Albert Sledzianowski
There are many evidences about unique emotions evoked by music in different individuals and that, the saccadic eye movements (saccades) and pupil size are sensitive to different emotions. The experiment presented in this article concerns both issues, as we investigated possible changes in parameters of saccades, caused by emotions induced by music listening. It was assumed, that found correlations will help to determine subject’s age and mood dependently on recorded saccade parameters. The purpose of this experiment was to find ability to distinguish and classify saccade parameters to one of defined groups of age (young/old) or mood (energetic/relaxed) induced by music.
We have measured saccades of two groups of 6 subjects in age below 30 and over 60, during musical sessions with energetic and relaxing music. Collected data were analyzed in search of possible correlations between characteristics of respondents and saccade parameters, using combination of different types of filters and classifiers from WEKA. Classifications showed statistical changes between age groups, in the latency (23.6% of difference) and in the pupil’s size (16, 6% of difference), both found extremely significant (P>0.0001). In case of Mood, results showed changes in the group of younger adults, in the latency (P=0.4532) and very significant for the amplitude (P=0.0001) and for the average velocity (P=0.0048). The best classification results were obtained for Age and Mood groups. Prediction of age group showed the accuracy of 91.4%. In case of Mood groups, obtained percentage of correctly classified instances was between 96.6 and 97%. For both types of groups, best predictions were obtained by Random Forest and Multilayer Perceptron. The results of classifications allowed to build the confusion matrix and decision trees based on values of saccades parameters and data of subjects. It showed differences in saccade processing between particular groups. Article tries to explain main differences in obtained results by SAT and LATER models, exemplifying the computational nature of human brain processing.
The comparison of predictions made on the basis of the obtained results, showed acceptable statistics for examined subjects, which may suggest further researches at the intersection of machine learning, human age, mood, and eye moves. The results of this experiment suggest usefulness of the Eye Tracking and Eye Movement parameters classifications in machine learning driven detection of human features.
இந்தக் கட்டுரையைப் பகிரவும்
English 
Spanish 
Chinese 
Russian 
German 
French 
Japanese 
Portuguese 
Hindi 