CN104899565A

CN104899565A - Eye movement track recognition method and device apparatus based on textural features

Info

Publication number: CN104899565A
Application number: CN201510293913.1A
Authority: CN
Inventors: 张成岗; 李春永; 岳敬伟; 屈武斌
Original assignee: Beijing Yunyi International Technology Co ltd; Institute of Radiation Medicine of CAMMS
Current assignee: Beijing Yunyi International Technology Co ltd; Institute of Radiation Medicine of CAMMS
Priority date: 2015-06-01
Filing date: 2015-06-01
Publication date: 2015-09-09
Anticipated expiration: 2035-06-01
Also published as: CN104899565B

Abstract

The invention provides an eye movement track recognition method and apparatus based on textural features. By acquiring an original eye movement track diagram recorded by an eye tracker and extracting features of the original eye movement track diagram, the extracted features are input into a classifier; a recognition result is acquired; and by carrying out recognition on the original eye movement track diagram, texture details of an eye movement track are more abundant, so that the recognition accuracy rate is improved.

Description

Eye movement track identification method and device based on texture features

Technical Field

The invention relates to an image processing technology, in particular to an eye movement track identification method and device based on texture features.

Background

Biometric identification is a technique of performing identity authentication by using a characteristic inherent to a human body through a computer. The eye movement track recognition technology is widely applied to unique psychological mechanisms of identity authentication and personal visual information processing.

In the prior art, the eye movement track identification method is a method for identifying based on the characteristics of a fixation point and an eye jump, namely, the biological characteristics of a complex eye movement mode, which are extracted based on a fixation point and an eye jump trajectory diagram, the fixation point and the eye jump trajectory graph are drawn by the fixation point and the eye jump data, so that the obtained fixation point and the eye jump trajectory are very sparse, wherein, the eye tracker acquires the original data of the eye tracking diagram, the sampling rate of the eye tracker is assumed to be 300HZ, the fixation point is a point at which the interval time between the current sampling point and the next sampling point is more than or equal to a certain preset threshold value, the threshold is typically 200ms, eye jump refers to a fast moving point between two fixation points, because the fixation point and the eye jump are artificially divided based on the original eye movement data, and the obtained fixation point and the eye jump track are very sparse, the identification accuracy is low.

Disclosure of Invention

According to the eye movement track identification method and device based on the texture features, the original eye movement track graph is identified, so that the texture details are richer, and the identification accuracy is improved.

The invention provides an eye movement track identification method based on texture features, which comprises the following steps: acquiring N original eye movement track graphs recorded by an eye movement instrument, wherein N is an integer greater than or equal to 1; extracting the characteristics of the N original eye movement track graphs; and inputting the characteristics of the N original eye movement track graphs into a classifier to obtain an identification result.

The characteristics of the N original eye movement track maps comprise the characteristics of M samples to be identified; the extracting the features of the N original eye movement track diagrams comprises the following steps: combining the N original eye movement track graphs into M samples to be identified, wherein each sample to be identified comprises L multiplied by L original eye movement track graphs, L is an integer which is greater than or equal to 1, M is an integer which is greater than or equal to 1, and the product of L multiplied by M is less than or equal to N; extracting features of each of the samples to be identified.

The combining the N original eye movement trajectory diagrams into M samples to be recognized includes: determining L multiplied by M original eye movement track graphs from the N original eye movement track graphs; and combining the L multiplied by M original eye movement track maps into M samples to be identified according to an L multiplied by L distribution mode.

The extracting features of each sample to be identified comprises: carrying out Gabor transformation on each sample to be identified; extracting the characteristics of each sample to be identified after Gabor transformation.

Performing Gabor transformation on each sample to be identified, including: converting each sample to be identified into a corresponding two-dimensional matrix; carrying out binarization on each two-dimensional matrix; using Gabor transformation functions with different frequencies f and directions theta to respectively carry out two-dimensional convolution operation on each two-dimensional matrix after binarization to obtain f multiplied by theta result matrixes, wherein the Gabor transformation functions are

Wherein,

f is the frequency of the sine curve in the Gabor transformation function, theta is the direction of the Gabor transformation function, phi is the phase difference and is the standard deviation of the Gaussian function, and gamma is the space proportionality constant.

