RU2696042C2 - Method and system for recording eye movement - Google Patents

Abstract

FIELD: measuring equipment.

SUBSTANCE: invention relates to measurement equipment and computer technologies, namely to technologies used to determine areas of gaze fixation or gaze when moving eyes, and can be used for objective assessment of processes of visual attention, in operator's activity, for control of computer interfaces by means of direction of sight. Disclosed is a method of determining gaze fixation areas during eye movement, characterized by that respondent with fixed position of head is presented with, at least, one stimulant image, during presentation of the stimulating image, covering the respondent's eye region with infrared radiation, recording a reflected signal for at least 50 ms to obtain at least 5 frames of the pupil of the pupil of the eye, processing the resulting video sequence with determining the coordinates of the pupil centre in each frame of the video, after which these coordinates are converted into a coordinate system of the stimulus image, forming a two-dimensional array of values of coordinates of the pupil centre in a chronological sequence corresponding to the frames of the video sequence, from which segmentation analysis is used to select at least three points considered to be a gaze fixation region. Also disclosed is a system for determining areas of gaze fixation during eye movement, including an infrared emitter, arranged to cover the respondent's eye region, a reflector filter configured to transmit visible light and reflect IR radiation, set at the level of the respondent's eyes, stimulus image presentation unit located behind the reflecting filter, a unit for recording the IR-radiation reflected from the pupil in the form of an image of the pupil of the eye, comprising at least five frames, including a video camera, and an image processing unit from a video camera, configured to process data by segmentation analysis and obtain areas of gaze fixation when moving eyes.

EFFECT: group of inventions enables processing large arrays of images with high reliability of highlight areas.

12 cl, 2 tbl, 7 dwg

Description

FIELD OF THE INVENTION

The present invention relates to measuring equipment and computer technology, in particular, to the technology used to determine the areas of fixation of the gaze (or gaze) during eye movement, and can be used to objectively evaluate the processes of visual attention, in operator activity, to control computer interfaces by directions of the gaze, in marketing, etc. It should be noted that fixing the gaze (or gaze) during eye movement is carried out with a fixed position of the respondent’s head relative to image being presented. The invention is applicable mainly for viewing static images.

State of the art

The dynamics of the change in gaze direction, as well as the duration of fixations at individual points of the object in question is an extremely informative parameter that characterizes the person’s attention in the process of cognitive activity [Velichkovsky VM, Pannasch et al Neurocognitive investigations of human eye movements: Interaction between levels. XX Conference of the Physiological Society "I.P. Pavlov". Moscow (2007), Franke, I.S., Pannasch, Velichkovsky B.M. et al (2008). Towards attention-centered interfaces: An aesthetic evaluation of perspective with eye tracking. ACM Transactions on Multimedia Computing Communications and Applications, 4 (3), Art. No. 18]. Extensive research in this area has led to the development of a number of technologies that allow optimizing human interaction with technical means, optimizing information flows in terms of the best perception of the information provided by a person, and controlling the attention of a person involved in complex production processes in order to reduce the likelihood of situations caused by the human factor .

The development of these technologies, as well as the implementation of research activities in the field of cognitive ergonomics, is directly related to the creation of systems for determining the direction of the gaze and the point of fixation of attention directly on the object in question (algorithms for identifying fixations and saccades). Currently, there are a number of commercial systems:

1) SMI (https://www.smivision.com);

2) Tobii (https://www.tobii.com);

3) Eye Tribe (https://www.theeyetribe.com),

allowing to determine these parameters, however, they have several drawbacks, which include restrictions on processing speed and other limitations of information processing algorithms.

In particular, the method and system for detecting saccades based on the determination of angular velocity, published on the psychological portal PsyJournals.ru http://psyiournals.ru/exp/2008/n1/Demidov_Jegallo_full.shtml [SMI equipment for detecting movements, is known from the prior art Eye: Test Drive - Experimental Psychology - 2008. Vol. 1, No. 1]. The system interprets the movement as a saccade in the event that the speed of movement in the middle of the path from one point to another exceeds a certain threshold value, for example, 75% ec. Otherwise, the system detects fixation. A similar algorithm is proposed, for example, by Smith and Hoog (Smeets & Hooge, 2003). This algorithm is recommended to be used for installations with a high sampling frequency (more than 200 Hz) in order to be able to recognize short saccades, which are very common, for example, when reading. An alternative to this algorithm is the case of using angular velocity values for the analysis of fixations (Rotting, 2001). If the angular velocity is below a certain value (usually less than 10 ° / sec), the system detects the fixation. The algorithm is used to cut off “noise” in the data.

