Emotion From Mouse and Keyboard

From AIRWiki

Jump to: navigation, search
Emotion from mouse and keyboard
Coordinator: AndreaBonarini (andrea.bonarini@polimi.it)
Tutor: SimoneTognetti (tognetti@elet.polimi.it)
Collaborator:
Students: GoranBurek (goran.burek@gmail.com)
Research Area: Affective Computing
Research Topic: Emotion from Interaction
Start: 2008/04/08
End: 2009/12/22
Status: Closed
Level: Ms
Type: Thesis
Goran Burek has defined a protocol to identify specific emotional states (e.g., attention) and collect and analyze data from different users, comparing them with biological data acquired during the interaction Emotion From Mouse and Keyboard.

Until April 2009 the following parts of project have been established:

  • In order to obtain biological data Goran has developed Affective Logger in C++ using developer API for the ProComp Infinity, the device in charge of collecting bio-signals (such as hart rate, skin conductance, body temperature etc.)
  • In January 2009 the first testing model has been abandoned due to lack of guarantee that user doing the test will find himself in the appropriate emotional state.The model was imagined to test the user for levels of concentration, stress and tiredness. To test each of three emotional states there was a phase of inducing the given state (e.g. playing annoying music to induce stress) and a phase of testing where user would do some simple cognitive tasks. The tasks were the same for every emotional state. Goal was to compare the data and performance between users and their states. But the model was abandoned because of rapid changes of emotional state and uncertainty of the duration of each emotional state.
  • The new model was accepted. The goal was to identify difference between concentration and tiredness. Uses will be monitored during the whole work day for some number of days. During the day they will have the liberty to do their work, but with exception that every hour they have to do a test. This test is composed of four tasks: a keyboard game, a text copy task, a mouse game and a search for a given link on a website. Data of our interest are typing behavior and mouse movement. The test is developed in form of php sites, results and performance of each task for each user is saved in database. During the day when user is monitored he has to activate a keylogger and mouselogger that collect his typing and movement and save it to the database. (for privacy reasons the keylogger will not record the exact characters the user has written but just a notion that a character has been typed). During the test time there will be another keylogger activated, this one will record everything that has been typed in order to have a better look at user's performance. Upon ending the test the mentioned keylogger will be turned of.
  • From the gathered data in the database the features are extracted (e.g. writing speed, key press duration, press frequency of specific keys (backspace, del, space, enter...)) The features will be analyzed in MATLAB. To see the promising features for the moment the test is performed on one user. After few days we expect to have preliminary view of the more and less interesting features and have some useful conclusions how to proceed.

The final work is available to the AIRWiki users from the discussion tab, or upon request o the Project coordinator.