Stress analysis from physiological data under pressure: WorkStress3D Dataset

This phrase describes a research approach that analyzes physiological signals to gain insights into psychological stress.

The dataset includes survey responses from 20 participants. In this data type, 4 different questionnaires were used: demographic information, instant questionnaires, a short form of the Positive and Negative Affect Schedule (PANAS) and general stress test.

Facial expression:

While preparing facial expression documents, data from the fer2013 and Cohn-Kanade datasets were also utilized in addition to the facial expressions collected from the participants.. We transformed the facial expressions collected from the participants into 48x48 dimensions and converted them to grayscale pixels to match the format of these datasets. There are two different documents in the facial expressions file:

facial_dataset_2_class.csv: In this data, the labels are assigned as stress-free (0) and stressed (1). When assigning these labels, positive emotions are defined as stress-free, while negative emotions are defined as stressed.

AudioData file: raw audio data

Physiological Signal: The physiological signals of participants were sampled at varying frequencies using the wrist-worn Empatica E4 sensor device. To resolve this issue, a downsampling technique was used to make the frequencies of the signals more uniform. Downsampling was chosen to achieve a balance between computing efficiency and accuracy. By downsampling all physiological inputs to 4 Hz, consistency and effective fusion were attained.

A "0" in the physiological signals file indicates stress-free conditions, while a "1" indicates the presence of stress. Additionally, the physiological signals file contains data files in three different window sizes: 15 seconds, 30 seconds, and 60 seconds of data windows.

All raw physiological signals recorded with the Empatica E4 armband are located in the EmpaticaSource.zip file.