*3.2. Apparatus*

To collect the data, we used the StressFoot prototypes. In addition, the Empatica E4 wristband [85] was used to collect the participant's EDA. All the study tasks were executed and visualised on a 13" Macbook Pro, which was connected to an extended DELL LED monitor. (see Figure 4).

**Figure 4.** Experimental setup: ( **A**) StressFoot prototype, (**B**) Empatica E4 wristband, ( **C**) Macbook Pro, (**D**) Dell LED monitor and (**E**) adjustable chair.

#### *3.3. Procedure & Tasks*

After explaining the aim of the study, participants were required to fill out consent forms. Participants were informed that the study aims to explore behavioural changes while performing tasks which induce stress and relaxation. We then collected demographic data and asked the participant to wear the apparatuses. The participants were then asked to sit in-front of the Macbook Pro with an extended display as shown in Figure 4. We allowed the user to adjust the chair and the distance of the table according to their personal preference. During the study, they were also allowed to cross their legs. However, we asked them not to take the foot up to the chair. Once the setup was complete and the user was comfortable, we asked them to complete four tasks. The first, as well as the third task, were stress-inducing tasks, while the second and fourth tasks intended to create relaxation. The task order remained consistent for all participants. After the completion of each task, the participant had to fill a questionnaire. On a 7-point Likert scale, the users had to rate their perceived stress level, energy level and how pleasant they found the task. A NASA Task Load Index (NASA TLX) [86] was conducted to calculate the overall workload per task.

#### 3.3.1. Task 1 [Stress]: Stroop Color and Word Test

The Stroop color and word test is a common stress test (Stroop Test) [87] used in many previous works [10,88,89]. For our study, we customised an open source MatLab-based Stroop Test tool to display four words ("Red", "Magenta", "Green" and "Blue") on the screen. Participants were then asked to type the first letter of the colour ('R', 'M', 'G', 'B') the displayed word was coloured with. After each answer, the program provides the result by displaying the word "correct" or "incorrect". The participant has to perform 5 rounds. Each round has 20 words with a mismatch of 80% between the word and the colour it is printed with. We use several tactics to induce stress among participants. As participants were required to provide answers as quickly as possible, we first induced stress by providing performance feedback on their speed and accuracy. We further elevated stress levels by exposing the participant to loud traffic noises through a headset. Moreover, an experimenter continuously observed the participant, and verbally commented on the participant's performance. Completing the stress task took 5 min.

#### 3.3.2. Task 2 [Relaxation]: Minesweeper Introduction Video

The second task aimed to induce relaxation among the user after the stress task. It was also intended to provide an introduction to the game, Minesweeper. For this, we selected an entertaining Minesweeper Let's Play Video from the platform youtube.com. The selection of this video was completed beforehand with 3 pilot study participants. We explicitly informed the user that the video is purely for relaxation purposes and also encouraged the user to relax and enjoy the video. The video took 5 min.

#### 3.3.3. Task 3 [Stress]: Minesweeper

Previous studies already utilized Minesweeper to induce a stress condition [90,91]. Therefore, we used an implementation, in which the user had to find 10 mines in an area of 10 × 10 fields within a time frame of 100 s. The user had to complete 10 rounds. To increase the stress level, the given time was reduced by 10 s every round the user successfully completed the task. The score was indicated at all times, and the user was challenged to complete at least 7 rounds successfully, which was deemed to increase stress. Moreover, similar to task 1, the experimenter provided verbal comments aiming to further elevate the stress level.

#### 3.3.4. Task 4 [Relaxation]: Nature Video

This is the second relaxation task, in which we presented a relaxing nature video for 5 min. Similar to the previous relaxation task, we selected this video based on user feedback from a previous pilot study.

## *3.4. Data Gathering*

The participant's plantar pressure data, as well as the accelerometer data, were collected as a time series with a sampling rate of 50 Hz and stored with the timestamp on an SD card. The EDA data was collected with a sampling rate of 4 Hz at the Empatica E4 wristband. From this data, the mean EDA and the average slope over a non-overlapping windows of 10 s were calculated. In addition to collecting sensor data, the questionnaire collected the user's perceived stress level, energy level and the level of pleasantness: "How much did you feel stressed during the task? How energetic did you feel during the task? How pleasant did you find the task?" We quantified the answers by letting participants rate their answer on a 7-pnt Likert scale (1: low, 7: high). Moreover, we conducted the NASA TLX to account for the induced task load. It consists of 6 questions which are equally weighted.

## *3.5. Data Analysis*

To identify whether our task induced stress or relaxation as we expected, we conducted a pairwise *t*-Test on user's perceived stress level, energy level, the level of pleasantness and the task load. In addition, to detect significant difference of average EDA slope we conducted another pairwise *t*-Test. Since the data was normally distributed according to Shapiro Wilk's test, parametric tests such as pairwise *t*-Test and oneway-Anova were used. The participant's plantar pressure data and IMU data were used to calculate features as mentioned above and used in model training.
