ES2338092A1 - Device and method of biometric recognition based on a movement of the hand in the space with a device that integers an accelerometer (Machine-translation by Google Translate, not legally binding) - Google Patents

Abstract

Device and method of biometric recognition based on a movement of the hand in space with a device that integrates an accelerometer. Device and method of biometric recognition based on the movement of the hand in space, which includes the enrollment stage and the access stage. The enrollment includes: obtaining signals according to the axes of space from the repetition of a manual gesture; sample and store them; process, each pair of signals sampled in each of the axes of the space, comprising: softly adapt the sampled signals to obtain two aligned sequences, interpolate and quantify the differences; and, as a final step, obtain and store a gestural pattern for authentication. The access stage comprises: obtaining access signals from a manual gesture; sample and store them; quantify differences between the resulting sequence and the gestural pattern; calculate the average deviation from the deviations obtained from each of the access signals with the gestural pattern; and decide access, as a final stage. (Machine-translation by Google Translate, not legally binding)

Description

Device and method of recognition biometric based on a movement of the hand in space with a device that integrates an accelerometer.

Field of the Invention

The present invention belongs to the technical field of biometrics and more specifically to the field of cryptobiometry as it allows the interaction of characteristics of Human morphology with cryptographic foundations. Sectors like biometric engineering or related to the safety of communications, especially to authentication tasks in systems of access, they would also belong to the technical field of the invention.

State of the art

The goal of a biometric system is the identification / authentication of people through some biological characteristic physically measured or of individual behavior, safely and non-invasively. He problem of identification and authentication of people is very old and has always tried to solve with different Media: sealed, titles, nameplates ... Today this is not enough and it is necessary to introduce new techniques of authentication and identification to ensure that a person is who He claims to be in a multitude of contexts.

The system to be patented is about perform a biometric authentication that brings the two together General characteristics of biometric authentication: physical characteristic and behavior.

There are many biometric techniques that they try to recognize a person for their physical characteristics (iris, face morphology, fingerprint ...) or by their behaviors (way of walking, way of writing, signature on-line...). It is of vital importance in this document, for its similarity, the biometric technique of the online firm. They have developed many works to improve this technique. In them explain the basis of verification by online signature. In this type of verification, it is about comparing so much that the drawing of the signature made is equal to one stored, as the way of making said signature matches how the user did it registered. To do this, different parameters are measured at the time of make a signature, such as writing speed, pressure exerted or the angle of the pen at each moment of the time in which the signature is made, among other features. These signatures can be done on a special screen that collects and analyzes all the necessary signals for analysis or on paper if the pen with which it is signed is able to measure the signals described above and send them to a server where it is done the analysis and verification of the signature.

In the MX2007007539 patent a system is collected which implements biometric authentication by signature electronics. This system includes an interface on a capable computer of storing the movement of a cursor on a screen of the computer and compare it with already stored signature patterns.

The object of the present invention is about Authenticate on a mobile device. Nowadays, There are many applications that can be accessed from a mobile terminal where an identification of the Username. For years, all security was entrusted to mobile devices to type a secret key that the user I knew However these keys can be easily forgotten, transferred, lost and even counterfeit, so the User authentication is compromised.

According to this, WO2008111012 patent deals with perform an authentication system on mobile devices that include a scanner to collect biometric data and send it to a server that will perform the analysis and pattern check necessary. The patent describes a mobile identification device associated with an authentication and identification system. The device can collect biometric data from a user and through a processor compare them with the biometric data stored by the legitimate user and give access to a system in case the Comparison is made successfully. Through a wireless connection, the processor can transmit the information associated with the identification to another terminal, such as a mobile phone. For it, it is necessary that the authentication server where the comparison have access to a database that stores both the Biometric data such as the identifications of each device mobile.

Focusing on the biometric technique to authenticate a user with a mobile device has been found US2006286969 and US2008005575.

In US2006286969 a system of remote authentication to authenticate users from a device mobile. The biometric technique used is the recognition of the voice. The system consists of a mobile phone that sends some Voice samples to an authentication device that connects to a database that stores mobile phone identities and the voice pattern associated with said telephone to perform a comparison and check that the user who is speaking for the Phone is the user registered in the system.

In US2008005575 a method and a device to authenticate a user on a mobile phone. While the user keeps the phone in his ear, a nearby microphone emits a signal in the user's ear and the speaker of the phone is able to measure the response of the ear to that signal. A processor analyzes the response signal and converts it into a signature, which uniquely identifies each person and can be used to authenticate.

The object of the present invention proposes the making a gesture with the mobile device that identifies a user, considering that this gesture is only going to be known by the user and also, because of its physical characteristics, it is going to perform differently from other people who could Try to repeat your gesture.

Several recognition techniques are known of gestures, in which a system is able to detect when a User makes a gesture. Some of these techniques use the gesture recognition to check if patients in a Residence are taking dinner food. For it, they use the video signals of the dining room and techniques of recognition of head and arm gestures, analyzed by Hidden Markov Models (HMMs). Preferably studies are conducted applying HMMs to recognize hand gestures corresponding to Arabic numbers from 0 to 9.