The extracting the features of each sample to be identified after Gabor transformation comprises the following steps: converting the f multiplied by theta result matrixes into f multiplied by theta one-dimensional vectors, respectively solving the mean and the variance of the f multiplied by theta one-dimensional vectors to obtain f multiplied by theta mean values and f multiplied by theta variances, and taking the f multiplied by theta mean values and/or the f multiplied by theta variances as the characteristics of the extracted sample to be identified.

Inputting the characteristics of the N original eye movement locus diagrams into a classifier to obtain a recognition result, wherein the recognition result comprises the following steps: randomly extracting M feature vectors from M feature vectors of M to-be-recognized samples of the N original eye movement locus diagrams as training samples, using the (M-M) feature vectors as test samples, training a classifier by using the training samples, inputting the features of the test samples into the trained classifier, and acquiring a recognition result, wherein one to-be-recognized sample corresponds to one feature vector, and M is an integer greater than or equal to 1.

The invention provides an eye movement track recognition device based on texture features, which comprises: the system comprises an acquisition module, a display module and a control module, wherein the acquisition module is used for acquiring N original eye movement track graphs recorded by an eye movement instrument, and N is an integer greater than or equal to 1; the characteristic extraction module is used for extracting the characteristics of the N original eye movement track maps; and the recognition module is used for inputting the characteristics of the N original eye movement track maps into the classifier and acquiring a recognition result.

The feature extraction module includes: the image processing unit is used for combining the N original eye movement track maps into M samples to be identified, wherein each sample to be identified comprises L multiplied by L original eye movement track maps, L is an integer which is greater than or equal to 1, M is an integer which is greater than or equal to 1, and the product of L multiplied by M is less than or equal to N; and the extraction unit is used for extracting the characteristics of each sample to be identified.

The picture processing unit is specifically configured to determine L × M original eye movement trajectory diagrams from the N original eye movement trajectory diagrams; and combining the L multiplied by M original eye movement track maps into M samples to be identified according to an L multiplied by L distribution mode.

According to the eye movement track identification method and device based on the texture features, the original eye movement track graph recorded by the eye movement instrument is obtained, the features of the original eye movement track graph are extracted, the extracted features are input into the classifier, the identification effect is obtained, the texture details are richer through identification on the original eye movement track graph, and the identification accuracy is improved.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.

FIG. 1 is a flowchart of a first embodiment of an eye movement trajectory identification method based on texture features according to the present invention;

FIG. 2 is a flowchart illustrating a second embodiment of the eye movement trajectory identification method based on texture features according to the present invention;

FIG. 3A is a sample to be identified after combining original eye movement trajectory graphs of a digital search test according to an embodiment of the present invention;

FIG. 3B is a sample to be identified after combining original eye-movement traces of a mind rotation test according to another embodiment of the present invention;

FIG. 4 is a flowchart of a third embodiment of the eye movement trajectory identification method based on texture features;

FIG. 5 is a detailed waveform diagram of a Gabor transform function;

FIG. 6 is a graph showing the difference between the recognition accuracy rates corresponding to different frequency eigenvalues;

FIG. 7 is a schematic structural diagram of a first embodiment of an eye movement trajectory recognition device based on texture features according to the present invention;

fig. 8 is a schematic structural diagram of a second embodiment of the eye movement trajectory recognition device based on texture features.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

Fig. 1 is a flowchart of a first embodiment of an eye movement trajectory identification method based on texture features, as shown in fig. 1, the method of this embodiment may include:

step 101: n original eye movement track graphs recorded by the eye movement instrument are obtained, wherein N is an integer larger than or equal to 1.

The method comprises the steps of firstly obtaining original data of an eye movement track graph, wherein the sampling rate of an eye movement instrument used for collecting eye movement data is 300HZ, and if the eye movement data is used for drawing, the obtained fixation point track is very sparse, so that most of texture features of visual information processing are lost. Since we analyze the gaze track of the subject as a texture, the richer the texture details are, the more the search features of the subject can be reflected, and if the gaze track is plotted with the original data of 300Hz, the more the texture details are, therefore, the eye movement track graph made by the original data will be used in this embodiment. And drawing by using the data acquired by the eye tracker in each test to obtain N original eye movement track pictures, wherein N is an integer greater than or equal to 1.

