Going into this summer, I’ve never done anything research related so it was a new experience for me. I am working in the Perform Laboratory under Gabriel Diaz. This lab is focused on understanding the mechanisms that able humans to execute everyday visually guided actions. The lab works a majority of the time with virtual reality and motion capture technology.
My research is focused on classifying eye movements using a supervised learning model. The types of eye movements I am concerned about are fixations, pursuits, and saccades. A fixation is what it sounds like, when your eyes are fixated on something and barely moving if at all. A pursuit is when your eyes are tracking a moving object. And then a saccade is when your eyes move from one thing to another thing very quickly. I started by acquiring a pre labeled data set from a previous experiment. From there I extracted 4 features: total duration, average acceleration, average velocity, and total amplitude. From there I got the statistics from those features along the different eye movements finding the mean and also getting the min/ max values. From there I needed to use a support vector machine(SVM) to train my model. SVM are models with algorithms that analyze data used for classification. Given a training set, the SVM will build a model that can assign new data points into a specific category. I trained the model using one of the pre labeled datasets. After the model was trained, I ran other data sets provided by the same experiment through the model and calculated the how accurate the model was. Through 10 trials the model correctly predicted the eye movement 96.5% of the time.