Patents have also been found US2009103780 and WO2009006173 related to methods to recognize standard gestures US2009103780 includes a method for pick up the gestures produced by one hand, based on illuminating the hand at first by the palm and the back, to get Its associated silhouette. From different illuminations infrared, a method is proposed to collect the different hand movements and identify them with a series of gestures previously stored in a database of gestures obtained from The same way.

WO2009006173 describes a method for electronically detect the response gesture of a user who You are listening to an announcer using a mobile device. To the make a specific gesture, the speaker is informed of an answer corresponding audible or visual.

Related to the idea of recognizing a person by performing a gesture has found the WO2007134433. It develops a method to authenticate a user when performing an action that includes the Manual manipulation of a device like a mouse. The authentication consists of obtaining the gesture made with the mouse by the user when pursuing a goal and compare it with the stored pattern of the user when he has pursued that goal previously.

Focusing the state of the art towards present invention, it should be noted that the realization of gestures for biometrically authenticate a person on a mobile device through gestures measured with an accelerometer is novel.

Regarding the use of accelerometers in devices mobile in authentication systems there is US2005226468 that proposes perform user authentication based on certain sensors biometrics that must be connected to the mobile device, and check that the authentication was successful on the basis to an accelerometer that collects data on how to catch the user device, making sure that it is not a machine the That tries to fool the system.

Also, in US2009030350 a method and a system to analyze patterns of ways of walking of a subject measuring head acceleration in the vertical direction while go To do this, use an accelerometer that is placed on the head of the user. The analysis includes the creation of a signature from of the acceleration data at the time of walking of the user.

The present invention also proposes the use of the patterns obtained by the performance of the gesture of the user for the generation or release of a key cryptographic In this regard, patents have been found DE102005010698 and KR749380-B1.

In DE102005010698 the construction of a secure cryptographic key independent of communication to Starting from the fingerprint. It is proposed to use this key for television communication on demand, with applications of payment by vision, protection of the child or verification of age.

A KR749380-B1 explains a method to generate a biometric key from a Biometric feature that does not change over time like the iris. Biometric information is received and preprocessed, some are extracted values and the associated cryptographic key is obtained through the grouping of values. The grouping error is corrected using a block of Reed-Solomon code. The key obtained can be applied to any cryptographic system.

It was therefore desirable a device of biometric recognition such as the one described herein invention to avoid the inconveniences existing in the prior state of the art systems.

Description of the invention

The invention solves existing problems in the state of the art by means of a device and a method of biometric recognition for non-invasive authentication without need for an additional device to perform said recognition.

In the present invention, a authentication by performing a gesture in space with a mobile device with an integrated accelerometer.

The present invention improves the manner of make the signature with respect to the one presented in MX2007007539 since it performed using a 3D gesture with a device that integrates a accelerometer, being completely different from that proposed in said patent and in all online signature systems that what they do It is a handwritten signature on some surface. In the case of The present invention also introduces a third component for perform the gesture, corresponding to the z plane of the space.

With respect to WO2008111012, the invention You do not need to use a mobile device to collect data biometric, since the accelerometer presented in the device is fully integrated into the mobile device, and the authentication / identification can be performed directly on the own device. In addition, biometric data to perform the verification of the authenticity of the person accessing are different.

US2006286969 and US2008005575 patents propose different methods to perform the authentication of a User on a mobile device. However, the proposed method proposes a completely different biometric technique, such as the making a gesture with the hand holding the device, in such a way so that the user will be able to repeat it without difficulty and In addition, any user who tries to imitate him will make the gesture in a subtly different way because of the different physical and behavioral characteristics (way of taking the mobile, way of turning the mobile) intrinsic to each person.

US2009103780 and WO2009006173 patents collect methods to identify the different gestures made by the User hands In this the approach of the proposed method is completely different, since it tries to recognize people by the gesture they make and how they do it, as opposed to recognizing gestures that people make. In addition, the methods to recognize the gestures are also completely different, since in the proposed system makes use of an accelerometer that picks up the gesture of the person unlike the lighting technique of US2009103780 or electronics for simple gestures of WO2009006173.

Regarding WO2007134433, the proposed device presents a behavioral identification of the user differently. The system described in the patent WO2007134433 identifies a user by the movements he makes with the mouse when pursuing a target while the system which is intended to patent identifies you by performing a own gesture with hand in space.

US2005226468 uses an accelerometer and performs biometric authentication on a mobile device, but the use of the accelerometer and the way of performing authentication are different. The patent system US2005226468 uses sensors connected to the mobile device to verify the identity of a user, using the accelerometer to verify that the device is held by a person. In a novel approach and in contrast to the method explained in US2005226468, the proposed method directly uses the accelerometer to perform authentication based on a gesture made by the user keeping the mobile device with integrated accelerometer in the hand, offering as an advantage that the device proposed does not need any external sensor, but uses the accelerometer integrated in the device itself to perform the checks
Biometric

US2009030350 patent proposes the use of a accelerometer to identify people by the way they walk. In the proposed device exposes the use of the accelerometer to collect the gestures made by keeping the mobile device with integrated accelerometer by hand, and in this way check the Identity of a user who wants to access the system.

