regularization algorithm is that the matrix D is searching for a point from the starting point (G)1,t1) To the end point (g)M,tN) The cumulative distance of the paths is minimal; in order to achieve cumulative distance optimization, the warping function of the DTW algorithm needs to satisfy the constraints of global and local constraints. The local constraint conditions that the regularization function needs to satisfy have the following three points:
(1) and (3) end point constraint, wherein the end point constraint requires that the starting points and the end points of the two sequences are consistent:
(2) monotone constraint, the generation of gesture data has a sequence, and the regularization function must ensure that the matching path does not violate the time sequence of the generation of the gesture data, so the following formula must be satisfied:
f(n+1)≥f(n)；
(3) and (3) continuity constraint, in order to ensure that the loss of the matching information is minimum, the warping function cannot skip any matching point, the optimization problem is solved through a DTW algorithm, and the cumulative distance of the optimal path can be obtained as follows:
in the formula, p (g)i,tj) Represents the minimum cumulative distance of the path traced by point (1,1) to point (i, j) of matrix D, and therefore p (g)M,tN) Is the minimum cumulative distance of the test sequence G (M) and the reference template sequence T (N), i.e., the DTW distance of the two.
Particularly, when the distance between two sequences is calculated by adopting a classic DTW algorithm, a larger matrix needs to be calculated and stored by using a dynamic programming method, the time complexity required by the calculation is O (mn), and in order to improve the calculation efficiency of calculating the distance between the gesture sequences by adopting the DTW algorithm, a global constraint window is introduced in the sequence bending calculation to avoid invalid path search. The CM-DTW algorithm uses a Sakoe-Chiba window to reduce the calculation of the distance between invalid data points, thereby improving the efficiency of calculating the two sequences. Under the global path constraint of fig. 1, the space required for the DTW algorithm to compute is not a complete matrix, but is limited to a strip-shaped area near the diagonal.
The Sakoe-Chiba global constraint can be understood as the point-to-point (g)i,tj) The restriction of the middle subscript ensures that j is more than or equal to i-f and less than or equal to i + f, and f is a constant; under the constraint of Sakoe-Chiba, the matching rule path of the test sequence G (M) and the reference template sequence T (N) for calculating the DTW distance is c1,c2,......,cK(wherein c isk(i, j)), as shown in fig. 2.
Because the lengths of the gesture data sequences may be different greatly, certain constraint needs to be made on the slope of the search path so that the subscript meets the requirement
Specifically, in the present embodiment, the normalization process (r (S) (S is the length of the gesture template sequence)) in step S126 has 3 cases:
i) when S ═ L, r (L) ═ r (S);
ii) when S > L, performing template shrinkage by down-sampling R (S); firstly, the ith data point r in R (S)iMapping to g in R (L)jThen the corresponding subscript mapping is implemented according to the following formula:
wherein
The data is rounded, if i corresponding to j is more than 1, the data points corresponding to i are averaged and assigned to gjOtherwise gj＝ri；
iii) when S < L, realizing R (S) extension by a linear interpolation method. Firstly, the ith data point r in R (S) is expressed according to the formulaiMapping to g in R (L)jThen all g' sjAre arranged in order of increasing j
Wherein
j is incremented from 1 if j < j1Then g isj＝gj1If j satisfies jk＜j＜jk+1Then the value can be calculated using the following equation:
in particular, the following experimental analysis was carried out in this example. The experimental simulation platform adopts Matlab R2017a (9.2.0538062), and the acquisition of experimental data is completed by a Wwaken p10 mobile phone. Experimental data Co-collection of 40 Ginseng, age distributed between 20 and 47 years, with 26 males and 26 females14 persons were born. When data are collected as required, the mobile phone is held by one hand, the front face of the screen of the mobile phone is upward, and the index finger of the other hand slides in a Z shape to unlock the screen according to the prompt of software. Each person collects 15 groups of gesture data in succession, for a total of 600 groups of gesture data. Of the 15 groups of gesture data collected by each person, 10 groups of data are used for training the gesture template of the algorithm, and the other 5 groups of data are used for testing the verification algorithm. The training data and the test data are subjected to preprocessing operations such as normalization and smoothing. In order to highlight the characteristics of the data, the highest weight is usually given to the data to be smoothed, and a weighted average of the data at the intermediate position is obtained as a result of smoothing. Therefore, the window size at the time of data smoothing processing is set to 5, while the weight values are sequentially set to 1,2,3,2, 1. Taking the value of the Sakoe-Chiba window constraint parameter f in the verification stage
First, the present embodiment performs identity authentication comparison between the standard template and the optimal template in the classical DTW algorithm.