However, the algorithms for extracting fixations by the criterion of speed and based on the variance of the points of view of the gaze have a number of limitations, the main of which is the criterion of the speed of information processing and the ability to adequately and accurately select the level of speed or coordinate value, against which the algorithm compares the current and previous values of the analyzed indicator . Using the proposed method, including the segmentation analysis algorithm, through which the data array is processed, requires less time. From this follows an increase in the processing speed of the array. This entails, accordingly, a reduction in the cost of machine resources. In general, it should be noted that the use of the proposed method leads to a significant reduction in errors in the selection of fragments of the array.

From this source of information there is also known a method and system for detecting fixations, based on the determination of variance and duration of fixation. A fixation is a set of neighboring samples located inside a space of a given diameter (usually a circle), with a total duration of at least a given. An example of this algorithm is the I-DT algorithm proposed by Salvuchi and Goldberg (Salvucci, Goldberg, 2000).

However, the disadvantages of these algorithms are the low processing speed of the coordinate arrays and inaccuracies in the allocation of relevant oculomotor events. The most critical aspect of the operation of the algorithms are errors of the first and second kind during detection in the data array of areas corresponding to oculomotor events (saccades and fixations). In addition, it is known that it is important to define “boundary conditions”, i.e. those spatial and temporal criteria by which event detection should begin. The segmentation algorithm used in the claimed invention allows for optimal detection. His work is based on the selection of "segments" that differ from the entire array. Thus, the approach captures the events themselves, and not their beginning and end, as the above algorithms do.

In addition, a detection method and system based on the determination of angular velocity using an additional criterion of the duration of fixations is also known from this information source. This method uses the principles of detecting eye movements from the two above algorithms.

However, it is the combination of the two approaches for detecting gaze fixations that creates restrictions on the indicator of data processing speed. The segmentation algorithm used in the claimed invention reduces the number of errors and, thus, increases the efficiency of extracting data fragments from the array corresponding to the commits and saccades.

The problem of detecting eye movements is particularly relevant in cases of access to automated software for data processing. In the early oculographic studies with manual data processing, the use of saccade detection and fixation algorithms was largely unnecessary, since the accuracy of the researcher himself, together with the resolution of the recording setup, themselves were some kind of detection algorithms.

Closest to the claimed is a method and system for tracking eye movement (Patent US 7522344). The solution is an optical system containing a projection mounted on the head (head-mounted) display with an optical system inside, used for IT tracking. In particular, the system contains the following structural elements:

- a double light source to provide the first and second IR emitters, respectively, for infrared illumination of the respondent’s eyes, allowing you to create several so-called “glints” - reflections on the surface of the eye, according to which the direction of view is detected. The system contains the ability to dual light - illumination of the eyes with two types of light to obtain an image of the pupil with alternating frames of positive and negative contrast. This is required to subtract one image from another to obtain maximum accuracy in determining the pupil for aitracking;

- an optical system for creating a head-mounted display of the optical part and the tracking part; however, these parts are combined to reduce the weight of the projection part. The optical system also includes first and second warm mirrors that reflect infrared rays and transmit visible light, and which are necessary for projecting an eye image onto the camcorder's matrix;

- a video camera for recording eye movements by recording several reflections of infrared radiation; wherein one radiation source is located on the axis of the camera, and the second outside the axis; during operation, the camera takes frames (with positive and negative contrast), alternating them, and then subtracting one from the other;

- a device for processing, which, based on a set of algorithms, determines the position of the pupil of the eye of the respondent based on information about the movement obtained from the light and dark images of the pupil recorded by the camera.

- a display for presenting images to the respondent.

The method of providing eye tracking inside a projection head-mounted display, presented in the materials of patent US 7522344, includes:

- the formation of multiple infrared beams of rays from the first and second source of infrared radiation to illuminate the pupil of the eye of the respondent;

- the use of the first and second warm mirrors together with the first and second source of infrared radiation, respectively, to obtain an image of the eye on the matrix of the camera;

- alternating recording of multiple reflections (glints) on frames with a pupil in positive and negative contrasts;

- obtaining data on eye movements from a light image of the pupil and a dark image of the pupil.