Patents DE 102005010698 and KR749380-B1 propose methods to generate a key unique cryptographic based on a biometric characteristic of user, such as the fingerprint on DE102005010698 or the iris in KR749380-B1. The proposed method has a completely different approach, as is the generation of a cryptographic key from the signals obtained from a gesture made by the user using a device with a integrated accelerometer. The use of a feature Static biometrics such as the iris or the footprint will always generate the same key applying certain algorithm. However, the system proposed offers the user the possibility of generating a password "false" that can happen as correct for any observer external, making a different gesture to the one that characterized him, maintaining all the advantages of key generation unique cryptographic from biometric patterns when performs the characteristic gesture of the user.

Finally, it should be noted regarding the existing literature in this field that identification through drawing a graph is different from the one proposed by the present invention since the realization of this graph is done in the three axes of space, not being necessary to use a paper or a screen where to make a handwritten signature.

The present invention proposes the use of a device that includes an accelerometer for tasks authentication from the realization of a gesture. The technique it includes both the recognition of a person when making a gesture determined, such as the possibility that the person creates a gesture own to identify in a system.

       \ newpage
    

Due to the accelerometer, you can get the speed variation signals of the device with which make the gesture, in the three dimensions of space. This form, a user is going to authenticate in the system making a gesture in the air holding the device that integrates the accelerometer To do this, you must first register with the system, making the gesture that will identify you so generate a gestural biometric pattern according to their identity. In the following access, the user will try to access the system repeating the gesture used in the registration phase. The Biometric authentication will be performed comparing the gesture performed by the user in the access with the biometric pattern of the gesture carried out in the registration phase, taking into account that only the Registered user will be able to accurately repeat the gesture with The one who identified himself.

This biometric authentication technique is novel. In fact, the realization of a three-dimensional gesture it makes it very difficult to copy, since a person who visualizes the gesture will not have the reference of a plane, as in the technique of the handwritten signature that has a fixed role in which firm. It should also be noted that a person by his characteristics physical will pick up the device and perform the gestures of a different way like wrist twists or arm movement, so even knowing how is the gesture that makes a person, it will be very difficult to fake, since each one will differently

On the other hand, the accelerometer is integrated in many mobile devices, such as phones, PDAs ... so no an external device is needed to perform the identification, as in the case of the firm that a Touch screen or nearby equipment to analyze movements. This feature is of relevant importance since currently many of these accelerometer devices have connection to Internet. Thus, this method provides a way to perform a online biometric authentication for secure communications without the need of any special peripheral.

Therefore, the recognition method biometric based on the movement of the hand in space, understands:

? one enrollment stage in turn comprising:

to)

obtain, from at least one accelerometer integrated in a biometric recognition device, an acceleration signal ( G_ {1x}, G_ {1y}, G_ {1z} ) from the realization of a hand gesture carrying in hand said device;

b)

repeat step a) n times;

C)

sample the acceleration signals previously obtained ( G_ {1x}, G_ {1y}, G_ {1z}; G_ {2x}, G_ {2y}, G_ {2z}; ... );

d)

store sequences sampled;

and)

process each pair of signals sampled on each of the axes of space, comprising:

\ sqbullet

adapt the signals sampled smoothly to get two aligned sequences, one for each gesture performed on a given axis of the space;

\ sqbullet

interpolate the errors of sequences aligned with the samples of the signals obtained in a) and in b) to correct the variations of speed in the repetition of the gesture;

\ sqbullet

quantify the differences between the sequences adapted and interpolated by each pair of gestures made on a certain axis of space;

F)

get and store a gesture pattern for user authentication;

an access stage which in turn understands:

g)

obtain, from at least one accelerometer integrated in a recognition device biometric, an access signal from the realization of a manual gesture carrying in hand said device;

h)

sample the signals from access;

i)

store the access sequence sampled;

j)

quantify the differences between the sampled access sequence and gestural pattern;

k)

calculate the average deviation from of the deviations obtained from each of the access signals with the gestural pattern;

l)

decide access by comparing the mean deviation with the gestural pattern.

Preferably, obtaining the signals of acceleration and access comprises the decomposition of the manual gesture acceleration in each of the three axes of the space.

Preferably, the adaptation is performed using a global sequence alignment method.

In addition, the gestural pattern may comprise the signals obtained in a) and b) and the average distance between quantified differences between each pair of gestures, where said average distance measures the similarity between the repetitions of the gesture.

In a preferred embodiment, the quantification of differences of the access stage is performed between the sequence of sampled access and stored signals as part of the pattern gestural

Preferably, the calculation of the deviation media is performed with each of the signals stored as part of the gestural pattern.

In a preferred embodiment, the decision of the access is made by comparing the average deviation with the distance average stored as part of the gestural pattern.

The biometric recognition device based on the movement of the hand in space, it includes:

at least one configured accelerometer to obtain acceleration signals;

configured processing media for the realization of a enrollment stage and a stage of access where:

\ circ

the enrollment stage understands:

m)

obtain, from at least one accelerometer integrated in a biometric recognition device, an acceleration signal ( G_ {1x}, G_ {1y}, G_ {1z} ) from the realization of a hand gesture carrying in hand said device;

n)

repeat the previous step n times;

or)

sample acceleration signals ( G_ {1x}, G_ {1y}, G_ {1z}; G_ {2x}, G_ {2y}, G_ {2z}; ... );

p)

store sequences sampled;

q)

process each pair of signals sampled on each of the axes of space, comprising:

\ sqbullet

adapt the signals sampled smoothly to get two aligned sequences, one for each gesture performed on a given axis of the space;

\ sqbullet

interpolate the errors of sequences aligned with the samples of the signals obtained in m) and in n) to correct the variations of speed in the repetition of the gesture;

\ sqbullet

quantify the differences between the sequences adapted and interpolated by each pair of gestures made on a certain axis of space;

r)

get and store a gesture pattern for user authentication;

\ circ

the access stage understands:

s)

obtain, from at least one accelerometer integrated in a recognition device biometric, an access signal from the realization of a manual gesture carrying in hand said device;

t)

sample the signals from access;

or)

store the access sequence sampled;

v)

quantify the differences between the sampled access sequence and gestural pattern;

w)

calculate the average deviation from of the deviations obtained from each of the access signals with the gestural pattern;

x)

decide access by comparing the mean deviation with the gestural pattern.

Preferably, the processing means are configured to obtain signals by means of decomposition of the acceleration of the gesture in each of the three axes of space.

       \ newpage
    

Preferably, the processing means are configured for adaptation using a method of global sequence alignment.

In addition, the processing means are configured to obtain the gestural pattern from the signals obtained in m) and n) and the average distance between quantified differences between each pair of gestures, where said average distance measures the similarity between the repetitions of the gesture.

In a preferred embodiment, the means of processing are configured for quantification of differences in the access stage between the access sequence sampled and the signals stored as part of the pattern gestural

Preferably, the processing means are configured to perform the deviation calculation average with each of the signals stored as part of the pattern gestural

In a preferred embodiment, the means of processing are configured to make the access decision comparing the average deviation with the average distance stored as part of the gestural pattern.

Brief description of the drawings

Then, to facilitate the understanding of the invention, by way of illustration but not limitation, will be described an embodiment of the invention that refers to a series of figures.

Figure 1 shows, on each axis of space (G_ {1x}, G_ {1y} and G_ {1z}), the decomposition of the signals obtained by an accelerometer integrated in a device when a person makes a gesture by holding said device with the hand.

Figure 2 shows, on each axis of space, two embodiments of the same gesture by the same user in different moments of time.

Figure 3 shows, on each axis of space, the result obtained after the alignment of the two signals shown in figure 2.

Figure 4 shows, on each axis of space, the result obtained after interpolation of the two signals adapted from figure 3.

Figure 5 shows, on each axis of space, the final result after analyzing a case in which the signals correspond to the realization of two different gestures.

Figure 6 shows a scheme of the device Biometric recognition consisting of an accelerometer and media of processing.

Detailed description of one embodiment

The object of the invention is to present a device and a biometric recognition method to authentication by using a device that includes an accelerometer to recognize a person by performing of a gesture.

Figure 1 shows the signals G_x, G_ {1y} and G_ {1z}, which are the decompositions on each axis of the signal space obtained by an accelerometer integrated in a device when a person performs an example gesture with the hand holding the device containing the accelerometer.

Figure 2 shows the similarity between two repetitions of performing the same gesture for the same user in different moments of time. It is observed that the form of the signals in each of the axes is very similar, and that a small global delay since the user is not able to start the gesture exactly at the same time in each of your repetitions. Further, a small local delay can be observed, which has to do with the impossibility of the same user repeating the same gesture at exactly the same speed, and can perform parts of the gesture a little faster or slower depending on the repetition.

Figure 3 shows the result of processing the signals of figure 2 with the algorithm that will be explained to continuation. In this figure 3 the two signals of the repetition of the same gesture by the same user are completely aligned based on having entered zeros in each of the sequences

Figure 4 shows the result of having interpolated the two adapted signals of figure 3. In this figure 4 you can check how the signals have been adapted, and visually it can be verified that the gesture is the same.

Figure 5 shows the final result of analyze a case in which the signals correspond to two gestures different. Despite having optimized the adaptation between the two signals, being different, the result is not good. Thus, when defining a distance to quantify the difference in signals, it can be deduced that the gestures are not equal.

Figure 6 shows a scheme of the device of biometric recognition.

As an example, and without pretending to be limiting its scope, a possible way of describing embodiment of the invention.

       \ vskip1.000000 \ baselineskip
    

1. Introduction

To identify, the user performs with a mobile device that integrates an accelerometer (for example a iPhone) a gesture in the air. From the signals collected by the accelerometer of the speed variation in each of the axes from space a pattern will be generated, which will identify the person.

The process of identifying a person through this biometric technique it has two phases: the phase of enrollment or creation of the pattern and the phase of the access attempt repeating the gesture that has been used for the creation of Pattern.

An embodiment is explained below. of each of the phases, applying a specific algorithm. These Same phases can be implemented in many other ways. In this document has chosen this one in particular because of the good results that are obtained and to the explanation clearly of the philosophy of technique that is presented.

       \ vskip1.000000 \ baselineskip
    

2. Enrollment phase

In this phase, the user has to repeat so as accurately as possible a gesture a certain number of times to that the system can generate a pattern of the gesture with which to compare the following user accesses.

One way to perform this phase is that the user perform by hand, holding the device that integrates the accelerometer, three times the same gesture with which you are going to Identify in the system on subsequent occasions.

The system is able to collect the data collected by the accelerometer in each of the gestures and after a analysis and processing generates a pattern that identifies the user and it allows to decide if the user is who he says he is when performing gesture to access in the access phase.

Each of the following will be explained below. necessary steps to perform this phase in detail:

2.1. Making gestures and obtaining signals

The device that integrates the accelerometer offers the following instruction to a user when enrolling in the system:

You have to perform the same gesture three times, taking into account the following factors:

?

He must be able to remember it and Repeat it in a simple way. It is advised to choose a gesture that remember something or that the user can do in a way natural.

?

It has to be enough complex so that another person who sees the gesture can not Copy it immediately.

?

It must last a maximum of six seconds.

For each of the three gestures ( G_ {1}, G_ {2}, G_ {3} ) the accelerometer of the device will obtain three different signals ( G_ {1x}, G_ {1y}, G_ {1z}, G_ {2x }, G_ {2y}, G_ {2z}, G_ {3x}, G_ {3y}, G_ {3z} ) corresponding to the accelerations of each of the gestures made in each of the three axes of space.

In Figure 1, the signals G 1x, G 1 and G 1z obtained from the performance of a gesture by a user can be observed. Figure 2 shows how a user has been able to repeat the same gesture. It is found that the shape of the signals on each axis is very similar but that they are not aligned and that there are narrower parts of the signal in one sample with respect to the other because the user repeats the gesture slightly varying its speed. These differences of the signals are those that will try to correct with the processing of the same ones that will be exposed later.

For the realization of the system it has been used a device with an accelerometer that samples the variations of the hand movement at a sampling frequency of 100 Hz, precise enough to obtain representative signals of the movement of the hand in space. Given that it has chosen a maximum gesture size of 6 seconds, the accelerometer It will generate a signal on each axis of 600 points. Yes the gesture performed by the user lasts less than the maximum time selected, the rest of the signal points will be filled with zeros.

The signals of the three axes of each gesture are grouped into one, forming a signal of 1800 points that will be the It is stored on the device. Therefore, in the device they will keep three signals corresponding to each gesture with the Accelerations in the three axes.

The pattern of the gesture made is formed by the three signals of each gesture and by certain parameters that they will obtain from the analysis and processing of the signals that will be explained then.

2.2. Signal Processing

The signals are going to be processed two by two, making a processing on each axis separately. This processing will try to adapt the signals to align and correct small variations in time when making the same gesture, of such that two gestures of one can be identified as equal simple way. For this step, it will be necessary to explain the following points:

?

Soft adaptation algorithm of two signals

?

Signal interpolation adapted.

?

Define a metric for quantify the difference of two adapted signals.

In the enrollment phase these three phases are will do for each pair of gestures in each of its axes, being therefore necessary nine signal processing. In the Next steps will explain how from the metrics obtained when processing the signals that form the pattern, will be defined the threshold of rejection and acceptance for the rest of the accesses.

Next, each of the points necessary to process a pair of signals of two gestures on the same axis will be detailed, for example, G_ {1x} and G_ {2x} .

2.2.1. Smooth signal adaptation

For the smooth adaptation of two signals it has been used a Global Sequence Alignment method. These methods are usually used to align two strands of DNA from such so that the resulting alignment is optimal. Is done by dynamic programming to obtain the best result of possible alignment based on a defined score.

To align two point sequences so optimal the algorithm can introduce three types of changes that improve the overall alignment of the sequences:

?

A mutation, which changes the value of one point of a sequence by another.

?

An insert, which adds one or Multiple positions in a sequence.

?

A deletion that removes a or more point positions of a sequence.

To find the maximum alignment of two sequences ( A = {a_ {1}, a_ {2}, ..., a_ {m}} and B = {b_ {1}, b_ {2}, ..., b_ {m}} Dynamic programming has been used following the following steps:

?

Creation of the matrix M :

A matrix M of range mxm is created . In the upper part of the matrix (row 0), the m values of the sequence A are placed , and in the left part (column 0) the corresponding is done with the values of the sequence B.

?

Movement through matrix M :

The objective of the algorithm is to find the path from M (0,0) to M (m-1, m-1) that maximizes the score of the aligned sequences. To do this, at each point of the matrix you have to find the maximum of the possible scores of the signals at each point, which are defined by equations (1), (2) and (3):

one

where h is a parameter that can be defined. The best results in this specific case have been obtained with h = 0.125 .

In the algorithm a part of decision by fuzzy logic, which is equivalent to the Δ of the equation (2). It has the expression of equation (4):

2

where \ mu = A ( i -1), x = B ( i -1) and \ sigma the Gaussian parameter to be defined; In this model the value \ sigma = 0.125 has been used.

?

Trail Generation :

The maximum evaluation of equations (1), (2) and (3) is done at each point of the matrix. Once done, start from point M (0,0) and if it comes to reach point M (m-1, m-1) by crossing the matrix points following the following rules for any point M (i, j) :

\ circ

If the maximum of the equations at point M (i, j) is that of equation (1), the next point that is evaluated is the M (i + 1, j) , horizontal movement.

\ circ

If the maximum of the equations at the point M (i, j) is that of the equation (2), the next point that is evaluated is the M (i + 1, j + 1) , diagonal movement.

\ circ

If the maximum of the equations at point M (i, j) is that of equation (3), the next point that is evaluated is M (i + 1, j + 1) , vertical movement.

The trail is a vector R = { r_ {1}, r2 , ...} that gathers the path that has been followed taking into account the previous rules to reach the point M (m-1, m-1 ) starting from M (0,0) . The trail will have a maximum size of 2m and will be filled according to the following criteria:

\ circ

A diagonal movement introduces a 1 on the trail.

\ circ

A horizontal movement introduces a 2 in the trail.

\ circ

A vertical movement introduces a 3 in the trail.

?

Alignment according to the trail :

Once the trail is obtained the aligned signals They can be obtained from it as follows:

They are called A '= { a' 1, a '2, ..., a' lr 'and B ' = { b '1, b' 2, ..., b '_ {lr} } to the new aligned signals to be constructed. These will have a length equal to the length of the vector R. Sequences A ' and B' are to be filled in with the corresponding values of the initial sequences A and B or with zeros so that the signals are perfectly aligned. The counters of the initial sequences A and B are called indexA and indexB . For each point r_ {i} of the trail the following is done:

\ circ

If r_ {i} = 1 then, a '= a_ {indexA} and b' i {b} {index B} . In addition, indexA = indexA +1 and indexB = indexB +1. No zeros are introduced in the sequences.

\ circ

If r_ {i} = 2 then, a '_ {= a_ {indexA} , b' i = 0 and indexA = indexA +1. A zero is entered in sequence B.

\ circ

If r_ {i} = 3 then, a 'i = 0, b' i = b_ {indexB} and indexB = indexB +1. A zero is entered in sequence A.

In this way it is achieved that the best possible alignment of sequences A and B , which correspond to sequences A ' and B' . The aligned sequences may have a length m <length <2 m depending on the alignment. In order not to work with signals of different lengths, it has been decided that the size of the aligned sequences is always 2 m , so the sequence is filled with zeros until such length is completed.

Figure 3 shows the result of process the signals of figure 2 with the described algorithm. Be check that the signals have been adapted based on the introduction of zeros in each of the sequences.

2.2.2. Interpolation

As a consequence of the previous step they have obtained two sequences aligned with possible ranges of zeros to that the alignment has been optimal. This step is to correct the mistake that those zeros introduce, interpolating those zeros to the values of the original sequences among which introduced those zeros. Interpolation is performed in each of the sequences aligned as follows:

A range of zeros is assumed that goes from the ith position to the jth position of the sequence A: I (i, j) \ Rightarrow a 'i = a' i + 1 =. .. = a 'j = 0.

Then, for each i <j it applies that a '_ {i} = \ frac {\ left (a' _ {+ 1} + a '_ {j + 1} \ right)} {2}.

In this way, they adapt more faithfully point to Point the two processed signals. The effect of the improvement it offers the interpolation step can be seen in figure 4. It you can check that the signals have adapted perfectly and visually it is observed that after this analysis the gesture is the same.

Due to the adaptation steps e interpolation, two optimally aligned signals are obtained and with the correction of small speed variations at the time of make the gesture Applying this same process to two signals generated from different gestures or the same gesture made by different people, it can be seen in figure 5 that the result at a glance is not so good, which will allow define a distance that quantifies the difference between signals, identify if the two signals that are processed are the result of The same user has made the same gesture or not.

2.2.3. Metric to quantify pattern differences aligned

Once the two sequences are interpolated, you can apply the Hamming distance point to point (equation 5) to Quantify how different the signals are:

3

2.3. Obtaining the gesture pattern

At the beginning of the enrollment phase the user perform the same gesture three times (G_ {1}, G_ {2}, G_ {3}) and get the signals corresponding to their acceleration on each axis. In the previous point it was explained how the sequence processing two to two, to finally get the distance δ that related them.

To obtain the pattern it is necessary to obtain the distances of each pair of signals on each axis, therefore, will be It is necessary to perform 9 processes to obtain: \ delta_ {1x, 2x}, \ delta_ {1y, 2y}, \ delta_ {1z, 2z}, \ delta_ {1x, 3x}, \ delta_ {1y, 3y}, \ delta_ {1z, 3z}, \ delta_ {2x, 3x}, \ delta_ {2y, 3y} and \ delta_ {2z, 3z}.

Once all distances have been obtained above, the distance between each of the gestures with equation 6:

4

From the distances \ delta_ {1,2}, \ delta_ {1,3} and δ_ {2,3}, the distance can be obtained mean of the pattern, which will be a very important parameter to define the access threshold as will be seen later. Distance Mean of the pattern is defined by equation 7:

5

The pattern of the gesture made by the user that it is stored on the device for user authentication in Following access consists of:

?

The three signals G_ {1}, G_ {2} and G_ {3} that include the accelerations in each axis of the repetition of the same gesture by the user.

?

The \ mu parameter that gives a idea of the same as the three repetitions of the gesture made by the user.

The convenient thing would be that being the difficult gesture to imitate by a person who can observe the user, the user Be able to repeat it very roughly. It is necessary that the value of \ mu is small to ensure that the created pattern will be complicated to imitate.

       \ newpage
    

Note that the above is just a way of perform the system but there are a lot of ways to generate and combine the signals to form a pattern.

       \ vskip1.000000 \ baselineskip
    

3. Access phase

The access phase will be carried out in three Steps:

3.1. Realization of the gesture and signal store

A user can try to authenticate in the system if the enrollment phase has previously passed. In that case, to have access to the system repeat the gesture from the which created its biometric pattern. The device will record the accelerations of the gesture in the three axes, obtaining the signals G_ {Ax}, G_ {Ay} and G_ {Az}.

3.2. Access signal and pattern processing

Next, an adaptation of signals, as explained in section 2.2.1, in each of the axes of the access signal with each of the three signals that form the biometric pattern, resulting in distances \ delta_ {A, 1}, \ delta_ {A, 2} and \ delta_ {A, 3}.

From them the deviation of the access signal with each of the signals that form the pattern, following equations (8), (9) and (10).

6

Finally, the average deviation is calculated as the average of the three deviations obtained in the previous step, such and as required in equation 11.

7

This calculated average deviation is the metric used to compare gestures with patterns. The smaller it is, more will approach the average distance of the pattern, and therefore will be more faithful to the pattern stored in the device.

3.3 Access decision

The user will be successfully authenticated in the system, if you have been able to perform the gesture with a deviation regarding the pattern smaller than the average deviations of the pattern signals; that is, if you meet the equation:

8

Finally it should be noted that the application industrial of this system. First, any technology of physical or logical access control can take advantage of this technology, since it allows to identify a person and recognize if This person is someone registered in the system.

On mobile devices, such as a mobile phone, it would not be necessary for example to enter the PIN number to unlock the phone, and this feature can Extrapolate to any system that requires an access code such as an ATM, a password stored in a computer, etc.

In electronic commerce, it can be used to authenticate a person similar to when signed in a establishment when buying with the credit card, but with the advantage of doing it from the terminal itself, therefore adding a greater security to transactions via Internet from terminals mobile phones

In addition to this, the industry could take advantage of the characteristics that this technology has for the firm electronic documents, for non-repudiation of transactions or for any other application that needs the rubric of a action.

       \ newpage
    

Likewise, as explained above, This technique can be used with cryptographic applications, being able, from the pattern of the gesture made, to be generated or release a cryptographic key that encrypts a document and Only the user, when repeating the gesture, can decipher it.

Finally in figure 6 a scheme is shown Basic biometric recognition device composed of a accelerometer (1) and processing means (2).

Once the invention is clearly described, again it is stated that the particular achievements described above are subject to modifications of detail as long as they do not alter the fundamental principle and the essence of the invention.

Claims (14)

1. Biometric recognition method based on the movement of the hand in space, characterized in that it comprises: ? one enrollment stage in turn comprising:

to)

obtain, from at least one accelerometer (1) integrated in a biometric recognition device, an acceleration signal ( G_ {1x}, G_ {1y}, G_ {lz} ) from the realization of a manual gesture carrying in the hand said device;

b)

repeat step a) n times;

C)

sample the acceleration signals previously obtained ( G_ {1x}, G_ {1y}, G_ {1z}; G_ {2x}, G_ {2y}, G_ {2z}; ... );

d)

store sequences sampled;

and)

process each pair of signals sampled on each of the axes of space, comprising:

\ sqbullet

adapt the signals sampled smoothly to get two aligned sequences, one for each gesture performed on a given axis of the space;

\ sqbullet

interpolate the errors of sequences aligned with the samples of the signals obtained in a) and in b) to correct the variations of speed in the repetition of the gesture;

\ sqbullet

quantify the differences between the sequences adapted and interpolated by each pair of gestures made on a certain axis of space;

F)

get and store a gesture pattern for user authentication;

an access stage which in turn understands:

g)

obtain, from at least one accelerometer (1) integrated in a recognition device biometric, an access signal from the realization of a manual gesture carrying in hand said device;

h)

sample the signals from access;

i)

store the access sequence sampled;

j)

quantify the differences between the sampled access sequence and gestural pattern;

k)

calculate the average deviation from of the deviations obtained from each of the access signals with the gestural pattern;

l)

decide access by comparing the mean deviation with the gestural pattern.

