Generic Physiological Features as Predictors of Player Experience

Abstract

Abstract. This paper examines the generality of features extracted from heart rate (HR) and skin conductance (SC) signals as predictors of self-reported player affect expressed as pairwise preferences. Artificial neural networks are trained to accurately map physiological features to expressed affect in two dissimilar and independent game surveys. The performance of the obtained affective models which are trained on one game is tested on the unseen physiological and selfreported data of the other game. Results in this early study suggest that there exist features of HR and SC such as average HR and one and two-step SC variation that are able to predict affective states across games of different genre and dissimilar game mechanics.

Multimodal PTSD characterization via the StartleMart game

“…In order to account for any day-variation effects, signals from the same patients, but taken on different days, are treated as separate individuals. In accordance with recommendations from earlier studies on SC signal processing[7,8,24], a number of features that summarize the key statistical characteristics of SC signals are extracted: Mean SC value (SCx), standard deviation of the SC signal (SC σ ), minimum SC value (SC min ), maximum SC value (SC max ), the difference between the maximum and minimum SC value (SC range ), the Pearson correlation between recording time (t) and SC values (R SCt ), the value of the first SC sample (SC α ), the value of the final SC sample (SC ω ), the difference and absolute difference between final and first SC value (SC ω−α ) and (|SC ω−α |), the time of the minimum SC value (t SCmin ), the time of the maximum SC value (t SCmax ), the absolute time (t) difference between the minimum and maximum SC values (|t SCrange |), the means of the absolute values of the first and second differences of the SC signal (SC |δ1| ) and (SC |δ2| ). An uncommonly used feature, the mean of the absolute first difference of the absolute first difference (|SC δ δ |), is added in an attempt to describe the tendency toward weak habituation in the signal.6.2 Features Extracted from Blood Volume PulseBVP features are also extracted from complete game sessions after inspection for artifacts.…”

“…A wide range of approaches exist for capturing stress using physiological, behavioral, and self-report data or combinations thereof. Earlier work on stress detection [6] has demonstrated how features extracted from raw physiological signals can be used to discern between a variety of emotional states in general [7] and in games [8], and previous work has presented designs and studies that build affective loops for PTSD treatment by coupling presented stimuli with PTSD symptom severity [5,9]. Informed by this previous research, our con-figuration captures indications of stress responses by continuously recording SC and BVP and by requesting self-reports from the player.…”

A taxonomy and state of the art revision on affective games

“…Scheirer et al [14] Skin conductivity, blood pressure and mouse patterns for affective analysis Sakurazawa et al [15] Skin conductance response as emotional state detector Mandryk et al [16][17][18][19] Efficiency of several physiological measures Hazlett and L. [20] Facial electromyography Nacke and Lindley [21], Nacke et al [22,23] Multiple measures and flow between affective states Perez Martínez et al [24] Generality of physiological features Ravaja et al [25], Drachen et al [26], Levillain et al [27], Wu and Lin [28], Gualeni et al [29], Vachiratamporn et al [30], Martey et al [31], Abhishek and Suma [32], Landowska and Wróbel [33], Li et al [34] Applications of physiological measures Giakoumis et al [35] Automated boredom detection Chanel et al [36,37], Nogueira et al [38] Machine-learning classifiers for emotional states Jones and Sutherland [39] Emotion detection from player's voice Garner and Grimshaw [40], Nacke et al [41], Nacke and Grimshaw [42] Effect of the sound in players' fear level Christy and Kuncheva [43] Computer mouse with affective detection Going a step further, Nacke and Lindley [21], Nacke et al [22,23] studied how to measure the global player experience while playing a game analysing the same physiological metrics as before: electromyography, electrodermal activity and so on. Regarding the player experience, the authors measured the flow between varied affective states, such as anxiety, apathy and boredom.…”

A Panorama of Artificial and Computational Intelligence in Games

“…General game AI → Player modeling: General AI approaches for data-driven player modeling are of utmost importance for the identification of general playing patterns that map to player styles and experiences in a context-free fashion. A notable example of such an attempt is the work of Martinez et al [138] in which detected physiological manifestations of player experience are cross-validated in both prey-predator and racing games yielding generic sets of player modeling attributes across those game genres. General game AI → Games as AI benchmarks: Developers of game-based benchmarks and competitions have much to gain from adopting some of the thinking of the AGI community, in particular by adding meaningful variation to the benchmarks so as to privilege generic solutions rather than domain-specific engineering.…”