In the first embodiment of the invention, because the texture features are extracted, the original data acquired by the eye tracker is used for drawing the eye movement locus diagram, so that more visual information can be reflected. In addition, the 3 × 3 combination is preferably adopted in the embodiment of the present invention, because of the limitation of the number of experimental data.

Step 102: and extracting the characteristics of the N original eye movement track graphs.

Feature extraction is to use a computer to extract image information and determine whether a point of each image belongs to an image feature. The result of feature extraction is to divide the points on the image into different subsets, which often belong to isolated points, continuous curves or continuous regions. Common image features include color features, texture features, shape features, and spatial relationship features.

The invention relates to an eye movement track identification method based on texture characteristics, wherein the texture characteristics are global characteristics and describe surface properties of scenes corresponding to images or image areas. Unlike color features, texture features are not based on the characteristics of the pixel points, which requires statistical calculations in regions containing multiple pixel points. In pattern matching, such regional features have great superiority, and matching is not unsuccessful due to local deviation. As a statistical feature, the texture feature often has rotation invariance and is resistant to noise.

In this step, the features of the N original eye-movement trajectory diagrams in step 101 are extracted for identification. The feature may be a mean, variance, or other statistical data.

Step 103: and inputting the characteristics of the N original eye movement track graphs into a classifier to obtain an identification result.

Inputting the features extracted in the step 102 into a trained classifier to obtain a recognition result; if the classifier aiming at the original eye movement track graph does not exist, the classifier is trained, the features extracted in the step 102 are input into the classifier for training, and then the picture to be tested is input for recognition. The classifier can be a weighted Euclidean distance, probability density estimation and support vector machine. In the embodiment, a support vector machine is used as a classifier, belongs to a generalized linear classifier, and is used for dividing a sample into two types by constructing a hyperplane, simultaneously realizing minimized empirical errors and maximized geometric marginal areas, and having high identification accuracy.

According to the eye movement track identification method based on the texture features, the original eye movement track graph recorded by the eye movement instrument is obtained, the features of the original eye movement track graph are extracted, the extracted features are input into the classifier, the identification effect is obtained, the texture details are richer by identifying on the original eye movement track graph, and the identification accuracy is improved.

Fig. 2 is a flowchart of a second embodiment of the eye movement trajectory identification method based on texture features, as shown in fig. 2, the method of this embodiment may include:

step 201: and acquiring N original eye movement track graphs recorded by the eye tracker.

This step is the same as the step method of the first embodiment, and is not described herein again.

Step 202: combining N original eye movement track graphs into M samples to be identified, wherein each sample to be identified comprises L multiplied by L original eye movement track graphs, L is an integer which is larger than or equal to 1, M is an integer which is larger than or equal to 1, and the product of L multiplied by M is smaller than or equal to N.

In this step, the features of the N original eye movement trajectory diagrams include features of M samples to be recognized, specifically, the N original eye movement trajectory diagrams are combined into M sample pictures to be recognized, each sample picture to be recognized includes L × L original eye movement trajectory diagrams, for example, the samples to be recognized include 2 × 2,3 × 3, and 4 × 4 original eye movement trajectory diagrams, when L is 1, the flow is shown in the implementation method of the first embodiment, at this time, the N original eye movement trajectory diagrams are combined into M samples to be recognized, and the calculation method of M is N% (L × L), that is, an integer obtained by remaining N pairs (L × L).

Step 203: features of each sample to be identified are extracted.

And after the samples to be identified are combined, extracting the characteristics of the samples to be identified. The extracted features are the same as in step 102 and will not be described here.

Step 204: and inputting the characteristics of the sample to be identified into the classifier, and acquiring an identification result.

Similar to step 103, the difference is that the original trajectory graphs in step 103 are combined into a sample to be identified, and after the characteristics of the sample are extracted, the sample is input into a classifier to obtain an identification result.