In the experiment, identity authentication is performed on training data through a classical DTW algorithm, wherein an optimal template refers to a template for obtaining an optimal WA by using the training gesture data as a template for authenticating other training gesture data, the proportional variation range of a threshold value is (0.5,1.5), and the variation amplitude is increased progressively by 0.1. The identity authentication results of the two templates for the training data are shown in table 1 and fig. 4, where tp (true positive) is the proportion of the valid user sample that is correctly authenticated, and fn (false negative) indicates the proportion of the invalid user that is authenticated as the invalid user.
TABLE 1 TP and FN values for identity authentication of standard templates and optimal templates
It can be seen from table 1 that the TP and FN values of the majority of users who are performing identity authentication using the standard template are higher than the optimal template, and the standard template has a better effect than the optimal template on the average of the TP and FN of 40 users. Fig. 4 shows that the broken line of the standard template is higher than that of the optimal template, which shows that the standard template can obtain a better accuracy. Combining the results of table 1 and fig. 4, applying a standard template to the user's gesture authentication will yield good results.
Secondly, the CM-DTW algorithm of the embodiment is compared with the running time of the classic DTW algorithm.
In the experiment, training data of 40 users are selected, the distance between the same user gesture and the distance between the user gesture and other user gesture data are calculated by adopting a CM-DTW algorithm and a classic DTW algorithm respectively, and corresponding calculation time consumption average values are recorded, as shown in FIG. 5.
As can be seen from fig. 5, the time consumed for the authentication of the same user and different users by the CM-DTW algorithm is obviously shorter than that of the classic DTW algorithm, so that the authentication efficiency of gesture recognition is improved. In the authentication stage, the CM-DTW algorithm limits the optimal regular distance between the tested user data and the user template data in a Sakoe-Chiba window, and compared with the classical DTW algorithm, the CM-DTW algorithm reduces the calculation amount of the regular distance. Therefore, the CM-DTW algorithm is applied to the mobile phone identity authentication process, the user authentication time can be reduced, and the user experience effect is improved.
And finally, carrying out authentication comparison between the CM-DTW algorithm and the classical DTW algorithm.
In the experiment, the CM-DTW algorithm firstly obtains 40 groups of templates and optimal threshold values of users through training of training data, and performs identity authentication on test data with the classic DTW algorithm, wherein the proportional variation range of the threshold values is (0.5,1.5), the variation range is increased progressively by 0.1, the obtained optimal WA index value is used as a final verification result, and the gesture verification results of the two algorithms are shown in table 2 and fig. 6.
TABLE 2 TP and FN values for identity authentication of CM-DTW algorithm and classical DTW algorithm
From table 2, it can be found that the FN average value of the CM-DTW algorithm identity authentication is slightly higher and the TP average value is slightly lower than that of the classical DTW algorithm. Comparing the TP and FN values of 40 groups of users by the two algorithms, wherein the TP and FN index values of the two algorithms are respectively high and low. The broken lines for the two algorithms are shown in fig. 6 as being closer, indicating that the WA values for the CM-DTW algorithm and the classical DTW algorithm differ less. By combining the index values in table 1 and fig. 6, compared with the classic DTW algorithm, the CM-DTW algorithm can ensure accuracy in the identity authentication of 40 groups of user gesture data.
According to the dynamic gesture identity authentication method for the smart phone, the efficiency of dynamic gesture authentication of the smart phone is improved through global regular path constraint, a standardized gesture template is obtained by stretching and contracting the gesture data length through a normalization method, the problem that the selection of the gesture template is not accurate enough is solved, and the authentication precision of gesture identity authentication is ensured.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The foregoing is directed to preferred embodiments of the present invention, other and further embodiments of the invention may be devised without departing from the basic scope thereof, and the scope thereof is determined by the claims that follow. However, any simple modification, equivalent change and modification of the above embodiments according to the technical essence of the present invention are within the protection scope of the technical solution of the present invention.
