**2. Background**

In this study several parameters have been analyzed: (a) the physiological sensing mode (ECG, TEB and EDA), (b) the window length, (c) the features extracted from each signal, (d) the number of features to obtain the best results, and (e) the type of classifier.

• Selection of Physiological sensing modality: In this part, we compare the physiological signal under study and determine which physiological signal provides more relevant information about the individual activity. The signals used are ECG, TEB, and EDA. It is possible to find numerous works in which these signals are used to detect stress, emotions, and activity in the literature. The ECG signal is used in some papers such as [23], where the obtained results sugges<sup>t</sup> that positive emotions lead to alterations in HRV, which may be beneficial in some illness treatment [19,31,32].

TEB is also used in some papers, though it is less useful than ECG and EDA signals. The work [25] demonstrated that its use is decisive to detect stress. In addition, most of the studies considered several signals, such as the paper [28] which contains the study on the correlation between heart rate, electrodermal activity and Player Experience in First-Person Shooter Games, concluding that their results indicate correlation between the physiological measures and gameplay experience, even in relatively simple measurement scenarios. Another work, [29] studies the individual differences within the electrodermal activity as subjects' anxiety, which concludes that in normal

subjects there are individual electrodermal differences as a function of trait-anxiety scores. However, few papers provide a deep study of features for the three signals, such as the use of these signals with the same purpose.


The features extracted from the TEB signal are used in some works such as, [24] where the approach is to study cardiovascular reactivity during emotional activation in men and women. Here, the TEB has been acquired together with ECG and the heart sound. In [56] the full respiratory signal was derived from the thoracic impedance raw data, like in our case.

Finally, the EDA signal is studied in several papers, [26,42,49,53,55,57,58]. A complete study about the EDA signal is shown in [41,59].