However, the proposed solution is characterized by a number of disadvantages. For example, double lighting and two warm mirrors increase the size of the system and create additional factors that distract the participant’s attention during the registration process. At the same time, the accuracy obtained by registering a pupil using an approximation based on several reflections of the backlight is lower due to the multiplication of the error associated with each individual reflection (glint). From an economic point of view, the proposed system has a number of disadvantages associated with cost appreciation due to the use of 2 warm mirrors and two IR emitters used in the system.

Due to the fact that analogues use event extraction algorithms that operate with boundary conditions (when searching for fixations), there is always a high probability of losing values from an array of data (coordinates of viewpoints). Losses can also be associated with omissions in data arrays. In such cases, the dispersion algorithm and the search algorithm according to a given threshold can irrelevantly evaluate the available artifact sections in the data table, which in turn leads to errors in obtaining the final output.

Disclosure of invention

The technical problem solved by the claimed invention is to overcome the disadvantages inherent in the analogues disclosed in the description of the prior art by creating a method and device for determining the areas of gaze fixation during eye movement, providing reliable data while reducing the duration of the recorded video sequence of images of the pupil of the eye and increasing the processing speed of the measured data.

Thus, the claimed solution, in addition to increasing the reliability of the results obtained, allows you to process data arrays in a shorter time, and therefore, using less hardware and software resources, in comparison with known methods.

The technical result of the invention is the ability to process large arrays of images with increasing reliability of the allocation of areas of fixation of the gaze. The claimed invention allows to detect precisely the data sections corresponding to the areas of gaze fixation, different from the rest of the array, and not the "edges" of these sections, which makes it possible to more likely determine the necessary fragments. In the case of aitracking, this is especially critical, since these events are just fixations and saccades. At the same time, the claimed solution meets the criterion of a greater speed of processing the data array and searching for relevant fragments corresponding to the fixations of the gaze.

The technical result is achieved by the fact that according to the method for determining the areas of gaze fixation during eye movement, at least one stimulus image is presented to the respondent with a fixed head position, the respondent's eye region is illuminated with infrared light during presentation of the stimulus image, the reflected signal is recorded to obtain a video sequence of pupil images eyes, process the resulting video sequence with determining the coordinates of the center of the pupil in each frame of the video sequence and subsequent formation in the biological sequence of a two-dimensional array of values of the coordinates of the center of the pupil corresponding to frames of a video sequence from which at least three points are distinguished by means of segmentation analysis, which are considered to be the area of fixation of the gaze.

For all identified fixation regions, their centers are additionally determined with subsequent calculation of the angular distance between their centers for neighboring regions and, if the angular distance is less than 1.5 °, the neighboring regions are combined into one with the determination of its center and the subsequent repetition of the described actions with respect to the remaining fixation regions . For all identified fixation regions, the boundary values of the fixation regions, which are considered the boundaries of saccades, are additionally determined. Registration of the reflected signal is carried out for at least 5 Ohm with obtaining at least 5 frames of video. Stimulus images are placed at the eye level of the respondent at a distance providing an angular distance between the centers of neighboring fixation areas of not more than 20-25 degrees. The coordinates of the pupil center are determined in the coordinate system constituting the video sequence of eye images recorded by the video camera (in the coordinate system of the video camera matrix), and the gaze coordinates in the stimulus image system or in the image coordinate system.

Stimulus images are placed at the eye level of the respondent at a distance of not less than 30 cm, which makes it possible to obtain a reflected image of the participant’s eye and calculate the coordinates of the gaze position according to the preliminary calibration using the coordinates of the pupil’s center on the camera’s matrix.

The technical result is achieved by the fact that the system for determining the areas of gaze fixation during eye movement includes an IR emitter placed with the ability to illuminate the respondent’s eye area, a reflector filter configured to transmit visible light and reflect IR radiation, installed at the respondent’s eye level, a stimulus image presentation unit located behind the reflector filter; a registration unit for infrared radiation reflected from the pupil in the form of a video sequence of images of the pupil of the eye containing at least at least five frames, including a video camera, and an image processing unit from a video camera, configured to process data through segmentation analysis and obtain areas of gaze fixation with eye movement.