Claims (3)
1. A dynamic gesture identity authentication method for a smart phone is characterized by comprising a gesture template training stage and an identity verification stage;
the gesture template training stage comprises the following steps:
step S11: collecting user information for training, carrying out data standardization and smooth denoising treatment on a plurality of pieces of collected user information, and selecting a candidate template;
step S12: normalizing the candidate templates by adopting linear up-down sampling to generate a standard gesture template of each user; authenticating the training set through a standard gesture template, and obtaining an optimal identity authentication threshold value under the constraint of a weighting accuracy index;
the authentication phase comprises the following steps:
step S21: collecting test gesture data of a user, and carrying out data standardization and smooth denoising on the test gesture data of the user;
step S22: calculating a CM-DTW regular distance by using the data processed in the step S21 and the user standard gesture template, wherein if the regular distance is smaller than the optimal identity authentication threshold value of the template, the test gesture data and the user standard gesture template data belong to the same user, otherwise, the test gesture data and the user standard gesture template data do not belong to the same user;
wherein, step S11 specifically includes: repeatedly collecting gesture data n times by a user, and collecting n groups of gesture data R1,R2,...,RnAcquiring m user data as interference gesture data in the same way; carrying out standardization and smooth denoising processing on n groups of data collected by a user, and dividing training data and test data;
wherein, step S12 specifically includes the following steps:
step S121: calculating regular distances between each group of training gesture data of the user and other training gesture data by using a DTW algorithm under the restriction of a Sakoe-Chiba window;
step S122: calculating the Mean value of regular distances between each group of gesture data and other gesture data of the same user as a standard threshold value of the gesture, wherein the Mean value of j groups of data of the ith user is Mean (i, j);
step S123: taking Mean (i, j) rate (i, j) as the optimal threshold value of the jth data of the ith user, and performing identity authentication on all other data; wherein rate (i, j) is the optimal threshold proportion of the jth group of data for the ith user;
step S124: based on the calculation result of step S121, under the optimal threshold classification of step S123, obtaining the identity authentication result of each group of gesture data, and evaluating it with a weighted accuracy index WA, calculating an identity authentication index value of each group of data, where WA (i, j) is a WA index value of a jth group of data of an ith user; calculating an index mean value AVG _ WA (i) of all data of the same user; when WA (i, j) ≧ AVG _ WA (i), the jth data is considered as the candidate template of the ith user; constructing a candidate template set by the gesture data meeting the requirement of the ith user candidate template, wherein the formula of the WA index is shown as the following formula:
in the formula, a is the legal gesture authentication as the legal gesture number, b is the illegal gesture authentication as the illegal gesture number, c is the illegal gesture authentication as the illegal gesture number, d is the illegal gesture authentication as the legal gesture number, and β is the corresponding weighted value, and the formula is expressed as follows:
step S125: averaging the lengths of all gesture data of the candidate template set to obtain an average value L;
step S126: in order to conveniently standardize the template, all gesture data in the candidate template set are subjected to stretching processing to be processed into a normalized template with the length L;
step S127: averaging the candidate templates processed in step S126 to obtain an average template, and using the average template as a standard template of the user
Wherein
A standard template of the ith user; will be provided with
Calculating regular distance with training gesture data at optimal threshold
To obtain an optimum WA (R)i) Index value, optimal threshold value
As a template
The optimal authentication threshold.
2. The method for authenticating the dynamic gesture identity of the smart phone according to claim 1, wherein in step S11 or step S21, when data collection is performed, an acceleration sensor and a pressure sensor built in the smart phone are used to obtain a data value during a sliding unlocking process of a user, and collection is started when the user slides an unlocked finger to touch a screen until the collection is finished when the finger leaves the screen; the data sequence collected from the beginning to the end is called gesture data, and the gesture data is defined as follows:
R＝{r1,r2,...ri,...,rn}；
in the formula, ri＝{ri1,ri2,ri3,ri4N is a 4-dimensional data, where r is 1,2i1、ri2、ri3X, Y, Z triaxial parameter value r of built-in acceleration sensor of mobile phonei4Is the value of the built-in pressure sensor of the mobile phone.
