System and method for assessing the authenticity of dynamic holograph signature

The invention describes an interactive computer system and a method for the acquisition and processing of biokinetic information associated with the signature, in order to assess the authenticity of the dynamic holographic signature. The invention has applicability in the field of behavioral biometrics, in situations for which there is the interest of the user to validate the expression of his own will by holographic signature.

It is generally accepted that in the procedures for validating the declared identity of a person, in addition to the administrative, information, biometric-physiological methods and technologies, methods and technologies belonging to the field of behavioral biometrics can be used as an additional link. Acquisition and recognition of the constituent elements of the holographic signature is a class of behavioral biometrics.

From the patents RO 141297 and ΕΡΊ 846868, the description of which is included herein by reference, a procurement solution and processing methods are known for the purpose of verifying the holographic signature, a set of dynamic information - accelerations, associated with the signature. They are captured, processed and compared by a computer system. The retrieval of acceleration signals by inertial acceleration sensors - MEMS, integrated in an electronic pen and the specific topology of sensor placement, facilitates the acquisition of spatial kinetic information, as well as the acquisition of contact microvibrations with writing paper. By the processing methods on the acquired acceleration signals, it is determined:

- the beginning and end moments of the signature, a method realized by the algorithmic processing of the data that represents the variation in time of a distance distance to the paper, combined with the variation of the parameters amplitude and frequency of the contact microvibrations produced by the interaction of the subject / pen / paper of written;

- sequences of invariants associated with the initial acceleration signals, as well as sequences of invariants associated with the derived components of the initial acceleration signals;

- the distance between two signatures by algorithmic comparison of the sets of invariant sequences by two parallel methods;

- the final answer is determined by a decision method in which, through the results of the comparisons with the entry signature, the specimens of the subject concerned, as well as the specimens of other subjects registered in the signature base, participate.

The system and methods in RO 141297 and EP 1846868 treat spatial kinetic phenomena, combined with contact microvibrations, by processing the accelerations captured with MEMS acceleration sensors, but they have the disadvantage that they do not include the capture and processing of graphical information, which combined with kinetic information. by sensory fusion it would increase the accuracy of the system and the quality of human-machine interaction. The present invention aims to remove the mentioned disadvantages, by integrating, together with the inertial MEMS acceleration sensors, an optical navigation sensor, under the conditions of the sensory fusion concept. This optimization is exploited both by the new methods of multimodal algorithmic processing, graphically and kinetic, of the captured signals, as well as for visual feedback to the user, with advantages that will be detailed in the description.

US Patent 7176906 B2 (Microsoft Corporation) in the field of writing thickness interpretation uses electronic methods based on ballistic principles. A pen with accelerometer / accelerometers is used to capture information. For the interpretation of the information, the variation of the pulse width provided by the accelerometer or the inclination of the pen is used. The method refers to the use of a pen containing accelerometers, but which does not contain any element for the direct description of its graphical trajectory. The method proposes, for the association of the thickness information of the writing with the trajectory of the pen, the use of external devices, complex and expensive, of graphic tablet type. There are no issues specific to signature verification.

In contrast to the aforementioned patent, the present invention contains a module for capturing 1 pen-type signature, which integrates, by sensory fusion, together with accelerometers and an optical navigation sensor for capturing the graphical trajectory. The topological concept of 3 integration of sensors in the pen is mainly defined in order to ensure the spatial synchronization of the axes of the acceleration sensors with the axes of the optical navigation. By functional integration in the pen of both categories of sensors, the time synchronization of the acquisition of the two types of data is ensured; self-referential accelerations and displacements. Spatial-temporal synchrony 7 of the acquisition of information is the principle that governs the sensory fusion, whose application in the present invention produces a better accuracy in verifying the authenticity of the signature. The function of interpreting the thickness of the writing is defined and is based on the variation of both the frequency of the contact microvibrations and the variation of their amplitude as an effect 11 of the dynamic pressure that appears between the pen and paper in the gestural phenomenon of! handwriting.

US Patent No. 4,433,499 B2 (DynaSig Corporation) describes a system of acquisition and 13 signature authentication that contains, in the instrument description, acceleration and pressure sensors, and the method of authentication is accomplished by encrypting the data and comparing 15 resulting codes, thus avoiding retention clearly of the specimens. The patent does not treat information of a graphic nature, neither in relation to the writing instrument, nor in the 17 authentication method.

U.S. Patent 7483018 (Microsoft Corporation) discloses a dual system and method 19 for entering graphic data into a computer-type pen, in which the reference printed model analyzed by the integrated image sensor is the source of absolute or relative coordinate 21. , depending on the embodiment and the context of use. The system generates graphic or position information, but does not contain components and methods 23 for holographic signature authentication. The lack of integration in the capture module, in this case pen, of the acceleration sensors diminishes the potential of application in the field 25 of the signature authentication.

US patent 7508384 B2 (Data Research Inc.) describes a system and method of 27 acquisition and recognition of hand movement in the context of writing on a surface - Digital Writing System. The capture of spatial kinetic information is achieved by using a 29 3-axis gyroscope, a 3-axis accelerometer integrated into the pen and a proximity sensor to the writing surface. The patent does not describe sensors for capturing information of a graphic nature 31 nor algorithmic methods of authenticating the signature. The 3D trajectory data are estimated indirectly by dynamic linear filtering based on Kalman estimators (filters). 33

The procedures for which it is necessary to verify the holographic signature are highlighted and applied by concepts, principles and customs belonging to the dedicated fields of social-legal sciences, behavioral psychology and human neurophysiology. The field of electronics and information technology contributes to increasing the security of verification procedures through 37 methods of acquisition, processing and recognition of the signals and forms associated with the signature.

The ease of use, the degree of accommodation and the degree of acceptability on the part of the user of the biometric process are some of the essential parameters, defined in the specialized literature [http://www.biometrics.org], which participate as arguments in the classification of technologies 41 biometric. Biometrics over the holographic signature is preferred by most people, other biometric methods. The signature is considered, first of all, through 43 social utility and personal perception, a means of self-protection, exclusive and personalized. The justification resides in the individualized nature of the motor phenomenon by which 45 the signature is obtained, of reflex type acquired, conditioned by the existence of their own interest and of the free will. The fact that the present invention addresses the acquisition and processing of signals 47 specific to the signature psychomotority by maintaining the classical writing support - the paper, or any similar textured medium, determines a high degree of acceptance and accommodation. 49

RO 126248 Β1

During the signature, the user's forearm rests on the elbow, on a horizontal writing table, so that only the palm of the hand and the fingers that handle the writing tool perform the spatial gesture and dynamics associated with the signature production. In this context, a self-referencing optical navigation (ONS) sensor, located in the pen, almost parallel to the writing plane and near the writing tip, has as navigational points the distributed quasi-uniform elements of the paper / support model. The dynamic image of the printed model elements is taken through the lens (L), by projection on the photosensitive area of ​​the optical navigation sensor. The field of view of the lens concerns a quasi-constant area, which contains enough printed pattern elements for the ONS sensor to determine the relative displacements (relative to their own x, y, and y coordinates) relative to a momentary one. Based on the differences between consecutive images, internally sampled by the ONS in the continuous image, the sensor provides pairs of relative displacements dx, dy, with a fixed periodicity of the order of milliseconds, controlled and synchronized by the microcontroller in the pen, with the digital analog capture and conversion of the signals. acceleration (see reference RO 141297 and EP 1846868), taken from MEMS sensors A and MEMS B. As a result, a computerized graphical representation of the ONS sensor trajectory is similar to the graph made by the pen tip. The graphical differences between the representation of the ONS trajectory and the graph made by the pen mine represent a morphism of the same motor phenomenon, determined by the dynamic inclination of the pen. Morphism individualizes the signature in a biometric sense. At the same time, due to the synchronism between the graphics made on paper and the trajectory capture via the ONS sensor, the sequence of events that make up the two representations: the trace left on the paper with the pen and the electronic one captured by the system can be stored electronically, including the trajectory of the passages for which the pen does not reach the support. writing. For these situations where the signature realization contains momentary lifting gestures of the pen on the writing paper, and the amplitude of these rises is sufficiently small, of the order of millimeters or less, the optical navigation sensor continues to keep the focus of the printed model elements printed on write paper and capture the trajectory. The information produced in these passages continues to be acquired by the system, as they contribute to the signature's individualization. This information forms part of the signature and is treated by the module of graphic processing and display in a first instance through their visualization in order to achieve the gestic-visual perception by the user, and in the final instance, of calligraphic visualization, they are treated as segments that does not require display.

The holographic signature is a psychometric act. Psychosociality is defined as the result of the integration of motor and mental functions under the effect of maturation of the nervous system, which concerns the relation of the subject with his body. Psychomotricity thus appears both as an aptitude and as a complex function of regulating individual behavior, including the participation of different processes and psychomotor functions, which ensure both the reception of information and the proper execution of the response act.

According to J. Piaget [Theory of Cognitive Development, 1952], the psychic and the motor are not two distinct categories, one of pure thinking and the other of physical and physiological mechanisms, but on the contrary, they are the bipolar expression of a single process, that of efficient, flexible adaptation. , to external conditions. between the external conditions of the motive act and the subjective conditions, the signature is not only an execution mechanism, but a loop type circuit, in which each stage, each detail of the operations is the immediate expression of the relations established between the individual and the environment, respectively, between the will of the individual manifested psychomotorically through semi-reflexive gestures (acquired) and between the projection support of the expression of the will, in this case the paper on which the signature is made.

RO 126248 Β1

The fact that a signature is produced by means of the writing tool that makes practically 1 common body with the subject, during the act, guided the inventive concept towards the principle of capturing the motor variables (accelerations and relative displacements), by self-referential sensors integrated in the pen 3 (the tool of writing), thus emulating the non-intrusive integration of the sensors in the subject. The self-referential concept translates into practice the need for a close connection 5 between the observer (the sensors) and the psychomotor act, thus capturing the information from the sensoriomotor perspective of the subject, in the context in which the purpose is the observation and appreciation 7 of the individuality of the psychomotor act - the signature of the respective subject. The writing paper has relatively standardized qualitative properties, which is why the expression of individuality through 9 the signature act will be affected to a small extent by any qualitative differences of the paper. 11

The cumulative attributes of the trajectory, captured by the ONS sensor, together with the sequentiality and the gesture, spatial captured by the MEMS acceleration sensors, constitute elements 13 which, realized by the free will of the subject, individualize the signature.

From the user's point of view, the role of visual feedback 15 in the system is dual:

- favors cognitive mechanisms of gesture perception, offering the user 17 the possibility of acquiring through training and visualization, of complex gestures, not materialized on paper, but belonging to the signature kinetics. These acquired gestures can be created voluntarily by the user, in order to further individualize the signature,

- facilitates the user's accommodation with the system specificity, through the 21 psychomotor mechanisms combined with those of the visualization.

The consistency and complexity of the gestures associated with the signature, acquired naturally in 23 time or by training, are limited only by the motor skills and the ingenuity of the user to compose a gesture as individualized as possible. 25

The present invention relates to a system for the acquisition and processing of information associated with the dynamic holographic signature, made on writing paper support or on any support with 27 similar texture, which has printed a quasi-uniform printed model, the system comprising an electronic pen, provided with two groups of inertial MEMS acceleration sensors, for 29 capturing kinetic information and microvibration contact information with the support, integrated In a personal computer, for the acquisition and processing of the signals associated with the 31 signatures, which can be connected in the network, along with other personal computers that have connected as peripheral an electronic pen, in which the electronic pen has 33 components: a self-referential optical navigation sensor, for capturing the series of data pairs (dx, dy), as momentary displacements necessary to reconstruct the trajectory of the pen 35 and which together with the kinetic information (ax, ay, bx, by), captured by the set of inertial MEMS sensors, are input information in the personal computer, for 37 processing by sensory fusion, as well as for the fulfillment of the conditions for taking the information from the perspective of the psychomotor and sensory representation of the pen user; 39 an infrared LED for illuminating the printed model; a focusing lens and projection of the current image of the model printed on the sensitivity area of ​​an optical navigation sensor 41, positioned so that the image projection of a region of the printed model is performed on the sensor's area of ​​sensitivity, the condition of spatio-temporal synchronism 43 necessary for the sensory fusion of the two categories of data: dx, dy, of a graphical nature and of the accelerations ax, ay, bx, by, being achieved by the axial topological quasi-alignment of 45 of the 3 origins of the coordinate axes of the M EMS sensors and the optical navigation sensor and the tip of the pen mine, and by planparallel topological quasi-alignment of the 47 sensitivity / coordinate axes, of the three sensory components from two M EMS sensors and one

RO 126248 Β1 optical navigation sensor, with the writing plane and, by sampling the information captured by the sensors, with a constant periodicity, in the range 1-8 ms, which ensures, by the high frequency, the synchronism in time of the acquisition of the graphic and kinetic details , the acquisition being controlled by an acquisition microcontroller that transmits the information in real time to the personal computer that hosts the algorithmic visualization methods, to reconstruct the thickness of the writing and to process and compare; for the purpose of collecting information from the perspective of the user's psychomotor and sensory representation, the set of MEMS sensors and the optical navigation sensor, integrated in the human-handled writing tool, capturing the 6 associated signals ax, ay, bx, by, dx, dy in systems of self-referential coordinates, respectively, their own x, y axes of sensitivity coordinates; and to methods for appreciating the authenticity of the dynamic holographic signature, consisting of a phase of calligraphic reconstruction and visualization of the thickness of the writing and of a phase of applying conversions and comparisons on the signals, ax, ay, bx, by, dx, dy , associated with the signatures in order to appreciate their authenticity.

An example of embodiment of the invention is given below, in connection with FIG. 1 ... 11, which represents:

FIG. 1, the functional block diagram of the system;

FIG. 2, the physical modules of the system;

FIG. 3, the diagram of the signature capture mode - pen;

FIG. 4, the functional scheme of the signature capture mode - pen;

FIG. 5, ONS topology, MEMS A, MEMS B, Mine pen;

FIG. 6, toplogical detail of the pen module;

FIG. 7, reconstruction of the total trajectory made by the pen;

FIG. 8, the reconstruction of the graphical trajectory in the moments of contact with the paper;

FIG. 9, the general framework for the conversion and comparison of signatures;

FIG. 10, synthesis of signature analysis algorithms;

FIG. 11, determining the slope difference between two consecutive segments of a plane curve represented by points,

The system described in the present invention is physically composed of the electronic pen module 1, which integrates the sensory assembly for information retrieval and a personal processing computer 2, for the information captured by the pen, by algorithmic methods.

Functionally, the system is shown in the block diagram of fig. 1. The system realizes the motor phenomenon produced by the hand-pen assembly, associated to the holographic signature, by sensory fusion of two categories of signals: acceleration, captured by MEMS A, MEMS B and displacement sensors, captured by the ONS optical navigation sensor. The specific topology regarding the location of the sensors in the pen 1, as well as the principle of sampling the acquired signals, ensures the condition of spatio-temporal synchronization between the signals of the same phenomenon. This condition is necessary for the realization of the concept of sensory fusion for the purpose of multimodal algorithmic processing.

The invention deals with the self-referential capture of the pen trajectory performed by an optical navigation sensor, integrated into the pen-type module, as well as the method of graphically displaying the pen trajectory in the first instance, by the calligraphic display of the pen trajectory for those segments of the trajectory in which the pen is in dynamic contact with the paper during the signature. in FIG. 7 represents the reconstruction of the total trajectory made by the pen, and in fig. 8 it represents the calligraphic reconstruction, with a variable thickness associated with the moments of dynamic contact of the pen mine with the paper / writing medium for the same signature. The electronic representation, on the monitor of the personal computer 2, corresponds approximately to the signature calligraphy produced by the pen mine on the writing medium.

RO 126248 Β1

The methods for evaluating the authenticity of the signature, according to the invention, refer to the 1 processing and multimodal, graphical and kinetic comparison, of the signals associated with the signatures, and will be described in detail in the embodiment. 3

The signals captured and displayed include, in a first phase of the visualization on the personal computer monitor 2, and the signature passages in the air, for example: the differences between fig. 5 7 and fig.8 representing the same signature. They contribute, together with those from the paper movement, to the assessment of the authenticity of the signature. 7

The psychomotor manifestations involved in producing the signature contain elements of individuality, which the system determines as series of invariants and data vectors evaluated 9 automatically by the methods of acquisition / processing / comparison. These methods can be added to those of patents RO 141297 and EP 1846868, the description of which is included herein by 11 references. At the same time, the hardware structure of connection, processing and decision (computer, server) or its variants described in the patents preceding the present, (RO 141297 and 13 EP 1846868), contains the necessary and sufficient elements to constitute the platform for running the integrated working algorithmic methods. with the pen capture module (1) described in the present invention.

The new concepts for capturing the signature, according to the invention, embodied in the electronic pen module 17 and the methods of processing and comparison are described in detail and detailed below. 19

In order to capture and reconstruct the graphical trajectory of the pen, for the purpose of biometric verification and for the purpose of visualizing the signature, the 21 ONS optical navigation sensor was integrated into the pen, instead of the threshold distance sensor C described in the patents RO 141297 and EP 1846868. In the example In order to achieve this, an optical navigation sensor from the 23 family was used

ADNS (AGI LENT manufacturer). The threshold distance detection function is taken over by the ONS sensor through its property of generating the output data only in the context in which the captured image is focused enough for the sensor to analyze the printed model, respectively, the ONS sensor is relative to the model paper. printed P at a distance 27 within a range d determined by the optical properties of the lens lens L. In a practical embodiment, this range d is 5 - 20 mm for a lens made of 29 polycarbonate (organic glass) with refractive index n = 1.5. The ONS sensor is a self-referential sensor, having the property to generate as the output the relative displacements dx, dy, of the 31 image projected by the L lens on its sensitivity area, in a well-determined time (the sampling period). The displacements dx, dy correspond to the displacements of the image projections 33 of the printed model, printed on paper. The image captured at the beginning of any sampling cycle is the momentary reference, for a pair of values ​​dx, dy. 35

The displacements dx, dy, are estimated from the coordinate system of its own x, y axis, of the sensor. The ONS sensor is in solidarity with the entire parallel plane topology of the axes 37 of the MEMS acceleration sensors (MEMS A and MEMS B - see RO 141297 and EP 1846868, integrated in the pen, according to the topology in Fig. 5. Sensors were used in the embodiment 39 acceleration inertials with two x, y axes, with the output range +/- 2g, with analog output, from the MEMS ADXL family (manufacturer Analog Devices). MEMS sensors continue to have the same 41 role as in the previous patent cited as reference, in addition, by the processing method to be described, the variation of the amplitude of the signal in the frequency band above 150 Hz, 43 corresponding to the microvibrations captured after contact with the paper will constitute an input for the method of calligraphic interpretation / visualization of the trajectory performed by the ONS 45 sensor integrated in the pen.

RO 126248 Β1

The displacements dx, dy captured by the ONS sensor are evaluated by the sensor by internal processing of multiple consecutive images captured during a sampling period, in the order of milliseconds - in the embodiment the sample period of the ONS sensor is 2 ms. The processed images represent the projection on the photosensitive area of ​​the sensor, through the L lens, of the printed model P printed on the paper / writing medium. The topology shown in fig. 5 (detailed in FIG. 6) achieves the optimum compromise between the condition of alignment of the axes x, y, coordinates / sensitivity, of the three sensory systems MEMS A, MEMS B and ONS, the condition of non-intersection of the tip of the pen with the visual field of the lens and the condition of minimizing the errors of navigation due to the dynamic inclination of the writing process. The ONS sensor has maximum sensitivity for images captured in the near-infrared band. Therefore the illumination of the printed model is realized by means of the IR LED with emission in the near infrared band, located as in fig. 5, detail in FIG. 6.

The triggering of the fixed cycle sampling cycle is controlled by a microcontroller μΟΑ, which manages the synchronous sampling of both the four acceleration signals from the MEMS sensors and the two dx and dy displacements generated by the ONS sensor.

At each time n, in computer 2, the algebraic sum of the elements of the series dx (n), dy (n) is performed, which represents the approximation of the projection of the ONS sensor path included in the pen 1, through a flat curve, on the writing support, in the context dynamic and cursive signature creation, with pen tip lifts on the paper / support that fall within the range (d = 5-20 mm) to maintain the image focus of the printed model P through the L lens on the photosensitive area of ​​the ONS sensor.

The set of graphical signals dx (n), dy (n), together with the set of acceleration signals digitized ax (n), ay (n), bx (n), by (n), is transmitted via USB protocol through through the specialized microcontroller pC-USB, included in the pen, to the computer 2, for the purpose of purchasing, viewing, processing and comparing the signature. The graphical data set has a constant periodicity, the well-determined time intervals ensuring phenomenological synchronization with the acceleration signals collected through the MEMS Ă and MEMS B acceleration sensors, which in turn have a constant periodicity, for example: 1 ms. The writing plan / paper has a printed model P, with quasi-uniform distribution. The role of this printed model is to be the source of the static marks necessary for the ONS sensor for the calculation of the pair of values ​​dx, dy, which expresses the relative displacement of the ONS sensor, performed during two successive readings. Finally, at the output of the pen module 1, a sampling cycle corresponds to six signals, transmitted in USB protocol to computer 2, respectively, four acceleration signals: a_x(n) - the digitized signal generated by MEMS A in the X direction of point A, a_y(n) - the digitized signal generated by M EMS A in the direction of point A, b_x(n) - the digitized signal generated by MEMS B in the direction of point B, b_y(n) - the digitized signal generated by MEMS B in the direction of point B and two of movement: d_x(n) - the signal generated by the ONS in the x direction, d_y(n) - the signal generated by the ONS in the y direction. The x, y directions correspond to the internal axes of the MEMS and ONS sensors.

The system achieves the following specific functionalities:

- capture the projection of the trajectory of the writing device (electronic ballpoint pen), in the context of making a signature on writing paper or on another medium with similar texture that has a diffused printed model P, with quasi-uniform distribution. This function is performed by means of the self-referential ONS optical navigation sensor, integrated in the pen with which a subject realizes the signature. The utility of this functionality is dual: it generates graphical information that is exploited by the system in the algorithmic decision methods.

RO 126248 asupra1 on the authenticity of the signature as well as the production by the system of the visual feedback, 1 displayed on the system monitor, necessary for the user in the procedure of registering the signature or to accommodate with the system and visual validation by the subject of an acquisition following 3 a be authenticated;

- detection of the beginning of the signature and its end by the combined processing 5 of two categories of signals: the displacement ones, captured by the ONS sensor, as well as those corresponding to the microvibrations captured by the MEMS A acceleration sensors integrated in the pen. 7 The method is based on the analysis of the amplitude and frequency parameters of the microvibrations that appear, together with the data on the self-referential trajectory of the pen, at the dynamic interaction 9 of the gesture-pixel-fiber elements of the paper. This functionality is implemented as an algorithmic method, being contained in the system application, residing in a computer 2; 11

- visualization of the pen trajectory. This takes place immediately, practically in real time, and following the processing are displayed also calligraphic effects (information) of the thickness of the trajectory 13 as well as the graphics of the traces produced by the paper writing mine. This functionality has the role of providing the human subject with the visual feedback necessary to accept the acquisition made by 15 systems. The algorithmic module corresponding to the visualization functionality is contained in the application of the signature authentication system, residing in a computer, and the visualization 17 takes place on the computer monitor;

- multimodal processing and comparison of the acquired signals / shapes Se 19 is performed by the algorithmic methods described below, having as input elements both the acceleration signals captured by the MEMS sensors and the signals / shapes taken by the 21 ONS sensor. The characteristics of the algorithmic methods, to work specifically in n-dimensional spaces (2D, 4D, 6D) with the 6 acquired signals, as well as the type of signals 23 (graphical and acceleration), determine the multimodal character of the system methods.

The pen and the methods of the present invention are functionally and unitarily associated, being realized 25 as the module and set of algorithmic methods contained in the application of the signature authentication system.27

The method of calligraphic reconstruction of the thickness of the writing displayed on the computer monitor 2 is implemented by the following steps that are carried out in the computer29 personal 2:

In a first stage, for the calligraphic reconstruction of the writing thickness, the trajectory is calculated 31 by the algebraic summation of the series of values ​​of the signals dx, respectively, dy, obtaining the coordinates of the momentary displacements captured by the optical navigation sensor 33 ONS, which is the trajectory graphically reproduced on the computer monitor 2.

The second step is to filter the acceleration signals a_x and_y, associated with a 35 signature. Digitized acceleration signals a_x and_y from the MEMS acceleration sensor A is filtered through a filter with a cutoff frequency of 150 Hz 37, obtaining the signals c_x and cy. Each of the signals a_x, a_y, c_x, c_y actually represent sample vectors represented as positive integers. 39 in step three, the values ​​of RMS v_x(n) and v_y(n) within a time interval (ni, n) and the results obtained will be associated with the time moment n corresponding to the instantaneous values ​​of signals c_x and c_y so:

<img file="RO126248B1_D0001.tif" />

<img file="RO126248B1_D0002.tif" />

where c_x(n) represents the sample of signal c_x at the time of time n.

RO 126248 Β1

The size resulting from the arithmetic mediation of the two effective values, v_x(ri) and v_y(n), is the one that contains the information used to reconstruct the thickness of the writing. within the range of the signal m (n) we consider the interval [a, b], where a and b represent the minimum, respectively, maximum thickness, associated with the writing at a given time. The values ​​a and b were determined experimentally.

In step four it is determined the case where the segments of trajectory for which the pen is not in contact with the writing medium are removed from the initial graphical representation and the thickness information is associated with the non-eliminated trajectory segments, thus making the calligraphic interpretation. Thus, if any of the values ​​v_x(n), v_y(n) or v_xW + v_y(«) _callţ _mej _mand_Cj ₀a threshold value p, determined experimentally, is considered

1.5 that the thickness associated with writing at time n is 0. For the times when this condition is met, the segments corresponding to the trajectory (fig.7) captured by pen 1 are deleted from the initial graphic representation (fig. 8) displayed on the computer monitor 2 , thus resulting in a calligraphic representation similar to that drawn by the pen mine on the writing medium.

The present invention also provides the following methods for recognizing signatures: SRA3, SRA5, SRA7, SRA8 (SRA = System Recognition Algorithm), in order to assess the authenticity of dynamic holographic signatures, see fig. 10.

They process the signals produced by the pen. To describe the processing algorithms, note the signals generated by the pen as follows:

- a_x: the signal generated by MEMS A in the direction of point A;

- a_y: the signal generated by MEMS A in the direction of point A;

- b_x: the signal generated by MEMS B in the direction X of the point B;

- b_y: the signal generated by MEMS B in the direction of point B;

- d_x: the signal generated by the ONS in the x direction;

- d_y: the signal generated by the ONS in the y direction.

Each signal is actually a sample vector represented in the form of whole numbers. This vector is a numerical representation of a waveform. All vectors of a given signature have the same length (same number of samples).

These four methods are based on the following common principles:

Each method contains two modules (see fig. 9).

Input data processing module. Through a series of operations, representative data is extracted from the input data that is used to store the data regarding specimen signatures and to represent the data constituted by the input signatures (original or false) to be recognized. We will call all of these operations the method of converting signatures 3.

Comparison module of two signatures, one specimen and one input. We will call all these operations the method of comparing signatures 4.

The assembly consisting of the Conversion Method and the Comparison Method will be called the Signature Recognition Method.

The conversion of the input signals (which the input data processing module does) to a format usable in the comparison process comprises the following steps:

a) Conversion of signature signals into invariants.

b) Weighting of the invariant sequences.

RO 126248 Β1

The specimen signatures are converted and stored in the signature base. Subsequently, 1 when an entry signature appears (original or false), it is converted and compared with the signatures from the signature base, calculating the distance between the entry signature and 3 specimen signatures thus determining whether the subject of the entry signature is the subject. specimen signature. 5

The four methods mentioned above are based on a general method of converting curves (approximated by right segments) into sequences of invariants, that is, of elements 7 that do not vary with respect to the frequency or amplitude of the signals (in the case of accelerations). or in relation to the scale (in the case of graphic signals). 9

The SRA3 and SRA5 methods work with curves represented in the plane (2 dimensions), the SAR7 method works with curves represented in a six-dimensional space and SRA8 11 works with curves represented in a four-dimensional space (fig. 10).

The SRA3 method of working with curves in the plane is described first. 13

The first step is to convert the signals into invariants.

Let a curve 5, given by points in the plane XOY according to fig. 11. This curve is 15 given by the signals d_x, d_y.

We consider on this curve three consecutive points Τ ', T_{i + 1}, T_{i + 2} of coordinates, 17 respectively, (χ, y), (x_{i + 1}, y₍₊₁), (x_{i + 2}, and,_?) where i = 1,2, ..., n-2 and n is the total number of points on the curve taken into account. The values ​​of x_it x_w, x_{j + 2} they come from the curve dx and the values ​​y ,, y_w. it's there_{i + 2} come from curve d_y.

We determine the slope p₍ of the vector TjT_{i + 1} with the OX axis, for example as follows: 21

If x_{i + 1} &gt; X, and y_{j + 1} = γ, then p, = 0.

If x_{i + 1} &gt; X and y_{i + 1} &gt; γ then p, = atan ((y_{i + 1} - y_;) / (x_{i + 1} - χ).) 23

If x,₊₁ = x „and y ,,&gt; γ then Pj = rr / 2.

If x_{j + 1} &lt;X | and y_{j + 1} &gt; y, then p, = n / 2 + asan ((Xj - x_im) / (Y_{i + 1} - Y)) · 25

If x_{i + 1} &lt;Xj and y_{i + 1} = y, then p, = n.

If x_{i + 1} &lt;X, and y_{i + 1} &lt;y, then p, = π + carefully - y_{i + 1}) / (x, - x_{i + 1})).27

If Xj₊₁ = Xj and y₍₊₁ &lt;y₍ then Pj = 3 * n / 2.

If x_{t + 1} &gt; x_t and y_{t + 1} &lt;y_t then μ- = 2 * π - atan ((y - y_{i + 1}) / (x_{i + 1} - χ)); 29

If x_{i +}i = N and y_{i + 1} = y then pj = 0.

(By atan () I noted the arctangent function.) 31

The slope p is calculated analogously_{i + 1} of the vector T_j+₁T_{i + 2} with the OX axis.

We then consider that in point T_{i + 1} we have a new X 'T coordinate system_{i + 1}Y 'and 33 determine the orientation of the vector T_{j + 1}T_{i + 2} In this coordinate system, or in other words we calculate: 35 η = lp₂ - p ^ I

We now determine two codes: 37 caj = the absolute orientation code cr, = the relative orientation code39

The absolute orientation code as, is obtained by dividing in S_of sectors the circle that would have the center at point T_{i + 1}. These sectors are numbered starting from the horizontal axis 41 passing through T_{i + 1} and heading towards X. We will have:

caj = int (e.g._{j + 1}/ (2 * n / S._of)) 43 (I noted by int () the whole part.)

The value of S_of is determined experimentally (for example it can be S_of = 8).45

The relative orientation code but; is obtained by dividing it into S_r sectors the circle that would have the center at point T_{i + 1}. These sectors are numbered starting from axis BX '. We will have: 47

If P_{i + 1} &gt; = Pj then cr, = ίηΐ (η / (2 * n / S_r))

RO 126248 Β1

If P₂ &lt;p₁ then c_r = S_f -1 - int {p / (2 * n / S_r)) finally, a single invariant code c is obtained, in the form:

Ci = CTi * S_of + horses

The value of S_r is determined experimentally (for example it can be S_r = 145).

Starting from each point of the curve, a sequence of invariants encoding the curve is determined by the analysis made on three consecutive points.

The second stage is a weighting step of the invariants.

Because an invariant is defined on the basis of 3 consecutive points on the curve T ,, T_{i + 1}, T_{i + 2}, the length of the invariant Lj (i = 1,2, ..., n-2) will be considered as:

A = ⁺ h + l - ⁺ - y_M)²

We define l_t as a total reference length and calculate it as follows:

i = n-2 / = 1

The weight w is now determined₍ of each invariant as the ratio between the length Lj of the invariant and the reference length L_r:

Wi-Lj / L,

Thus each invariant will be represented by the pair (q, w_f) which can be packed in a single C code, in the form:

O, = Ci + Wi * S_of * S_r

The SRA5 method is similar in principle to the SRA3 method, meaning that all the analysis of a curve in the plane. However, there will be two curves that are analyzed as two separate components. The first curve will be given to_x and_y and the second curve will be given by b_x and b_y. In this way, the curve corresponding to an acceleration (in point A, respectively in point B) becomes a flat curve, independent of time, but having an order of travel given by the sequence of the samples (ie the points on the curve) over time.

We now describe the SRA7 method of working with curves in a space of six dimensions. The first step is to convert the signals into invariants.

We consider a curve given by points in a space with six dimensions (numbered 0, 1,2, 3, 4, 5) and three consecutive points on this curve T ,, T_{i + 1}, T_{i + 2} coordinates,

<td colspan="4">respectively:</td>

<td>T</td><td>(u_HQ, u_l1 , u_II2, u_ij3, μ₄</td><td>'^ ί δ)</td><td></td>

<td>Tj +₄</td><td> (^Ui + 1, Q&gt; ^Ui + 1.14 ^Ui + 1,2&gt; ^Ui + 1,3 '</td><td>^UIM4 '</td><td>^Ui + 1.5.</td>

<td>Tf + 2</td><td> (^Ui + 2,0&gt; ^Ui + 2,1 · ^Ui + 2,2 · ^Ui + 2,3&gt;</td><td>Ui + 2,4&gt;</td><td>^Ui + 2,5)</td>

waves = 1, 2, ..., n-2, and n is the total number of points on the curve taken into account.

U values_{and D}, u_{£ i + 1)}, u_{j + 2 0} come from the curve a_x, u values_i4, u_{i + u}, u_{l + 21} come from the curve a_y, u values_{and 2}, u_{j + 12}, u_{i + 22} come from the curve b_x, the values ​​μ ₃, u_{i + 13}, u_{i + 23} come from the curve b_y, u values_i4, u_{i + 1}4, u_tut.₂ 4 come from curve d_x and u values_{i 5}, u_{i + 15}, u _i2;. come from curve d_y.

We determine the angle p, of the vector with the vector T_{i + 1}T_{j + 2} as follows: We first calculate two values ​​s_t and s₂ with the formulas:

And ⁼ (Uj + ι, ρ - u_{i 0}) * (u_{i + 2 0} - U_{l + 10}) + (U_{l + 1j1} - Ui J * (u_{i + 21} - u_{i + 11}) + (^Ui + 1,2^{_ u}i, 2) (^Ui + 2.2 Ui + v) ⁺ (^Ui + 1.3 U_{and 3}) * (U_{i + 2 3} - U_{i + 1 3}) + (U.- 4 - u_i4) * (U_{and + 214} - u_{i + 14}) + (^Ui + 1.5 ^Ui, s) (^Ui + 2.5 ^Ui + 1.5)

RO 126248 Β1:

$2 ⁼ &lt;(^uÎ + io - μ, o)² + Ki.1 - μ, ι)^{2 +} (^UK2 - μ, / ⁺(^Ui + 1,3 ^{U +} (μ + 1.4 μ, 4) ⁺ (^Ui + 1 5 μ, δ)) ((Ui + 2i0 - μ * ί, ό)² + (μ + 2, ι - μ₊ι, ι)^{2 +} (Mfr2,2 - μ + 1,2)² + (ui + 2j3 - u_{i + 1 (3})² + (ui + 24 - ui + t4)² + (u, 2 .. - u_M5)²)

If₂ # O and&gt; = O, then pj = arccosțs., /).

If₂ # O and s₄ &lt;O then p₁ = π + arccos (s₄ /

<img file="RO126248B1_D0003.tif" />

If₂ = 0 then p, = 0.

We now determine the code c of the invariant ic, = p, / (2 * n / S_r) where the value S_r is determined experimentally (for example it can be S, = 7).

The second stage is a weighting step of the invariants.

Because an invariant is defined on the basis of 3 consecutive points on the T-curve_hT_{i + 1}, T_{i + 2}, we will consider "the length of invariant L, (i = 1,2, ..., n-2) as:

A ^_ ^(^Μί + ϊ 0 ^_ ^&gt;,0) ^{+ _} ^,1) ⁺ (Ί + ϊ ^ 2 ^!/ :) ⁺ (^Μή1,3 ^_ Α, ΐ) ⁺ (^Μί + 1.4 “Α, ΐ) ⁺ (^M5 » ^Ui, s} ⁺

<img file="RO126248B1_D0004.tif" />

We define L_t as a total reference length and we calculate it as follows: 17 i = n = -2

L, = ΣΤ, / = 1

Now determine the weight w, of each invariant as the ratio between the length 21 L, of the invariant and the reference length L_r:

W₍ = L / L_r23

Each invariant will therefore be represented by the pair (c ,, w.) Which can be packed in a single code C, in the form:

C_and = c_and + w_and* S where S is a value greater than the largest c, (for example, S = 32768 is taken) .27

The SRA8 method is analogous to the SRA7 method but works only in four given dimensions_x, a_y, b_x, b_y.29

By the methods SRA3, SRA5, SRA7, SRA8, the input signals are translated into rows of invariants, each invariant having a certain weight (cost). Comparing two signatures, depending on the respective algorithm, means finding the distance between two rows of symbols, belonging to the two signatures that are compared. To calculate this distance it is considered that the most suitable algorithm is the Levenshtein distance. Finally, if the result (Levenshtein distance) is D, then the considered distance (normed) d will be: 35 ____D

Σ cosr, + cosz) where cost, = w ,, and cos = Wj, represent the costs of the invariants of the two components.