Fig. 3A is a sample to be recognized after combining original eye movement locus diagrams of a digital search test according to an application example of the present invention, and fig. 3B is a sample to be recognized after combining original eye movement locus diagrams of a mind rotation test according to another application example of the present invention, where the combined sample to be recognized is shown in fig. 3A and fig. 3B, respectively.

Specifically, combining N original eye movement trajectory diagrams into M samples to be recognized includes: determining L multiplied by M original eye movement track graphs from the N original eye movement track graphs; and combining the L multiplied by M original eye movement track maps into M samples to be identified according to the L multiplied by L distribution mode. The method for determining the L × M original eye movement trajectory diagrams may adopt a random extraction method, or may be selected manually, for example, 40 original eye movement trajectory diagrams are originally provided, 36 pictures are randomly extracted from the original eye movement trajectory diagrams to form 4 samples to be identified, each sample to be identified is arranged according to the distribution mode of the horizontal 3 original eye movement trajectory diagrams and the vertical 3 original eye movement trajectory diagrams, and the samples to be identified after the original eye movement trajectory diagrams of different application examples are combined are respectively shown in fig. 3A and fig. 3B. The digital search test comprises a randomly generated 7-digit user number and 10 groups of winning numbers, and the user needs to match the user number with the 10 groups of winning numbers and select several prizes in the user number. Psychological rotation is a space representation power conversion capability for imagining self or object rotation, and is also an important scale for evaluating space intelligence. The adopted test mode is to judge whether two three-dimensional figures are overlapped through rotation of a certain angle, if the two three-dimensional figures can be overlapped, the same result is selected, and if the two three-dimensional figures can not be overlapped, the different result is selected.

The gaze trajectories of a single digital search task under test may not have a certain stable characteristic due to a large difference, but the gaze trajectories of a plurality of digital search tasks under test combined together may have a stable characteristic that reflects a gaze characteristic under test. The eye movement track identification method of the embodiment adopts a mode of combining the eye movement track diagrams for multiple times for identification, and has stable characteristics, so that the identification result is more stable and accurate.

Further, extracting features of each sample to be identified comprises: carrying out Gabor transformation on each sample to be identified; extracting the characteristics of each sample to be identified after Gabor transformation.

The Gabor transform belongs to windowed Fourier transform, and the Gabor function can extract related features in different scales and different directions of a frequency domain. In addition, the Gabor function is similar to the biological action of human eyes, so that the Gabor function is often used for texture recognition and achieves better effect. And converting the result obtained by Gabor transformation into a one-dimensional vector according to the result obtained by Gabor transformation, and then obtaining a mean value and a variance.

Fig. 4 is a flowchart of a third embodiment of the eye movement trajectory identification method based on texture features, and as shown in fig. 4, the performing Gabor transform on each sample to be identified includes:

step 301: and converting each sample to be identified into a corresponding two-dimensional matrix.

The sample to be identified is represented as a two-dimensional matrix, which means that the gray value of each pixel in the sample image to be identified is represented as one value of the matrix.

Step 302: and carrying out binarization on each two-dimensional matrix.

The binarization of the two-dimensional matrix means that all values in the two-dimensional matrix are replaced by 0 or 1 according to a certain rule, that is, the element values in the two-dimensional matrix are set to be 1 if the element values are greater than 127 and set to be 0 if the element values are less than 127, and the values of 0 and 1 are inverted to reduce the calculation amount of the Gabor transform.

Step 303: using Gabor transformation functions with different frequencies f and directions theta to respectively carry out two-dimensional convolution operation on each two-dimensional matrix after binarization to obtain f multiplied by theta result matrixes, wherein the Gabor transformation function is

Wherein,

In this embodiment, the two-dimensional matrix data is processed by using 40 kinds of Gabor transform functions of 5 × 8, i.e. 5 different frequencies and 8 different directions, wherein the 5 different frequencies areThe 8 different directions are [1,2,3,4,5,6,7 ]]8 × π. The other is 15 × 8, that is, 120 Gabor transform functions with 15 different frequencies and 8 different directions are used to process two-dimensional matrix data, wherein the 15 different frequencies areThe 8 different directions are [1,2,3,4,5,6,7 ]]8 × π. The embodiment adopts two-dimensional convolution operation, namely windowed Fourier transform, the window function is a Gaussian function, and the method can extract relevant characteristics in different scales and different directions of a frequency domain, is similar to the biological action of human eyes, and has a good effect on texture recognition.

Further, extracting the features of each sample to be identified after Gabor transformation includes: and converting the f multiplied by theta result matrixes into f multiplied by theta one-dimensional vectors, respectively solving the mean and the variance of the f multiplied by theta one-dimensional vectors to obtain f multiplied by theta mean values and f multiplied by theta variances, and taking the f multiplied by theta mean values and/or the f multiplied by theta variances as the characteristics of the extracted sample to be identified.

Specifically, the Gabor is converted into a one-dimensional vector according to a result obtained by Gabor conversion, and then the mean and the variance of the one-dimensional vector are obtained and used as the features of a sample to be identified, and all the features can be used when a classifier is used for classification, namely, the mean and the variance are both used as feature values, or any one of the features can be used, namely, only the mean or only the variance is used. A set of frequency f and direction θ parameters of the Gabor transform function are transformed once corresponding to each other to obtain a two-dimensional matrix result, so that the dimension of a feature vector obtained by a sample to be identified is as follows: frequency (f) group number x direction (theta) group number x feature vector number (mean, variance), one sample to be identified corresponds to one feature vector.

Fig. 5 is a specific waveform diagram of a Gabor transform function, and as shown in fig. 5, the 5 kinds of Gabor transforms with frequencies and 8 directions (5 × 8) of 39 × 39 are performed on a sample to be identified to obtain 40 result matrices, and a mean value and a variance are calculated for the obtained result matrices to serve as eigenvalues of one-time transform. In an embodiment provided by the present invention, two feature values, i.e., a mean value and a variance, are extracted, and 5 × 8 Gabor transformations are adopted, so that the dimension of the obtained feature vector is 5 × 8 × 2 — 80.

In the embodiment, the mean value and the variance of a result matrix obtained by Gabor transformation of different frequencies and directions are adopted for one sample to be identified and used as the characteristics of one extracted characteristic vector, and one characteristic vector is composed of a plurality of characteristic values and can reflect rich and stable characteristics of the sample to be identified, so that a foundation is laid for subsequent identification accuracy.

Further, on the basis of the third embodiment, inputting the features of the N original eye movement trajectory diagrams into the classifier, and acquiring the recognition result, including: randomly extracting M feature vectors as training samples from M feature vectors of M to-be-recognized samples of N original eye movement locus diagrams, using the (M-M) feature vectors as test samples, training a classifier by using the training samples, inputting the features of the test samples into the trained classifier, and obtaining a recognition result, wherein one to-be-recognized sample corresponds to one feature vector, and M is an integer greater than or equal to 1.

Specifically, classifying the extracted features using a classifier includes: randomly extracting a certain proportion from the feature vectors of a plurality of samples to be identified of each person to be used as training samples, and using the rest of the training samples as test samples; training the classifier by using the training samples; and identifying the test samples by using a classifier, wherein one sample to be identified corresponds to one feature vector. In one embodiment of the present invention, a person has 4 feature vectors, so we randomly draw one feature vector as a test sample and the other 3 feature vectors are training samples.

The data used in this application example is from 23 subjects of digital search test, each subject is subjected to 40 times of digital search test, and each 9 times of test is combined into one sample to be identified of 3 × 3, so that 23 × 4 samples to be identified are totally obtained, and 23 × 4 feature vectors are obtained. One of the 4 feature vectors obtained from each test is randomly extracted as a test sample, the other 3 feature vectors are extracted as training samples, the 23 extracted feature vectors are classified by using a support vector machine method, the classes corresponding to the feature vectors corresponding to the 23 test are respectively 1,2 and 3 … … 23, and the classes can be classified into the corresponding test classes to be correct. The classification accuracy of 20 repeated classification tests is shown in table 1:

TABLE 1

0.695652	0.73913	0.782609	0.73913	0.826087
					0.73913	0.782609	0.652174	0.652174	0.695652
0.869565	0.695652	0.695652	0.826087	0.826087
					0.652174	0.695652	0.869565	0.913043	0.608696

Table 1 mean correct rate for classification: 0.7478.

in another example of an application, we have attempted to identify an eye trajectory graph for a cardiac rotation test. The same Gabor transformation parameters were used, and the classification accuracy is shown in table 2:

TABLE 2

0.565217	0.478261	0.565217	0.565217	0.608696
					0.695652	0.695652	0.434783	0.521739	0.565217
0.565217	0.521739	0.478261	0.695652	0.73913
					0.608696	0.608696	0.521739	0.565217	0.565217

The average recognition accuracy is: 0.5783.

in another set of tests, we tried another set of Gabor transform parameters, namely Gabor transforms of 15 frequencies and 8 directions (15 × 8) on the sample to be identified, resulting in a feature vector dimension of 15 × 8 × 2 — 240. The test was also repeated 20 times, and the recognition results are shown in table 3:

TABLE 3

0.789474	0.894737	0.842105	0.894737	0.894737
					0.842105	0.894737	0.842105	0.894737	0.789474
0.789474	0.842105	0.842105	0.894737	0.947368
					0.789474	0.947368	0.789474	0.789474	1

The average accuracy of classification is: 0.8605.

at the last classification, the classification accuracy reaches 1, which indicates that all test samples are correctly classified as the extracted feature value increases. Therefore, the eye movement trajectory identification method of the embodiment has higher identification accuracy for the test samples of different application examples, and the identification accuracy is increased along with the increase of the feature value of the extracted sample.

Further, the influence of the feature values of different frequencies on the classification result is different, and fig. 6 is a difference value of the recognition accuracy corresponding to the feature values of different frequencies, as shown in fig. 6, which is mainly used to illustrate that the feature values of different frequencies contribute differently to the recognition accuracy. The difference in recognition accuracy in the Y-axis is the recognition accuracy for classification using all frequency features minus the accuracy for classification using the absence of one frequency feature. The value of Y is greater than 0, which indicates that the classification effect is better with the frequency characteristic value than without the frequency characteristic value. In addition, the accuracy of the test sample size and eye movement data may also affect the classification results.

Fig. 7 is a schematic structural diagram of a first embodiment of an eye movement trajectory recognition device based on texture features, as shown in fig. 7, the device includes:

an obtaining module 41, configured to obtain N original eye movement trajectory diagrams recorded by an eye movement instrument, where N is an integer greater than or equal to 1;

the feature extraction module 42 is configured to extract features of the N original eye movement trajectory diagrams;

and the recognition module 43 is configured to input the features of the N original eye movement trajectory diagrams into the classifier, and obtain a recognition result.

The apparatus of this embodiment may be used to implement the technical solution of the method embodiment shown in fig. 1, and the implementation principle and the technical effect are similar, which are not described herein again.

Fig. 8 is a schematic structural diagram of a second embodiment of the eye movement trajectory recognition device based on texture features, and as shown in fig. 8, the feature extraction module 42 includes:

the picture processing unit 421 is configured to combine N original eye movement trajectory diagrams into M samples to be recognized, where each sample to be recognized includes L × L original eye movement trajectory diagrams, L is an integer greater than or equal to 1, M is an integer greater than or equal to 1, and a product of L × M is less than or equal to N;

an extracting unit 422, configured to extract features of each sample to be identified.

The apparatus of this embodiment may be used to implement the technical solution of the method embodiment shown in fig. 2, and the implementation principle and the technical effect are similar, which are not described herein again.

Further, the picture processing unit 421 is specifically configured to determine L × M original eye movement trajectory diagrams from the N original eye movement trajectory diagrams; and combining the L multiplied by M original eye movement track maps into M samples to be identified according to the L multiplied by L distribution mode. The implementation principle is similar to the corresponding method, and the details are not repeated here.

