Video games are interactive software able to arouse different kinds of emotions in players. Usually, the game designer tries to define a set of game features able to enjoy, engage, and/or educate the consumers. Through the gameplay, the narrative, and the game environment, a video game is able to interact with players' intellect and emotions. Thanks to the technological developments of the last years, the gaming industry has grown to become one of the most important entertainment markets. The scientific community and private companies have put a lot of efforts on the technical aspects as well as on the interaction aspects between the players and the video game. Considering the game design, many theories have been proposed to define some guidelines to design games able to arouse specific emotions in consumers. They mainly use interviews or observations in order to deduce the goodness of their approach through qualitative data. There are some works based on empirical studies aimed at studying the emotional states directly on players, using quantitative data. However, these researches usually consider the data analysis as a classification problem involving, mainly, the game events. Our goal is to understand how the feelings, experienced by the players, can be automatically deducted, and how these emotional states can be used to improve the game quality. In order to pursue this purpose, we have measured the mental states using physiological signals in order to return a set of quantitative values used to identify the players emotions. The most common ways to identify emotions are: to use a discrete set of labels (e.g., joy, anger), or to assess them inside an n-dimensional vector space. Albeit the most natural way to describe the emotions is to represent them through their name, the latter approach provides a quantitative result that can be used to define the new game status. In this thesis, we propose a framework aimed at an automatic assessment, using physiological data, of emotions in a 2-dimensional space, structured by valence and arousal vectors. The former may vary between pleasure and displeasure, while the latter defines the level of physiological activation. As a consequence, we have considered as most effective to infer the players’ mental states, the following physiological data: electrocardiography (ECG), electromyography on 5 facial muscles (Facial EMG), galvanic skins response (GSR), and respiration intensity/rate. We have recorded a video, during a set of game sessions, of the player's face and of her gameplay. To acquire the affective information, we have shown the recorded video and audio to the player, and we have asked to self-assess her/his emotional state over the entire game on the valence and arousal vectors presented above. Starting from this framework, we have conducted two sets of experiments. In the first experiment, our aim was to validate the procedure. We have collected the data of 10 participants while playing at 4 platform games. We have also analyzed the data to identify the emotion pattern of the player during the gaming sessions. The analysis has been conducted in two directions: individual analysis (to find the physiological pattern of an individual player), and collective analysis (to find the generic patterns of the sample population). The goal of the second experiment has been to create a dataset of physiological information of 33 players, and to extend the data analysis and the results provided by the pilot study. We have asked the participants to play at 2 racing games in two different environments: on a standard monitor and using a head mounted display for Virtual Reality. After we have collected the information useful to the dataset creation, we have analyzed the data focusing on individual analysis. In both analyses, the self-assessment and the physiological data have been used in order to infer the emotional state of the players in each moment of the game sessions, and to build a prediction model of players' emotions using Machine Learning techniques. Therefore, the three main contributions of this thesis are: to design a novel framework for study the emotions of video game players, to develop an open-source architecture and a set of software able to acquire the physiological signals and the affective states, to create an affective dataset using racing video games as stimuli, to understand which physiological conditions could be the most relevant in order to determine the players' emotions, and to propose a method for the real-time prediction of a player's mental state during a video game session. The results to suggest that it is possible to design a model that fits with player's characteristics, predicting her emotions. It could be an effective tool available to game designers who can introduce innovative features to their games.
EMOTIONS RECOGNITION IN VIDEO GAME PLAYERS USING PHYSIOLOGICAL INFORMATION / M. Granato ; school director: P. Boldi ; advisor: D. Gadia. - Milano : Università degli studi di Milano. DIPARTIMENTO DI INFORMATICA Giovanni Degli Antoni, 2019 Feb 01. ((31. ciclo, Anno Accademico 2018.
|Titolo:||EMOTIONS RECOGNITION IN VIDEO GAME PLAYERS USING PHYSIOLOGICAL INFORMATION|
|Supervisori e coordinatori interni:||BOLDI, PAOLO|
|Data di pubblicazione:||1-feb-2019|
|Settore Scientifico Disciplinare:||Settore INF/01 - Informatica|
|Citazione:||EMOTIONS RECOGNITION IN VIDEO GAME PLAYERS USING PHYSIOLOGICAL INFORMATION / M. Granato ; school director: P. Boldi ; advisor: D. Gadia. - Milano : Università degli studi di Milano. DIPARTIMENTO DI INFORMATICA Giovanni Degli Antoni, 2019 Feb 01. ((31. ciclo, Anno Accademico 2018.|
|Digital Object Identifier (DOI):||http://dx.doi.org/10.13130/granato-marco_phd2019-02-01|
|Appare nelle tipologie:||Tesi di dottorato|