If the number of data values are too much, then it becomes very difficult to analyse. If the dimensions are too large and grows leaps and bounds it would be very hard to perform operations on those values. To get rid of this humongous problem, a matrix can be created to show the correlation between pairs of eye values. Patterns can then be obtained from the matrix which helps analyse the data. It removes all the clutter and allows for an efficient procedure of calculation. A new dimension of data can be brought into picture by an user performing the task. The values are displayed both horizontally and vertically in a matrix cell. Some kind of grouping is required to form a structure among the pairs of data.
For demonstration purposes, we have decided to use three common metrics in eye tracking: average fixation duration (M2), average saccade length (M3), and average task completion time (M4). Additionally, we used the similarity of transition matrices (M1) as a metric 11. A 3D transition matrix is computed based on all sub-scanpaths with three subsequent fixations. More precisely, the fixation labels were used to generate a 3D transition matrix by counting how often a particular label sequence occurred. Thereby, a 3D transition matrix was created for each scanpath, which was interpreted as a 1D vector. The similarity calculation is now reduced to calculating the Euclidean distance of two vectors.
For each metric, we calculate similarity values svMi for each pair of participants Pl,m (for a given stimuli s). We denote the similarity values as svl,m,M1, svl,m,M2, svl,m,M3, and svl,m,M4.

In total, for 20 participants, we will have 80 different values for one participant in the form of PlsvMi, where 1?l?20 and 1?i?4. Our visualization technique uses dimensional-stacked rectangular sub-grids, where we added a sub-grid in each matrix cell. The sub-grid contains the calculated similarity values svl,m,Mi. They are stored in clock-wise direction. There are many visualization techniques that can be used to show multi-dimensional data. In general, it is very easy to use a matrix-based visualization scheme for visual exploration as shown in Figure 1, but it becomes more challenging with an increasing number of attributes. This is in particular the case, if a larger number of matrices is involved, where each encodes the information of one metric. To overcome the problem of comparing several matrices, we introduced a matrix-based visualization that we used to explore the features extracted from eye movement data. A matrix cell is used to encode multiple similarity values, following the concept of dimensional stacking: embedding dimensions within other dimensions.

We Will Write a Custom Essay Specifically
For You For Only \$13.90/page!

order now 