The IR emitter is configured to emit at a wavelength of 860 nm, capturing a region of ± 40 nm. Such a range of radiation falls into the maximum spectral sensitivity of the video camera matrix, but remains invisible to the human visual analyzer. The IR emitter is placed with the possibility of illuminating the eye area with direct or reflected radiation from the filter-reflector. The reflector filter is made in the form of a warm mirror, which consists of 55-65 alternating layers of materials: TiO ₂ and SiO ₂ , and is located at a distance of no more than 10 cm from the respondent’s pupil. A warm mirror transmits visible light and reflects light in the infrared range . The video camera is equipped with a matrix having spectral sensitivity, the maximum of which lies in the region of 650-850 nm, in particular, characterized by the spectral sensitivity curve shown in FIG. one.

According to (Lemire, 2007; Bellman, Roth, 1969), storing dynamically changing data requires more accuracy (in bits) than for processing constant values. Therefore, to implement segmentation analysis, a 64-bit algorithm is used, not a 32-bit one. A linear algorithm compared with, for example, the Douglas-Pecker algorithm gives a gain in time, because the time required for the calculations grows linearly in this case (quadratically in the Douglas-Pecker algorithm).

A distinctive feature of the proposed solutions is the use of segmentation analysis to extract oculomotor events from the coordinate data of positioning the gaze in the image. These data are obtained by processing the video sequence of images of the pupil of the eye obtained from the video camera upon presentation of the stimulus material, as well as recalculating coordinates based on calibration data from the coordinate system of the matrix of the video camera into the coordinate system of the stimulus image presented on the monitor.

Brief Description of the Drawings

The invention is illustrated by drawings.

In FIG. 1 shows a spectral sensitivity curve of a video camera.

In FIG. 2 schematically presents the inventive system and method using an ittracker, including a conditional image of the respondent, a computer system unit that controls the operation of the software and hardware, presentation of the necessary visual stimuli on the monitor screen, and also implements the selection of events from the coordinate data array (using the segmentation algorithm ); emission filters (warm mirrors) are also marked on the diagram, with the help of which an image of the eye, a video camera and IR illumination, located outside the respondent’s visual area, and a monitor screen are displayed, on which stimulus images are presented to the respondent.

In FIG. 3 shows a block diagram of a segmentation analysis algorithm. An array of data (gaze position coordinates) is sent to the input, which is successively approximated by polynomials of varying degrees. Moreover, at each step, the accuracy of the approximation is estimated. At a certain iteration, the optimal form of the polynomial is determined by the criterion of minimal error (the error is minimal). For this polynomial, I realize the segmentation of the input array into fragments corresponding to gaze fixations. The algorithm is implemented on the basis of dynamic programming, which gives a gain in data processing speed.

In FIG. Figure 4 shows the optical scheme (movement of the infrared beam) used to record the areas of gaze fixation during eye movement. The diagram shows the projection of the image of the eye (and, accordingly, the pupil against the background of the iris) on the video camera matrix in the IR spectrum due to the illumination of the eye with IR illumination and reflection of the image from a warm mirror on the video camera matrix. Warm mirrors are indicated on the diagram, through which the respondent can freely see the image presented to him on the monitor screen, a video camera.

In FIG. Figure 5 shows an example of visualization of oculomotor events extracted from the initial data (fixations and saccades) plotted on the image considered by the respondent. The radii of the circles are greater, the longer the fixations of the gaze were. The image also shows the gaze position for each frame of the video sequence (each point corresponds to one frame). From the positions of the gaze in each frame, the positions of fixations are obtained by the method of segmentation analysis, for which the duration and coordinate of their center are determined.

In FIG. 6. The image of an eye with a pupil approximated (fitted) by an ellipse is presented. At the intersection of the axes of the ellipse, there is a key point for which the current coordinates are determined.

In FIG. 7 shows a representation of the series of coordinates of the position of the center of the pupil used for the segmentation analysis algorithm.

The positions in the drawings indicate: 1 - respondent, 2 - emission filters (warm mirrors), 3 - IR emitter (backlight), 4 - video camera, 5 - monitor screen, 6 - computer with module 7 installed on it that implements the claimed method, representing a memory block with software, 8 - a computer mouse, 9 - displaying images on a screen, 10 - IR rays, 11 - capturing a pupil’s video image.

The implementation of the invention

When describing the implementation of the invention, the following terms and definitions are used.

Fixation - stopping the gaze during oculomotor activity when viewing the image (physiological). From the point of view of mathematics, fixation is a fragment of an array of coordinates corresponding to the period of time when the gaze pauses when viewing an image.

The gaze fixation point is the place (coordinates) of the respondent's gaze positioning at the moment of fixation, i.e. gaze stop.