3. The method of claim 1, wherein in steps S11 and S21, the data normalization specifically comprises:
the gesture data is standardized by adopting a z-score standardization method, and the result after the standardization of the gesture data is shown as the following formula:
in the formula, R is a data sequence of the gesture data set R, μ is a mean value of R, σ is a standard deviation of R, and m represents the number of gesture data in the set R.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910571581.7A CN110298159B (en) | 2019-06-28 | 2019-06-28 | Smart phone dynamic gesture identity authentication method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910571581.7A CN110298159B (en) | 2019-06-28 | 2019-06-28 | Smart phone dynamic gesture identity authentication method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110298159A CN110298159A (en) | 2019-10-01 |
| CN110298159B true CN110298159B (en) | 2021-04-20 |
Family
ID=68029245
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910571581.7A Active CN110298159B (en) | 2019-06-28 | 2019-06-28 | Smart phone dynamic gesture identity authentication method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110298159B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110866231B (en) * | 2019-10-14 | 2022-06-24 | 北京邮电大学 | Server, intelligent terminal and implicit identity authentication system and method |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103442114A (en) * | 2013-08-16 | 2013-12-11 | 中南大学 | Identity authentication method based on dynamic gesture |
| CN106411952A (en) * | 2016-12-01 | 2017-02-15 | 安徽工业大学 | Telekinetic-dynamic-gesture-based user identity authentication method and apparatus |
-
2019
- 2019-06-28 CN CN201910571581.7A patent/CN110298159B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103442114A (en) * | 2013-08-16 | 2013-12-11 | 中南大学 | Identity authentication method based on dynamic gesture |
| CN106411952A (en) * | 2016-12-01 | 2017-02-15 | 安徽工业大学 | Telekinetic-dynamic-gesture-based user identity authentication method and apparatus |
Non-Patent Citations (4)
| Title |
|---|
| ULTRASOUND BASED GESTURE RECOGNITION;Ivan Tashev .et al;<Microsoft Research One Microsoft Way>;20171208;page 1-6 * |
| 基于ＤＴＷ的时间序列流相似性搜索方法;陶洋 等;《计算机工程与设计》;20171231;第38卷(第12期);第3292-3294页 * |
| 基于DTW算法的多模板组合手机身份认证方法;李翔宇;《福建电脑》;20180430(第4期);第8-9页 * |
| 基于手机触摸屏传感器多点触摸身份认证算法;庞永春 等;《计算机应用》;20150610;第35卷(第6期);第1781-1784页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110298159A (en) | 2019-10-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Soltane et al. | Face and speech based multi-modal biometric authentication | |
| Fallah et al. | A new online signature verification system based on combining Mellin transform, MFCC and neural network | |
| KR101252454B1 (en) | Apparatus and method for selecting representative fingerprint template | |
| KR102449839B1 (en) | Authentication method based on the ecg, authentication apparatus, learning method for ecg based authentication and learning apparatus | |
| EP2523149A2 (en) | A method and system for association and decision fusion of multimodal inputs | |
| CN103927532B (en) | Person's handwriting method for registering based on stroke feature | |
| CN104887263B (en) | A kind of identification algorithm and its system based on heart sound multi-dimension feature extraction | |
| CN106340298A (en) | Voiceprint unlocking method integrating content recognition and speaker recognition | |
| CN102045162A (en) | Personal identification system of permittee with tri-modal biometric characteristic and control method thereof | |
| MX2012010637A (en) | Biometric information processing device. | |
| US20140195232A1 (en) | Methods, systems, and circuits for text independent speaker recognition with automatic learning features | |
| CN109657528A (en) | Fingerprint identification method and the electronic device for using fingerprint identification method | |
| DS et al. | Implementing and analysing FAR and FRR for face and voice recognition (multimodal) using KNN classifier | |
| CN110298159B (en) | Smart phone dynamic gesture identity authentication method | |
| CN106650685B (en) | Identity recognition method and device based on electrocardiogram signal | |
| CN113241081B (en) | Far-field speaker authentication method and system based on gradient inversion layer | |
| Vasuhi et al. | An efficient multi-modal biometric person authentication system using fuzzy logic | |
| Soltane et al. | Soft decision level fusion approach to a combined behavioral speech-signature biometrics verification | |
| KR101208678B1 (en) | Incremental personal autentication system and method using multi bio-data | |
| CN114840834A (en) | Implicit identity authentication method based on gait characteristics | |
| Ye et al. | A study on the individuality of finger vein based on statistical analysis | |
| Chaitanya et al. | Verification of pattern unlock and gait behavioural authentication through a machine learning approach | |
| Sukor et al. | Speaker identification system using MFCC procedure and noise reduction method | |
| Yang et al. | Person authentication using finger snapping—a new biometric trait | |
| Wahid et al. | A Gaussian mixture models approach to human heart signal verification using different feature extraction algorithms |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |