ES2361686A1 - Device and stress detection method through physiological signs. (Machine-translation by Google Translate, not legally binding) - Google Patents

Abstract

Device and method of stress detection by physiological signals comprising an enrollment stage and an access stage. The enrollment includes: obtaining, for different mood states of the user, hr (heart rate) and gsr (galvanic skin response) signals; store the signals; execute a clustering algorithm to obtain a stress pattern associated with the user and store said stress pattern for later identification. The access stage comprises: performing identification tasks of the user; obtain the stress pattern corresponding to the previously identified user; obtain the hr and gsr signals of the current mood of the user; store the obtained access signals; execute a clustering algorithm to obtain an access stress pattern; compare the access stress pattern with the stored stress pattern for the user and decide the degree of user stress. (Machine-translation by Google Translate, not legally binding)

Description

Device and method of stress detection by physiological signals.

Field of the Invention

The present invention belongs to the field of security, to public safety, as in access control to buildings and services, to computer security, in communications and electronic commerce and also the security related to the sector automobile, aeronautical and domestic.

State of the art

Numerous related works are known with the detection of stress based on physiological signals. Some present a stress detection study based solely on existing temperature on the finger (Finger Temperature, ST), along with fuzzy logic algorithms (Fuzzy Logic) and Case-Based Reasoning (CBR). Others present a system capable of monitoring the degree of stress of a person to through the way the individual writes on the keyboard. Bliss Biometric technique is known as Keystroke dynamics.

Heart Rate (HRV) It has also been considered as an indicator of stress in being human and has been extensively studied and analyzed. Between the most important work a monitoring system is presented of stress on a girreles architecture distributed in sensors smart. The heart rate is recorded in different individual body locations using localized sensors in the user's clothes.

However, it is not usual to study a certain physiological aspect but a group of them to get a much more complete information about the mood. On this is work that studies a wide variety of signals and parameters, as well as the combination of these.

Following the line of the works presented previously, in addition to the skin temperature, they are known jobs where other signs such as resistance are considered Galvanic skin (GSR, Galvanic Skin Response) and pressure blood pressure and heart rate (BVP, Blood Volume Pulse). All these signals are characterized by being acquired non-invasively, not obstrusive and because its variation against stress is predictable.

One of the aspects associated with the signals Physiological is also pupillary dilation (PD, Pupil Dilation) and eye movement There are articles that not only consider the above signals, but also take into account previously presented signals, i.e., GSR, BVP, ST, etc ... With These signals not only manage to detect stress but also the user attention capacity (measured through movement same eye). In addition, the possible inference of the intention of the individual, a very interesting result for future applications with computers and for improved interaction machine man (HCI, Human Computer Interaction).

Notably, a study of all the previous physiological signals along with other very related to stress, but certainly invasive and difficult to acquire, as is the case of the Tomography by emission of positrons (PET, Positron Emission Tomography), Magnetic Resonance Functional (FMRI, Functional Magnetic Resonance Imaging), Electroencephalography (EEG, Electroencephalography), as well as others less invasive signals such as electromyograms (EMG) or the breathing rate In this case, fuzzy logic is used to elucidate the degree of stress of a user and also presents a very oriented approach to the improvement of man interaction machine.

In KR705984-B1 patent use the previous physiological signals to improve the concentration and the meditative abilities of a given user by controlling the stress state of it.

Due to the growing interest in security, a method and a stress detection device was desirable as that described in the present invention, based on the use of different biometric techniques on signals physiological in order to protect access systems and Provide user security. In fact, no system of identification is able to detect if the person presenting your Biometric feature (iris, fingerprint, face ...) is forced to do it, that is, if the user registered in the system is being used as a key by an illicit individual. By the method and the proposed stress detection device, can be detected this situation, since the access system of the ability to extract information from the user who accesses to decide if said individual is under a highly highly stressful

Description of the invention

The present invention proposes a method and a non-invasive stress detection device that can be integrated into Any access system with biometric identification.

In stress detection studies, mentioned in the prior art, based solely on the existing temperature on the finger (Finger Temperature, ST), it proposes a decision system based on fuzzy logic to decide The status of a person. However, the algorithm used in the proposed method, not only uses fuzzy logic for your system decision but it is also able to offer a range (between 0 and 100), in short, a percentage of how stressed you are the person (0 nothing, 100 a lot). With which, the functions performed according to known studies.

