Analysis and categorization of eye tracking data describing scanpaths
Described herein are various technologies pertaining to analysis of eye tracking data. A head and/or eyes of an observer who is viewing a visual stimulus is monitored, and eye tracking data that is representative of the path of the eyes of the observer over time (a scanpath) is generated. The eye tracking data is time-series data that defines the location of the focal point, or other measurable characteristics, of the eyes of the observer on the visual stimulus over time. A feature vector is constructed based upon the eye tracking data, where the feature vector is representative of the eye tracking data, and is thus representative of the scanpath. The feature vector is compared with other feature vectors to identify scanpaths that correspond to the scanpath represented by the feature vector.
1. A computing system, comprising:
at least one processor; and
memory that stores computer-executable instructions that, when executed by the at least one processor, cause the at least one processor to perform acts comprising:
receiving eye tracking data for a first scanpath, wherein the eye tracking data comprises data points having positional coordinates and corresponding timestamps;
segmenting the eye tracking data into a plurality of segments, wherein the plurality of segments comprises a first segment and a second segment, wherein the first segment comprises an entirety of the eye tracking data and the second segment consists of less than the entirety of the eye tracking data;
constructing a first p-dimensional feature vector based upon the first segment;
constructing a second p-dimensional feature vector based upon the second segment;
constructing an n-dimensional feature vector based upon features of the eye tracking data, wherein the n-dimensional feature vector comprises the first p-dimensional feature vector and the second p-dimensional feature vector;
determining that the first scanpath corresponds to a second scanpath based upon the n-dimensional feature vector; and
responsive to determining that the first scanpath corresponds to the second scanpath, outputting an indication that the first scanpath corresponds to the second scanpath.
2. The computing system of claim 1 , wherein determining that the first scanpath corresponds to the second scanpath further comprises:
comparing the n-dimensional feature vector with a second n-dimensional feature vector, the second n-dimensional feature vector representative of the second scanpath; and
determining that the first scanpath corresponds to the second scanpath based upon the comparison of the n-dimensional feature vector with the second n-dimensional feature vector.
3. The computing system of claim 1 , wherein determining that the first scanpath corresponds to the second scanpath further comprises:
clustering n-dimensional feature vectors into a plurality of clusters based upon a clustering parameter, wherein the n-dimensional feature vectors comprise the n-dimensional feature vector and the second n-dimensional feature vector, and further wherein a cluster in the plurality of clusters includes the n-dimensional feature vector and the second n-dimensional feature vector; and
determining that the first scanpath corresponds to the second scanpath based upon the cluster including the n-dimensional feature vector and the second n-dimensional feature vector.
4. The computing system of claim 3 , wherein the clustering parameter is a minimum number of members to form a cluster.
5. The computing system of claim 3 , wherein the parameter is a neighborhood radius.
6. The computing system of claim 3 , wherein the eye tracking data is collected in response to a visual stimulus being presented on a display.
7. The computing system of claim 6 , the acts further comprising:
receiving second eye tracking data describing the second scanpath, wherein the second eye tracking data comprises second data points having second positional coordinates and corresponding second timestamps.
8. The computing system of claim 7 , wherein the second eye tracking data is collected in response to the visual stimulus being presented on the display.
9. The computing system of claim 1 , wherein the eye tracking data is segmented into the plurality of segments based upon time.
10. The computing system of claim 1 , wherein features of the n-dimensional feature vector comprise:
a first geometric median of eye tracking data in the first segment; and
a second geometric median of eye tracking data in the second segment.
11. A method executed by a computer system that includes a processor and memory, the method comprising:
receiving eye tracking data that represents a first scanpath, wherein the eye tracking data is time-series data that represents focal points of eyes of an observer over a visual stimulus over time;
segmenting the eye tracking data into a plurality of segments, wherein the plurality of segments comprise a first segment and a second segment, wherein the first segment consists of a first number of data points in the time-series data, the second segment consists of a second number of data points in the time-series data, and further wherein the first number of data points is different from the second number of data points;
constructing a first p-dimensional feature vector for the first segment;
constructing a second p-dimensional feature vector for the second segment;
constructing an n-dimensional feature vector that represents the first scanpath, wherein the n-dimensional feature vector comprises the first p-dimensional feature vector and the second p-dimensional feature vector;
determining that the first scanpath corresponds to a second scanpath based upon values of the n-dimensional feature vector; and
responsive to determining that the first scanpath corresponds to the second scanpath, outputting an indication that the first scanpath corresponds to the second scanpath.
12. The method of claim 11 , wherein determining that the first scanpath corresponds to the second scanpath comprises:
comparing the n-dimensional feature vector with a second n-dimensional feature vector, the second n-dimensional feature vector represents the second scanpath; and
determining that the first scanpath corresponds to the second scanpath based upon the comparison of the n-dimensional feature vector with the second n-dimensional feature vector.
13. The method of claim 12 , wherein determining that the first scanpath corresponds to the second scanpath further comprises:
clustering n-dimensional feature vectors into a plurality of clusters, wherein a cluster in the plurality of clusters comprises the n-dimensional feature vector and the second n-dimensional feature vector; and
determining that the first scanpath corresponds to the second scanpath based upon the n-dimensional feature vector and the second n-dimensional feature vector being included in the cluster.
14. The method of claim 11 , wherein the n-dimensional feature vector includes:
a first value for a first geometric median of the first number of data points in the time-series data; and
a second value for a second geometric median of the second number of data points in the time-series data.
15. A computer-readable storage medium comprising instructions that, when executed by a processor, cause the processor to perform acts comprising:
receiving eye tracking data that represents a first scanpath, wherein the eye tracking data is time-series data that represents focal points of eyes of an observer over a visual stimulus over time;
segmenting the eye tracking data into a plurality of segments, wherein the plurality of segments comprise a first segment and a second segment, wherein the first segment consists of a first number of data points in the time-series data, the second segment consists of a second number of data points in the time-series data, and further wherein the first number of data points is different from the second number of data points;
constructing a first p-dimensional feature vector for the first segment;
constructing a second p-dimensional feature vector for the second segment;
constructing an n-dimensional feature vector that represents the first scanpath, wherein the n-dimensional feature vector comprises the first p-dimensional feature vector and the second p-dimensional feature vector;
determining that the first scanpath corresponds to a second scanpath based upon values of the n-dimensional feature vector; and
responsive to determining that the first scanpath corresponds to the second scanpath, outputting an indication that the first scanpath corresponds to the second scanpath.
16. The computer-readable storage medium of claim 15 , wherein determining that the first scanpath corresponds to the second scanpath comprises:
comparing the n-dimensional feature vector with a second n-dimensional feature vector, the second n-dimensional feature vector represents the second scanpath; and
determining that the first scanpath corresponds to the second scanpath based upon the comparison of the n-dimensional feature vector with the second n-dimensional feature vector.
17. The computer-readable storage medium of claim 16 , wherein determining that the first scanpath corresponds to the second scanpath further comprises:
clustering n-dimensional feature vectors into a plurality of clusters, wherein a cluster in the plurality of clusters comprises the n-dimensional feature vector and the second n-dimensional feature vector; and
determining that the first scanpath corresponds to the second scanpath based upon the n-dimensional feature vector and the second n-dimensional feature vector being included in the cluster.
18. The computer-readable storage medium of claim 15 , wherein the n-dimensional feature vector includes:
a first value for a first geometric median of the first number of data points in the time-series data; and
a second value for a second geometric median of the second number of data points in the time-series data.
19. The computer-readable storage medium of claim 15 , wherein the eye tracking data is segmented into the plurality of segments based upon time.