The fixation area is a set of points of gaze position and, accordingly, their coordinates characterizing the place of gaze fixation, i.e. its stops, as well as the corresponding region of space on the coordinate plane of the image considered on the monitor screen.

Saccade - a quick spasmodic movement of the eyeballs, moving the look from one subject to another (fiziol.). From the point of view of mathematical analysis, a saccade is a section of the trajectory (and the corresponding fragment of the array of data of coordinates of the gaze position) connecting adjacent fixations.

Stimulus image - the image presented on the screen or in real conditions, which the respondent is looking at.

A warm mirror is a term that denotes a filter that is transparent to visible light and reflects infrared radiation.

Segmentation analysis is the mathematical algorithm described in Lemire D. A better alternative to piecewise linear time series segmentation // Proceedings of the 2007 SIAM International Conference on Data Mining. - Society for Industrial and Applied Mathematics, 2007. - C. 545-550.

allowing to divide a continuous series of data into fragments taking into account optionally defined conditions.

1) Sequences (or series) are called data arrays of the following form:

(x ₀ , y ₀ ), ..., (x _n-1 , y _n-1 ), where the values of x are the coordinate values corresponding to the time samples.

2) Segmentation indices z ₀ , ..., z _k - marks dividing the sequence into intervals.

3) S ₁ , ..., S _k are the intervals determined by the segmentation indices S _i = {(x _i , y _i ) | z _i-1 ≤i <z _j }

Segmentation analysis includes:

Определения сегментационной ошибки по формуле:

Definitions of segmentation error by the formula:

, где Q - это квадрат регрессионной ошибки

.

where Q is the squared regression error

.

Поиск значений Q, которые вычисляются по формуле

(p(x_r)-y_r)², где минимум определяется по полиномиальному p заданной степени.

Search for Q values that are calculated by the formula

(p (x _r ) -y _r ) ² , where the minimum is determined by the polynomial p of a given degree.

The main task in assessing the allocation of segments and determining the error of segmentation analysis is the most suitable approximation (search for the best polynomial).

The calculation algorithm is a set of dynamic programming actions for the optimal search for segments that make up the general data array, during which the following actions are implemented:

The first part of the dynamic programming algorithm for finding optimal segmentation of time series into intervals having degrees 0, ..., N-1.

1. INPUT: temporary series (x _i , y _i ) of length n.

2. INPUT: model complexity k and maximum degree N (N = 2 => constant and linear).

3. INPUT: Function E (p, q, d) that calculates the fitting error of a polynomial of degree d in the series (x _p , x _q ) (constant time).

4. R, D, P ← k * n + 1 matrices (initialized to 0).

5. for r∈ {0, ..., k-1} do

6. {r scans rows of matrices}

7. for q∈ {0, ..., n} do

8. {q scans matrix columns}

9. Finding the minimum R _{r-1-d, P} + E (p, q, d) and storing its values in R _{r, q} and the corresponding d, p in two steps in D _{r, q} , P _{r, q} for 0 <d <min (r + 1, N) and 0 <p <q + 1, provided that R → ∞ for negative series except R _-1,0 = 0

10. RETURN The value of the matrix R, the degree of the matrix D, the segmentation points of the matrix P.

The second part of the dynamic programming algorithm for finding optimal segmentation.

1. INPUT: k * n + 1 matrices R, D, P.

2.x ← n

3.s ← empty list

4. while r> 0 do

5. p ← Pr, x

6.d ← Dr, x

7. r ← r-d + 1

8. incremental interval from p to x of degree d to s

9.x ← p

10. RETURN Optimal segmentation s.

Thus, the algorithm extracts the optimal fragments (segments) of the numerical series of coordinates from the entire sequence characterized by a certain time period.

The following is a detailed description of the method and system for determining the areas of gaze fixation during eye movement. To confirm the conformity of the proposed solution to the requirement of "industrial applicability", as well as to better understand the essence of the invention, examples of its specific implementation are presented below, which do not limit the claimed invention.