Those of ordinary skill in the art will understand that: all or a portion of the steps of implementing the above-described method embodiments may be performed by hardware associated with program instructions. The program may be stored in a computer-readable storage medium. When executed, the program performs steps comprising the method embodiments described above; and the aforementioned storage medium includes: various media that can store program codes, such as ROM, RAM, magnetic or optical disks.

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An eye movement track identification method based on texture features is characterized by comprising the following steps:

acquiring N original eye movement track graphs recorded by an eye movement instrument, wherein N is an integer greater than or equal to 1;

extracting the characteristics of the N original eye movement track graphs;

and inputting the characteristics of the N original eye movement track graphs into a classifier to obtain an identification result.

2. The method according to claim 1, wherein the features of the N original eye trajectory maps comprise features of M samples to be identified;

the extracting the features of the N original eye movement track diagrams comprises the following steps:

combining the N original eye movement track graphs into M samples to be identified, wherein each sample to be identified comprises L multiplied by L original eye movement track graphs, L is an integer which is greater than or equal to 1, M is an integer which is greater than or equal to 1, and the product of L multiplied by M is less than or equal to N;

extracting features of each of the samples to be identified.

3. The method of claim 2, wherein said combining the N original eye movement trajectories into M samples to be identified comprises:

determining L multiplied by M original eye movement track graphs from the N original eye movement track graphs;

and combining the L multiplied by M original eye movement track maps into M samples to be identified according to an L multiplied by L distribution mode.

4. The method of claim 2, wherein said extracting features of each of said samples to be identified comprises:

carrying out Gabor transformation on each sample to be identified;

extracting the characteristics of each sample to be identified after Gabor transformation.

5. The method of claim 4, wherein said Gabor transforming each of the samples to be identified comprises:

converting each sample to be identified into a corresponding two-dimensional matrix;

carrying out binarization on each two-dimensional matrix;

using Gabor transformation functions with different frequencies f and directions theta to respectively carry out two-dimensional convolution operation on each two-dimensional matrix after binarization to obtain f multiplied by theta result matrixesSaid Gabor transform function is

Wherein,

6. The method according to claim 5, wherein the extracting the features of each sample to be identified after Gabor transformation comprises:

converting the f multiplied by theta result matrixes into f multiplied by theta one-dimensional vectors, respectively solving the mean and the variance of the f multiplied by theta one-dimensional vectors to obtain f multiplied by theta mean values and f multiplied by theta variances, and taking the f multiplied by theta mean values and/or the f multiplied by theta variances as the characteristics of the extracted sample to be identified.

7. The method of claim 6, wherein inputting the features of the N original eye movement trajectory diagrams into a classifier to obtain a recognition result comprises:

randomly extracting M feature vectors from M feature vectors of M to-be-recognized samples of the N original eye movement locus diagrams as training samples, using the (M-M) feature vectors as test samples, training a classifier by using the training samples, inputting the features of the test samples into the trained classifier, and acquiring a recognition result, wherein one to-be-recognized sample corresponds to one feature vector, and M is an integer greater than or equal to 1.

8. An eye movement track recognition device based on texture features, comprising:

the system comprises an acquisition module, a display module and a control module, wherein the acquisition module is used for acquiring N original eye movement track graphs recorded by an eye movement instrument, and N is an integer greater than or equal to 1;

the characteristic extraction module is used for extracting the characteristics of the N original eye movement track maps;

and the recognition module is used for inputting the characteristics of the N original eye movement track maps into the classifier and acquiring a recognition result.

9. The apparatus of claim 8, wherein the feature extraction module comprises:

the image processing unit is used for combining the N original eye movement track maps into M samples to be identified, wherein each sample to be identified comprises L multiplied by L original eye movement track maps, L is an integer which is greater than or equal to 1, M is an integer which is greater than or equal to 1, and the product of L multiplied by M is less than or equal to N;

and the extraction unit is used for extracting the characteristics of each sample to be identified.

10. The apparatus according to claim 9, wherein the picture processing unit is specifically configured to determine L × M original eye trajectory diagrams from the N original eye trajectory diagrams; and combining the L multiplied by M original eye movement track maps into M samples to be identified according to an L multiplied by L distribution mode.