Several of the studies, also mentioned in the state of the art, are based on detecting stress through several signals (BVP, GSR, PD, etc ...) validating the sample with different populations and using the same Stroop test to induce the stress. However, the technique developed by the method proposed, requires only two signals (GSR and Heart Rate) to perform stress detection, unlike the compendium of signals that are needed in the mentioned studies. This allows a less invasive capture, and the placement of fewer sensors around the individual's body, facilitating the integration of the sensors in a future device. In addition, the population used is much wider (80 individuals versus 32 or 6 individuals), as well as tests to induce stress are more complete, seen from the cognitive point of view as well as from The physiological point of view. Finally, the concept of pattern of stress, original of the system presented, is not named in None of these studies.

Other works carried out present a study very deep stress, however, they analyze so many signs that violate the condition of non-invasiveness, deforms that any system is less invasive than the one used in such works. Hence, in the proposed system, not only the simplicity of the captured signals, but the algorithmic used to analyze the Signs outperform known works in simplicity and efficiency. In addition, the stress pattern proposed in the present system results Novel compared to previous studies.

In addition to the above, there is a difference fundamental between the state of the art and the device and method proposed. Both the proposed device and method are complete and not simply algorithms, whose objective is to minimize the size, consumption and cost, in addition to allowing the use of this technology in applications, both centralized, and Distributed

Three features of the proposed device and method not appreciated in works previous:

\ sqbullet

Stress Pattern : In the previous works consulted, there is no talk of a pattern of stress, information that may reflect the behavior of a particular individual when faced with anomalous or stressful situations.

\ sqbullet

Algorithmia : The output offered by the detection and decision algorithms provide a percentage expressing how stressed a certain user is (0 nothing, 100 a lot), unlike other jobs where you can only talk about whether or not there is stress, but Not to what extent.

\ sqbullet

Validation of the experiments performed : Due to their completeness, the experiments performed to detect stress have been validated in a larger database, unlike those used in most literatures (80 individuals). In addition to stressful tests, they offer a greater degree of safety when it comes to inducing only stress and not other spurious sensations such as anxiety or fear, which can be considered as noise when you want to detect only stress.

       \ vskip1.000000 \ baselineskip
    

Therefore, and according to what is presented in the state of the art, there are some devices and systems able to detect stress, however, none of them offers the management of the information provided by said stress. In others words, literature stagnates in mere stress detection, without carrying out any decision once the system has been able to detect it.

In addition, the stress pattern is novel since that only talk about stress detection in the sense of indicate whether or not there is stress, however, this system is capable of provide a degree of stress, indicating with a percentage the rate of stress.

The proposed system is capable not only of detect stress, but make decisions based on that stress, performing certain actions depending on the mood of the individual.

In addition, stress detection systems previously implemented in the state of the art, they study and explore many properties to achieve this end, without However, in this method, only two signals, the heart rate, HR (Heart Rate), and skin conductance, SC (Skin Conductance) or GSR (Galvanic Skin Resistance), will be the ones that offer the information enough to know if the user is stressed. That implies, that unlike the works presented previously, the device responsible for stress detection, it is easily integrable and can perform its function non-invasively, quickly and efficient.

The method of stress detection by physiological signals, includes:

?

a enrollment stage that to in turn includes:

to)

get, for different states mood of the user and from some sensor means, some HR (Heart Rate) and GSR (Galvanic Skin) signals Response);

b)

store the signals obtained;

       \ global \ parskip0.950000 \ baselineskip
    

C)

run a clustering algorithm that allocates a cluster for each mood of the user and that obtains, for each cluster, from the signals included in the cluster, a centroid, the standard deviation of the HR signals and the standard deviation of the GSR signals forming a pattern of stress associated with the user;

d)

store the stress pattern in ones storage media for identification of Username;

       \ vskip1.000000 \ baselineskip
    

?

an access stage that in turn understands:

and)

perform identification tasks of Username;

F)

obtain, from the identification made by the user, the stress pattern stored in the storage media corresponding to the user previously identified;

g)

obtain, from the media sensors, HR (Heart Rate) and GSR (Galvanic Skin) signals Response) corresponding to the current mood of the user who intends to access the system;

h)

store access signals obtained in g);

i)

run a clustering algorithm that allocates a cluster for each mood of the user and that obtains, for each cluster, from the signals included in the cluster, a centroid, the standard deviation of the HR signals and the standard deviation of the GSR signals forming a pattern of access stress associated with the identified user;

j)

compare access stress pattern with the stress pattern stored for the user;

k)

determine the degree of stress of user, comprising:

one)

calculate, for each signal x obtained in g) and for each mood of the user, the following function: one

2)

determine the degree of stress of the user as the mood of the user who fulfills the following function: 2 where N is the number of moods of the user.

       \ vskip1.000000 \ baselineskip
    

In a preferred embodiment, the step of access includes updating the stored stress pattern for the user in the storage media if the pattern of Access obtained for the identified user does not belong to any previously stored stress pattern.

In the same way, in one embodiment Preferably, the clustering algorithm used is the k-means.

Preferably, the comparison of the pattern of stress obtained with the stored stress pattern is performed by fuzzy logic. Performing in this way, the comparison of patterns comprises comparing the centroids of the pattern of access and centroids stored in the means of storage.

The stress detection device by physiological signals, includes:

?

sensor means configured to obtain HR (Heart Rate) and GSR signals (Galvanic Skin Response);

?

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

\hskip1cm

\circ la etapa de enrolamiento comprende:

 \ hskip1cm 

circ the enrollment stage comprises:

l)

get, for different states moods of the user and from the sensing means, the signals HR (Heart Rate) and GSR (Galvanic Skin Response);

m)

store the signals obtained;

n)

run a clustering algorithm that allocates a cluster for each mood of the user and that obtains, for each cluster, from the signals included in the cluster, a centroid, the standard deviation of the HR signals and the standard deviation of the GSR signals forming a pattern of stress associated with the user;

       \ global \ parskip1.000000 \ baselineskip
    

or)

store the stress pattern in ones storage media for identification of Username;

\hskip1cm

\circ una etapa de acceso que a su vez comprende:

 \ hskip1cm 

\ circ an access stage which in turn comprises:

p)

perform identification tasks of Username;

q)

obtain, from the identification made by the user, the stress pattern stored in the storage media corresponding to the user previously identified;

r)

obtain, from the media sensors, HR (Heart Rate) and GSR (Galvanic Skin) signals Response) corresponding to the current mood of the user who intends to access the system;

s)

store access signals obtained in r);

t)

run a clustering algorithm that allocates a cluster for each mood of the user and that obtains, for each cluster, from the signals included in the cluster, a centroid, the standard deviation of the HR signals and the standard deviation of the GSR signals forming a pattern of access stress associated with the identified user;

or)

compare access stress pattern with the stress pattern stored for the user;

v)

determine the degree of stress of user, comprising:

one)

calculate, for each signal x obtained in g) and for each mood of the user, the following function: 3

2)

determine the degree of stress of the user as the mood of the user who fulfills the following function: 4 where N is the number of moods of the user.

       \ vskip1.000000 \ baselineskip
    

In a preferred embodiment, the means of processing are configured to update the stress pattern stored for the user on the storage media if the access pattern obtained for the identified user does not belong to any previously stored stress pattern.

In the same way, in one embodiment preferably, the processing means are configured to run the k-means algorithm as an algorithm of clustering

Preferably, the processing means are configured to compare the stress pattern obtained with the Stress pattern stored by diffuse logic. So, the media processing are configured to compare patterns using the comparison of the centroids of the access pattern and the centroids stored in storage media.

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 represents the space formed by the spaces H (R) and G (R) , from the capture of the HR and GSR signals.

Figure 2 represents the result of the k- means algorithm with a centroid associated with each of the distributions obtained in Figure 1.

Figure 3 shows a vision in three dimensions of the stress pattern.

Figure 4 represents an overview of the diffusion decision system where the distributions in the different spaces H (R) and G (R) are appreciated. The vertical line indicates an average stress value.

Figure 5 shows the enrollment scheme of a user, their stress pattern is created and stored for a back access

Figure 6 shows the access scheme to system of a registered user, their signals are captured physiological, they are contrasted with the pattern stored in the base of data and, if the information matches some of the states stored for the user, their degree of stress is determined, but, Your pattern is updated.

Detailed description of one embodiment

The object of the invention is to present a device and a method of stress detection by signals biological to detect the amount of stress found in a user in the enrollment or access stage before a system with authentication requirement

Figure 1 represents the space formed by the spaces H (R) and G (R) . In this way, when trying to represent the capture of the HR and GSR signals, a cloud of points will be obtained, from which the parameters could be obtained later (\ zeta, \ sigma_ {h}, \ sigma_ {g} ) captured, where \ zeta corresponds to the centroid of the point cloud, \ sigma_ {h} with the standard deviation of the signals of H (R) and \ sigma_ {g} with the standard deviation of the signals of G ( R) . The figure shows the different moods with different symbols: diamond (base state), circle and square (altered state) and asterisk (post-stress state).

Figure 2 represents the result of the k- means algorithm, where for each of the distributions obtained in Figure 1 there is an associated centroid. The symbols remain the same as in Figure 1, that is: diamond (base state), circle and square (altered state) and asterisks (post-stress state).

Figure 3 shows a three-dimensional view of the stress pattern, where the different centroids \ zeta_ {i} can be seen, and the different distributions according to the deviations \ sigma_ {h_ {i}} and \ sigma_ {g_ {i }}, for i = 1, 2, ..., k with k equal to the number of different moods. Three states can be clearly seen: relaxed (base), altered and post-stress.

Figure 4 represents a global vision of the diffuse decision system where the distributions in the different spaces H (R) and G (R) can be seen in the first and second column, where, depending on the input value, a degree of stress, shown in the last column, depending on the output of diffuse stress ε.

The first row corresponds to the state base, the second and third with the altered state and the fourth with the post-stress state

The figure shows, in its first column, Gaussian distributions for different moods. These distributions have a mean equal to the different centroids \ zeta_ {i}, and dispersion \ sigma_ {h_ {i}} (for the first column of subfigures) and \ sigma_ {g} (for the second column of subfigures), for i = 1, 2, ..., k with k equal to the number of different mood states. Therefore, the first column of subfigures represents the distributions for the HR signals in the different mood states, and the second column of subfigures represents the distributions for the GSR signals, also in the different mood states. The fifth row, in both HR and GR columns, represents the case in which for any heart rate (HR), with a high GSR, the result of stress will be equally high.

In the last column of the figure, it represent the distributions of stress, for each of the five cases analyzed as well as the result of total stress. You choose a triangular representation for its simplicity, representing low values for low stress mood situations, and values High for situations with high stress. In the fourth row, the corresponding to the post-stress situation, it corresponds to a stress outflow with higher values than values obtained under stress situations, because in a post-stress situation, the user is still stressed even though the stressful agent has disappeared. Besides, the second and third row have two different stress outputs, simply because of the fact that later on at the start final (last subfigure of the third column) the situations stressors have more weight to the final solution than the non stressful

Partial exits of stress (the five first subfigures of the last column of figure 4) are obtained simply by the minimum of the values reached previously, in each of the rows. Those values are correspond to the vertical red lines, and are the values read by a sensor connected to the system where the distributions They contrast with the iris pattern.

With this you get the contributions of each one of the states to the partial exits of stress corresponding, and they are added finally obtaining the centroid of the final figure, indicating the value of stress obtained.

The vertical line indicates the value indicated by ε, which in this case indicates an average stress value.

Figure 5 shows the enrollment scheme of a user, when their stress pattern is created, \ wp, being this one stored for later access.

Figure 6 shows an overview of the system access scheme, in which the user already registered presents again its physiological signals, which will be contrasted with the pattern \ wp stored in the database.

The method developed for the detection of Stress, requires the following stages:

1. Capture of physiological signals

To perform the signal capture physiological, HR (Heart Rate) and GSR (Galvanic Skin Resistance), are use any commercial device capable of capturing and processing The above signals. The environment where the capture takes place of the samples does not require any specific specification, since aspects such as lighting, humidity and temperature do not affect the process, as long as they are within the margins indicated by the devices that perform the sampling.

The capture stage is performed in the phases from:

\ sqbullet

System training : To distinguish between a normal situation (referred to as the base state from now on) and a stress situation (altered state), the system must know how a certain user (such as their stress patterns) behaves in both situations . For this, different tests will be carried out with a certain cognitive load (simple mathematical operations, distinction between sounds with different height, etc ...), taking measurements at both times: before and after the tests.

\ sqbullet

Access to the system : Once the system has been able to store how the user behaves in both situations (storage that is achieved by recording the HR and GSR signals), the identification of the degree of stress that the individual possesses begins.

       \ vskip1.000000 \ baselineskip
    

2. System training and pattern extraction stress

To train the system, it is necessary first of all make the system associate the behavior that is going to be registered, that is, the HR and GSR signals in both states, to a certain user, because each person has a stress pattern different, and therefore user identification is done essential.

When a user shows up for the first time before the system, proceed to place on the fingers of the hand the relevant devices to obtain the HR and GSR signals. During enrollment, tasks with a certain load will be performed cognitive Specifically the Stroop test along with mere mathematical operations can induce a state of stress sufficiently differentiated from the base state. HR and GSR signals they are registered at all times, being stored as they are processed by commercial devices, that is, as a signal one-dimensional length dependent on the time the enrollment process

Once these signals have been left registered, the stress pattern is extracted, which It will be used later to determine if this same user presents An anomalous state.

In general, h \ in H (R) and g \ in G (R) respectively represent the HR and GSR signals, which can take real values, R , and whose range of values will be determined by the devices used. Cartesian space is defined
H (R) \ times G (R) , which has the information provided by the above signals in the plane, and which will allow for a simpler way to group the information provided by both variables.

On said space, supervised clustering techniques are performed, using the k- means algorithm.

This algorithm is responsible for analyzing the data obtained and associated with each of the states and get a centroid Said centroid, represented by \ zeta is associated with each of the states, which for each of the states There is a different \ zeta.

In addition to this, the algorithm provides a standard deviation for each of the states, whereby there are also two deviations associated with each of the spaces H (R) and G (R) . These values are represented by \ sigma_ {h} and \ sigma_ {g}, which are the typical deviations associated with H (R) and G (R) respectively. After executing the k- means algorithm, the stress pattern of the person who has enrolled in the system is finally obtained, which is simply the terns (\ zeta_ {i}, \ sigma_ {h_ {i}}, \ sigma_ { g_ {i}}), \ forall i \ in N , where there may be as many states as desired, although the minimum is that there are two: one relaxed state (base state) and one stressed state (altered state).

Finally, said pattern, called \ wp, must be updated as the user accesses the system, so that it adapts to the mood of the user and be able to detect stress peaks more accurately, apart from detecting the high level of stress.

The fact of offering a pattern of stress, something with which to contrast, it is something completely innovative when compared with what is implemented in the literature.

       \ vskip1.000000 \ baselineskip
    

3. Access to the system

Finally, and based on the pattern \ wp, the user accesses the system and again presents the HR and GSR biometric features. These traits carry out the same process above, that is, they are registered and treated by means of a k- means algorithm to obtain the list (\ zeta, \ sigma_ {h}, \ sigma_ {g}). Said tern is contrasted with the pattern \ wp by diffuse operators, in such a way that an output is obtained, also diffuse, called \ varepsilon and that offers a value in the interval [0,1], indicating the degree of stress of the person. This function? Compares the centroids obtained by \ wp with the recently acquired input centroid, and based on the distributions to the fuzzy rules defined, it offers as an output what state corresponds to the greatest degree and to what extent it resembles said state.

The diffuse output ε is obtained after calculating, for each signal x obtained from the user and for each mood of the same, the following function: 5 This function offers an output in the interval [0,1]. To determine the degree of stress of the user, it is calculated 6 for N equal to the number of moods of the user.

In this way, with the data presented two things can happen: first, that the value presented to the entry belongs to one of the states defined above, this is, or relaxed or stressed, base or altered, respectively. Yes this is the case, the system would act as explained to continuation in order to define the degree of stress Find the user. However, if the value presented is not considered as a value recorded by the system, then said system starts updating the stored pattern taking into account The new value arrived. If for a casual that value is never again repeated, it will have been an anomalous situation of the individual, but if on the contrary said value is repeated over time, it will imply that said state is the true base state of the individual, and surely its distribution of values is closer of the distribution of values of the relaxed base state that of altered state

Whereupon, the system first learns and train the system with data extracted from the user and then, and as the user accesses the system more and more, the system will be able to learn the behavior of the individual that access and adapt to the best, so that the more time passes, the deeper the learning about the person and It will be much easier to detect stress.

Then, without pretending to be limiting in the scope of the invention, an embodiment of the proposed method with which optimal results have been obtained.

First, it is necessary to enroll the user Who wants to belong to the system. To do this, you need to create your stress pattern, which is about the behavior of HR signals and GSR in different situations of base state (relaxation) and altered state (stress). This requires a device integrated that is capable of measuring HR and GSR. Said device may be any existing commercial device that offers the possibility of storing and processing these signals. Specifically, the psychophysiological record of heart rate and skin conductance was carried out by using the I-330-C2 PHYSIOLAB (J&J Engineering) with capacity for 6 channels (including signals from EMG, ECG, breathing, heart rate, conductance and resistance of the skin).

Specifically, this device allows high integrability in a future stress detection system, since The reading is done by placing sensors (small metal plates of about 1cm radius) on the fingers of the right hand, although it can be used in any hand.

To carry out a stress detection is it is necessary to know how a certain user behaves before a normal state (base state) and a state of stress (altered state). For this it is necessary to induce the state of relaxation and stress. The tests performed to implement this system consisted of four stages:

\ sqbullet

First stage : Induce a state of relaxation where the user is measured by their HR and GSR responses. Said state of relaxation consists simply in sitting the individual, with the sensors connected, and recording the previous signals.

\ sqbullet

Second stage : The user is asked to breathe quickly and deeply, performing hyperventilation until the sensation is annoying. After that precise moment in which the user indicates his discomfort, the samples are registered.

\ sqbullet

Third stage : After the second stage, and once the user has regained normal breathing, he is invited to conduct a talk before a camera that will record his performance. For two minutes the user is asked to think about a given topic. However, at the time of recording, your mood is recorded again using the HR and GSR signals, later telling you that it is not necessary to perform the test.

\ sqbullet

Fourth stage : Finally, the user is asked to relax as the tests have ended, re-registering here the previous biometric characteristics.

       \ vskip1.000000 \ baselineskip
    

These stages perfectly distinguish three states: state of relaxation (first stage), altered state (second and third stage), post-altered state (fourth stage). The tests performed are specific to this method; However, any test with considerable cognitive ability (such as mathematical operations, color distinction, Stroop test, etc ...) can be used to induce stress in the same way as previous tests.
beef.

       \ newpage
    

With these signals, the Cartesian space H (R) \ times G (R) is constructed . With the previous data, the Ar-means algorithm is executed, providing as a result (and for each state) the centroid of said distribution, as well as the deviation in both spaces ( H (R) and G (R) ). These centroids can be obtained with any clustering algorithm.

In short, it seeks to divide a certain data set <( h 1, g 1), ..., ( h n, g n)> \ in H (R) G G (R) in k partitions ( k < n ), called S = { S 1, ..., S k} \ subsete H (R) \ times G (R) in such a way that the sum of intracluster squares is minimized, or in other words:

7

where x _ {j} represents each of the signals obtained and \ {j} mu distance to the centroid selected in each iteration.

       \ vskip1.000000 \ baselineskip
    

Since there are four states, and it is known exactly what signals belong to each state, this information to make a first partition of the whole set of data in four partitions, \ Pi = {\ pi_ {1}, \ pi_ {2}, \ pi_ {3}, \ pi_ {4}}.

For each partition \ pi_ {1}, \ forall i \ in {1, 2, 3, 4}, this algorithm is applied, with k = 1, since it is known in advance that all points of said partition belong to the same mood. With all this, the centroids of the partitions in \ Pi, called Z = {\ zeta_ {1}, \ zeta_ {2}, \ zeta_ {3}, \ zeta_ {4}} are obtained. In addition to this, for each partition \ pi_ {i} the deviation of the data in the space H (R) \ times G (R) is obtained, that is, \ sigma_ {h} \ times \ sigma_ {g} = < (\ sigma_ {h_ {1}}, \ sigma_ {g_ {1}}), ..., (\ sigma_ {h_ {4}}, \ sigma_ {g_ {4}})>.

At this point, the stress pattern is defined as the set of triples (\ zeta_ {i}, \ sigma_ {h_ {i}}, \ sigma_ {g_ {i}}), \ forall i \ in {1, 2, 3, 4}, and will be represented by the letter \ wp.

Finally, and concluding this stage of User enrollment, pattern storage is required. To store the pattern, it is necessary to reserve in certain memory space to save the pattern \ wp and a user identifier, For example your name.

The \ wp pattern occupies 12x32 bits, since each pattern element (either \ zeta_ {i}, \ sigma_ {h_ {i}} or \ sigma_ {g_ {i}}), is represented by a float element, and set that there are four states, and three elements per state, in total \ wp occupies 384 bits, 48 bytes in memory.

Once the user has been enrolled and assigned a stress pattern \ wp, the system is accessed. To access the system, the user must present the fingers of his hand for the placement of the previous sensors. For a period of 10-20 seconds, the recording of the HR and GSR signals, h and g respectively, is performed.

Once the sample has been obtained, the patterns are compared by using fuzzy logic. For them, the previous partitions \ Pi are modeled, as fuzzy variables \ tilde {\ Pi} = {\ tilde {\ pi} _ {1}, \ tilde {\ pi} _ {2}, \ tilde {\ pi} 3, \ tilde {\ pi} 4. These variables depend on the triplets (\ zeta_ {i}, \ sigma_ {h_ {i}}, \ sigma_ {g_ {i}}), \ forall i \ in {1, 2, 3, 4}, associated to the pattern \ wp as follows:

8

where x are the signals obtained from the user.

       \ vskip1.000000 \ baselineskip
    

In order not to overload the nomenclature, it will be assumed the following as of now:

9

In addition, the fuzzy functions \ tilde {\ pi} _ {i} provide a value in the range [0,1], indicating the degree of belonging of the signals ( h, g ) to the partition \ tilde {\ pi} _ {i}. To aggregate all the information coming from the four states, the fuzzy output is defined that will also provide in the interval [0,1], the degree of stress in the user. This variable is based on rules defined based on the fuzzy functions \ tilde {\ pi} _ {i} and is expressed as: 10 100 where N is the number of moods of the user.

       \ newpage
    

This decision system based on fuzzy logic therefore provides a certain degree of stress in the user, based on the stress pattern \ wp that the individual.

Since each person has a certain stress pattern that is variable over time, this pattern of stress should be refreshed as the user accesses To the system. To perform this update, a new model is modeled \ wp 'stress pattern that cools following the following structure:

eleven

where? will be approaching one as the user accesses the system, through the rule next:

12

where n is the number of times the user has accessed the system.

       \ vskip1.000000 \ baselineskip
    

The idea is that the algorithm is capable of adapt to the different stages that human beings suffer throughout of your life, and therefore the system does not crash despite that the individual is going through a stressful stage in his life, nothing to do with the situation that someone is forcing him to Access a certain system.

With all this, and unlike literature existing, the system can be used as a live subject detector in biometric applications and devices, and applications derived from its use extend to a large number of fields, from home automation to computer security applications more demanding.

The improvements presented in this system are based not only in the detection of stress (because stress is detected only with two signals), but it presents a complex system and intelligent decision behind said detection system stress able to act accordingly in case an individual is stressed abnormally, therefore being a very complete system for fraud detection, detection of Frustration or even behavioral detection.

In this way, the industrial application of this Technology presents several fronts. First, any access technology (either control of a building, or access to a protected information) you can take advantage of this technology, since which allows to detect if a person is faced with a situation stressful and therefore at risk of fraud for the system.

The ease of integration of these devices in mobile devices will allow the inclusion of this method in any access system, and in any system of control, such as the cockpit of an airplane pilot, car, truck, etc ...

Its industrial application also contemplates the use of these devices for any type of application electronic, such as electronic transfer, voting electronic or access to compromised computer systems. Also, industrial applications related to home automation are derive from the fact of being able to equip a certain system (in the case of home automation in general) of intelligence enough to understand and understand the mood of the user, and its follow up. This would allow in the last stay to analyze continuously your mood and bring some kind of control doctor so you can control it.

Finally, at the business level it has multiple applications, because the productivity of an industry / company rests in the productivity of each of its workers, and said individual productivity is diminished when individuals suffer a high degree of stress Therefore, a detection system of stress will allow the stress control of each of the employees / workers getting maximum performance from them, minimizing your stress levels.

Once the invention has been clearly described, notes that the particular realizations above described are subject to modifications of detail provided that do not alter the fundamental principle and the essence of the invention.

Claims (10)

1. Method of stress detection by physiological signals, characterized in that it comprises:

?

a enrollment stage that to in turn includes:

to)

get, for different states mood of the user and from sensor means (1), HR (Heart Rate) and GSR (Galvanic Skin) signals Response);

b)

store the signals obtained;

C)

run a clustering algorithm that allocates a cluster for each mood of the user and that obtains, for each cluster, from the signals included in the cluster, a centroid, the standard deviation of the HR signals and the standard deviation of the GSR signals forming a pattern of stress associated with the user;

d)

store the stress pattern in one§ storage means (3) for the identification of the Username;

?

an access stage that in turn understands:

and)

perform identification tasks of Username;

F)

obtain, from the identification made by the user, the stress pattern stored in the storage media (3) corresponding to the user previously identified;

g)

obtain, from the media sensors (1), HR (Heart Rate) and GSR (Galvanic) signals Skin Response) corresponding to the user's current mood which intends to access the system;

h)

store access signals obtained in g);

i)

run a clustering algorithm that allocates a cluster for each mood of the user and that obtains, for each cluster, from the signals included in the cluster, a centroid, the standard deviation of the HR signals and the standard deviation of the GSR signals forming a pattern of access stress associated with the identified user;

j)

compare access stress pattern with the stress pattern stored for the user;

k)

determine the degree of stress of user, comprising:

one)

calculate, for each signal x obtained in g) and for each mood of the user, the following function: 13

2)

determine the degree of stress of the user as the mood of the user who fulfills the following function: 14 User moods.

         \ vskip1.000000 \ baselineskip
      

2. Method of stress detection by physiological signals, according to the preceding claim, characterized in that the access stage comprises updating the stress pattern stored for the user in the storage means (3) if the access pattern obtained for the user identified does not belong to any previously stored stress pattern. 3. Method of stress detection by physiological signals, according to any of the preceding claims, characterized in that the clustering algorithm used is k-means. 4. Method of stress detection by physiological signals, according to any of the preceding claims, characterized in that the comparison of the stress pattern obtained with the stored stress pattern is performed by diffuse logic. 5. Method of stress detection by physiological signals, according to claim 4, characterized in that the comparison of patterns comprises the comparison of the centroids of the access pattern and the centroids stored in the storage means (3). 6. Stress detection device by physiological signals, characterized in that it comprises:

?

sensor means (1) configured to obtain HR (Heart Rate) and GSR signals (Galvanic Skin Response);

?

processing means (2) configured for the realization of a enrollment stage and a access stage where:

 \ hskip1cm 

circ the enrollment stage comprises:

l)

get, for different states mood of the user and from the sensor means (1), the HR (Heart Rate) and GSR (Galvanic Skin) signals Response);

m)

store the signals obtained;

n)

run a clustering algorithm that allocates a cluster for each mood of the user and that obtains, for each cluster, from the signals included in the cluster, a centroid, the standard deviation of the HR signals and the standard deviation of the GSR signals forming a pattern of stress associated with the user;

or)

store the stress pattern in ones storage means (3) for the identification of the Username;

 \ hskip1cm 

\ circ the access stage which in turn comprises:

p)

perform identification tasks of Username;

q)

obtain, from the identification made by the user, the stress pattern stored in the storage media (3) corresponding to the user previously identified;

r)

obtain, from the media sensors (1), HR (Heart Rate) and GSR (Galvanic) signals Skin Response) corresponding to the user's current mood which intends to access the system;

s)

store access signals obtained in r);

t)

run a clustering algorithm that allocates a cluster for each mood of the user and that obtains, for each cluster, from the signals included in the cluster, a centroid, the standard deviation of the HR signals and the standard deviation of the GSR signals forming a pattern of access stress associated with the identified user;

or)

compare access stress pattern with the stress pattern stored for the user;

v)

determine the degree of stress of user comprising:

one)

calculate, for each signal x obtained in r) and for each mood of the user, the following function: fifteen

2)

determine the degree of stress of the user as the mood of the user who fulfills the following function: 16 where N is the number of moods of the user.

         \ vskip1.000000 \ baselineskip
      

7. Stress detection device by physiological signals according to claim 6, characterized in that the processing means (2) are configured to update the stress pattern stored for the user in the storage means (3) if the access pattern obtained for the identified user does not belong to any previously stored stress pattern. 8. Stress detection device by physiological signals, according to any of claims 6-7, characterized in that the processing means (2) are configured to execute the k-means algorithm as a clustering algorithm. 9. Stress detection device by physiological signals, according to any of claims 6-8, characterized in that the processing means (2) are configured to compare the stress pattern obtained with the stress pattern stored by diffuse logic.

         \ newpage
      

10. Stress detection device by physiological signals according to claim 9, characterized in that the processing means (2) are configured to compare patterns by comparing the centroids of the access pattern and the centroids stored in the storage media (3).