The respondent is asked to sit in a chair and fix his head on the frontal-chin stand opposite the monitor screen. The monitor screen is placed at a comfortable vision distance from the respondent's eyes (40-70 cm), although the technical capabilities of the system do not limit this indicator to the stated limits. Then the respondent is asked to undergo a calibration, which is a presentation of a set of points on the screen with previously known coordinates. This procedure is necessary to obtain reference values for recalculating the coordinates of the positions of the gaze while viewing any other images during the experiment. Presentation, registration and processing of data (both presentation of calibration and stimulus images, as well as registration of the coordinates of the pupil's position) can be implemented using available software tools with functionality for presenting and recording measurement results in files. Also, the algorithm of segmentation analysis should be implemented in the software in order to isolate the fixations and the corresponding saccades for subsequent visualization of these results (events) by applying them to the presented image.

In the process of presenting a stimulus image, the respondent's eye area is illuminated with infrared light. The IR emitter is located next to a digital camera that records the reflected image of the eye. IR rays, reflected from a warm mirror, illuminate the respondent's eye, the reverse beam projectes the resulting image onto the video camera matrix. A product may be used as a warm mirror, for example, see http://www.photooptic-filters.com/hot_mirrors/. IR radiation allows you to achieve maximum contrast of the pupil image relative to the iris, which increases the likelihood of correct pupil allocation used in the program image extraction algorithms.

In the process of viewing the stimulus image, the image of the respondent’s eye is recorded in the form of a reflected infrared signal to obtain a video sequence of images of the pupil of the eye, and the resulting video sequence is processed with the coordinates of the center of the pupil in each frame of the video sequence. IR radiation after double reflection from a warm mirror (the first path IR radiation passes from the emitter and gets into the eye, the second, reflected from the eye, returns to the camera lens, which is located next to the IR emitter. Reflected IR radiation transfers the image of the eye to the matrix of the video camera that registers the given image. A pupil is selected from each frame of this video sequence and the coordinates of its center are determined. Next, these coordinates are converted into the coordinate system of the image that the respondent was looking at, and Fixation by the method of segmentation analysis is distinguished from these coordinates.

The pupil extraction algorithm (Kim SI et al. A fast center of pupil detection algorithm for VOG-based eye movement tracking // Engineering in Medicine and Biology Society, 2005. IEEE-EMBS 2005. 27th Annual International Conference of the. - IEEE, 2006 . - C. 3188-3191 .; Zhu D., Moore S.T., Raphan T. Robust pupil center detection using a curvature algorithm // Computer methods and programs in biomedicine. - 1999. - T. 59. - No. 3. - C. 145-157.) From the image, it analyzes each frame and searches for a relevant area determined by a number of criteria (the darkest image, elliptical shape) and approximates the pupil with an ellipse, while determining its center. The coordinates calculated for the center of the ellipse are recalculated from the coordinate system of the matrix of the camera into the coordinate system of the image (see Fig. 6). Coordinates can be uploaded to a table, for example, in the form:

From this array of coordinates, the segments corresponding to gaze stops - fixations are extracted by the method of segmentation analysis. The visualization of these sections can be presented in the form shown in FIG. 7.

The results of the segmentation algorithm allow us to select sections (segments) of a fixed coordinate in an optimal way. After the implementation of all the blocks marked in the figure, an array of segment values corresponding to gaze fixations is found, for them (subtraction), the durations corresponding to these fixations and the center coordinates of these fixations are determined, forming the final output array of values.

The final table contains the characteristics of fixations and saccades (including the coordinates of the centers of fixations). To implement the proposed method, software can be used that allows you to visualize the result obtained by applying to the image that the respondent examined, visualization of fixations simultaneously with the coordinates of the gaze positions in the image coordinate system (see Fig. 5).

To confirm the achievement of the claimed technical result, a model experiment was conducted. The respondent himself noted which elements of the image he was looking at, as well as in what sequence. Simultaneously, an analysis was carried out on the basis of an expert method for evaluating video images of the eye image, as well as in comparison with recording on the SMI 250Hz system. It was found that the number of undetected fixations in the SMI 250Hz system is higher compared to the claimed solution using segmentation analysis (2 lost fixations on the recording with a duration of 5 minutes for SMI 250Hz compared to all highlighted on the claimed IT tracker).

A model experiment was also carried out in which a model of an artificial eye fixed motionless was used. Recording was carried out for one minute (60 sec.) Followed by analysis of the scatter of the fixation points of the gaze. It was found that the stability of gaze fixation under such conditions is very high: the error in the position of gaze fixation in such a model experiment was ± 1 pcl, which corresponds to a relative error of 0.01%. This is evidence of the accuracy of both the hardware and the accuracy of the algorithm for extracting fixations from the source data.