2. Biometric recognition method based on the movement of the hand in space, according to the preceding claim, characterized in that obtaining the acceleration and access signals comprises the decomposition of the acceleration of the manual gesture in each of the three axes from space. 3. Biometric recognition method based on the movement of the hand in space, according to any of the preceding claims, characterized in that the adaptation is carried out by means of a global sequence alignment method. 4. Biometric recognition method based on the movement of the hand in space, according to any of the preceding claims, characterized in that the gestural pattern comprises the signals obtained in a) and b) and the average distance between the quantified differences between each pair of gestures, where said average distance measures the similarity between the repetitions of the gesture. 5. Biometric recognition method based on the movement of the hand in space, according to any of the preceding claims, characterized in that the quantification of differences in the access stage is performed between the sampled access sequence and the signals stored as part of the gesture pattern. 6. Biometric recognition method based on the movement of the hand in space, according to any of the preceding claims, characterized in that the calculation of the average deviation is performed with each of the stored signals as part of the gestural pattern. 7. Biometric recognition method based on the movement of the hand in space, according to any of the preceding claims, characterized in that the access decision is made by comparing the average deviation with the average distance stored as part of the gestural pattern. 8. Biometric recognition device based on the movement of the hand in space, characterized in that it comprises: at least one accelerometer (1) configured to obtain acceleration signals; processing means (2) configured for the realization of a enrollment stage and a access stage where:

\ circ

the enrollment stage includes:

m)

obtain, from at least one accelerometer (1) integrated in a biometric recognition device, an acceleration signal ( G_ {lx}, G_ {ly}, G_ {lz} ) from the realization of a manual gesture carrying in the hand said device;

n)

repeat the previous step n times;

or)

sample acceleration signals ( G_ {lx}, G_ {ly}, G_ {lz}; G_ {2x}, G_ {2y}, G_ {2z}; ... );

p)

store sequences sampled;

q)

process each pair of signals sampled on each of the axes of space, comprising:

\ sqbullet

adapt the signals sampled smoothly to get two aligned sequences, one for each gesture performed on a given axis of the space;

\ sqbullet

interpolate the errors of sequences aligned with the samples of the signals obtained in m) and in n) to correct the variations of speed in the repetition of the gesture;

\ sqbullet

quantify the differences between the sequences adapted and interpolated by each pair of gestures made on a certain axis of space;

r)

get and store a gesture pattern for user authentication;

         \ vskip1.000000 \ baselineskip
      

\ circ

the access stage understands:

s)

obtain, from at least one accelerometer (1) integrated in a recognition device biometric, an access signal from the realization of a manual gesture carrying in hand said device;

t)

sample the signals from access;

or)

store the access sequence sampled;

v)

quantify the differences between the sampled access sequence and gestural pattern;

w)

calculate the average deviation from of the deviations obtained from each of the access signals with the gestural pattern;

x)

decide access by comparing the mean deviation with the gestural pattern.

         \ vskip1.000000 \ baselineskip
      

9. Biometric recognition device based on the movement of the hand in space, according to claim 8, characterized in that the processing means (2) are configured to obtain signals by decomposing the acceleration of the gesture in each of the The three axes of space. 10. Biometric recognition device based on the movement of the hand in space, according to any of claims 8-9, characterized in that the processing means (2) are configured for adaptation by means of a global sequence alignment method. 11. Biometric recognition device based on the movement of the hand in space, according to any of claims 8-10, characterized in that the processing means (2) are configured to obtain the gestural pattern from the signals obtained in m) and n) and the average distance between the quantified differences between each pair of gestures, where said average distance measures the similarity between the repetitions of the gesture. 12. Biometric recognition device based on the movement of the hand in space, according to any of claims 8-11, characterized in that the processing means (2) are configured to quantify differences in the access stage between the sequence of sampled access and stored signals as part of the gestural pattern. 13. Biometric recognition device based on the movement of the hand in space, according to any of claims 8-12, characterized in that the processing means (2) are configured to perform the calculation of the average deviation with each of the signals stored as part of the gestural pattern. 14. Biometric recognition device based on the movement of the hand in space, according to any of claims 8-13, characterized in that the processing means (2) are configured to make the access decision comparing the average deviation with the distance average stored as part of the gestural pattern.