Thus, an advantage of the present invention is the ability to obtain data on fixations by the minimum number of points in the "cloud" (ie, the grouping of points corresponding to fixation) - starting from three detected through the use of segmentation analysis. At the same time, from the point of view of physiology, the fixation parameters should be from 50 ms and higher. In case the perception of any complex cognitive information is studied, the parameters of fixation should be at least 60-80 ms. The algorithm for detecting fixations by a threshold level (speed, coordinate) is also characterized by high speed, but when using it there is a problem with choosing a threshold (what can be considered a “cut-off”, according to which it can be judged that the speed has exceeded a critical level and then the event is already a saccade. it follows that an algorithm working on this principle is largely fast, but suffers from the presence of errors of the first and second kind in high dependence on the chosen threshold level, which, if you choose Xia expert, suffers subjectivity, and the automatic algorithm can not just pick up the marked parameter. That is, when using such an algorithm is often detected in the fixing of those places where they do not exist and, on the contrary, are not allocated fixation, which objectively exist.

Compared with the method based on the use of dispersion, the inventive method using segmentation analysis is implemented faster in speed, as described above, and therefore saves computer resources when obtaining the same speed results with less hardware power. The basis of this approach is dynamic programming, which allows you to divide a large task into sub-tasks of a lower level and implement calculations with high speed. For other algorithms (for example, for dispersion), resource consumption (i.e., the use of hardware capacities, such as RAM, processor resources, etc.) grows according to a quadratic law depending on the number of coordinates (points) that make up the input data array.

Claims (12)

1. A method for determining the areas of gaze fixation during eye movement, characterized in that at least one stimulus image is presented to the respondent with a fixed head position, the respondent's eye region is illuminated with infrared radiation, and the reflected signal is recorded for at least 50 ms with obtaining at least 5 frames of the video sequence of images of the pupil of the eye, process the resulting video sequence with the coordinates of the center of the pupil in each frame of the video sequence, after which coordinates converted into the stimulus image coordinate system, form a two-dimensional array of the pupil center coordinate value in chronological sequence corresponding to the video sequence frames from which segmentation analysis is isolated by at least three points, that said area of fixation. 2. The method according to p. 1, characterized in that for all identified areas of gaze fixation, their centers are additionally determined - gaze fixation points, with subsequent calculation for adjacent regions of the angular distance between their centers and if the angular distance is less than 1.5 °, the neighboring areas are combined into one with the definition of its center and the subsequent repetition of the described actions in relation to the remaining areas of fixation. 3. The method according to claim 1, characterized in that for all identified areas of fixation of the gaze, additionally determine the boundary values of the areas of fixation, which determine the boundaries of saccades. 4. The method according to p. 1, characterized in that the stimulus images are placed at eye level of the respondent at a distance with an angular distance between the centers of neighboring fixation areas of not more than 20-25 degrees. 5. The method according to p. 1, characterized in that the coordinates of the center of the pupil are determined in the coordinate system of the stimulus images. 6. The method according to p. 1, characterized in that the coordinates of the center of the pupil are determined in the coordinate system of the images that make up the video image of the eye. 7. A system for determining the areas of gaze fixation during eye movement, including an IR emitter placed to illuminate the respondent’s eye area, a reflector filter configured to transmit visible light and reflect IR radiation, mounted at the respondent’s eye level, a stimulus presentation unit image located behind the reflective filter, the registration unit of the infrared radiation reflected from the pupil in the form of a video sequence of images of the pupil of the eye containing at least five frames, including a video camera Yeru, and an image processing unit with a video camera configured to data processing by the segmentation analysis and yield regions when eye gaze fixation movement. 8. The system according to p. 7, characterized in that the infrared emitter is placed with the possibility of illuminating the eye area directly or reflected from the reflector filter by infrared radiation. 9. The system according to claim 7, characterized in that the filter reflector is made in the form of a warm mirror that transmits visible light and reflects light in the infrared range, which includes layers of alternating materials: TiO ₂ and SiO ₂ . 10. The system according to claim 7, characterized in that the video camera is equipped with a matrix having spectral sensitivity, the maximum of which lies in the region of 650-850 nm. 11. The system according to claim 7, characterized in that the filter reflector is located from the pupil at a distance of not more than 10 cm 12. The system according to claim 7, characterized in that the infrared emitter is configured to emit at a wavelength of 860 nm, capturing a region of ± 40 nm.

Images