Thu, Apr 3, 2025
| فارسی
Volume 26, Issue 3 (Autumn 2024)
Advances in Cognitive Sciences 2024, 26(3): 28-39 |
Back to browse issues page
Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
Hatef B, Pirzad Jahromi G, Ramezanpour S. Measuring the degree of stress based on the Stroop test using the electrical resistance of the skin. Advances in Cognitive Sciences 2024; 26 (3) :28-39
URL: http://icssjournal.ir/article-1-1718-en.html
URL: http://icssjournal.ir/article-1-1718-en.html
1- Neuroscience Research Center, Baqiyatallah University of Medical Sciences, Tehran, Iran
Abstract: (587 Views)
Introduction
Stress is the biological reaction of the human body and is the result of pressure, emotional, mental, or physical challenges such as unexpected and threatening events. Stress can become a serious problem in people's lives and destroys a person's performance, behavior, responsibility, and ability to think. Stress hormones create a sense of fight or flight; thus, we can react to dangerous situations. Despite the fact that stress is useful in some situations and is a sign of health, high stress can have negative effects in the long term. Therefore, identifying the state of stress in everyday life can be a preventive solution to control stressful factors and prevent bad stress from persisting and its complications. Today, various tools are used to measure the sympathetic axis of stress, such as measuring heart rate in smartwatches, measuring galvanic skin resistance (GSR) in lie measuring devices, or more complex tools that use a combination of them and even breathing rate. GSR is defined as a change in the electrical properties of the skin. The signal can be used to capture autonomic nerve responses as a parameter of sweat gland function. The Stroop Color and Word Test (SCWT) is a neuropsychological test widely used for experimental and clinical purposes. It evaluates the ability to control cognitive interference. Meanwhile, wearable artificial intelligence (AI) has emerged as a valuable tool for this purpose. The study hypothesized that the GSR could classify the Stroop test as a stress condition for grades from easy to hard.
Methods
The method consisted of two parts: The first part: Designing and assembly of instrument: 1) Skin conductance sensors module (GSR), 2) STMf032k6t6 module, and 3) Ac (USB to Serial) module. The GSR module connects to an STM32 microcontroller via an output pin on port A1, which performs the analog-to-digital conversion. This conversion transforms the voltage into a digital number that is sent to the recording module. A USB-to-serial connection is utilized to communicate with the Rx base. Additionally, MATLAB software is employed to read GSR data. The GSR sensor records stress by detecting changes in electrical activity, which occur due to sweat gland activity. To accurately capture these changes, the electrodes must be highly sensitive and capable of transmitting the data to the recording device. Skin conductance is measured using easily applied skin electrodes. Essentially, the GSR sensor functions like an ohmmeter, as it measures the electrical conductivity between two points. Second part: Design the stress protocol, data recording, and learning machine to categorize stress levels. In this study, the Stroop test was used to detect Word and Color. The 21 healthy young volunteers sat in a chair in front of the monitor. The participant was asked to respond to the stimuli (color or word recognition) of the six-color Stroop test (blue, red, green, yellow, purple, and white) by left-clicking the computer mouse. The test consisted of six sets of 40 with intervals between 4 s, 3 s, and 2 s trials. GSR signals recorded at three levels of low, medium, and high stress are also included in the data. After recording and for each set, GSR data was segmented into 10-second segments; features were extracted from each segment, including the following: Mean, median, SD, Max, Min, Low and high-frequency band average, and SD, Slope, and Zero crossing of GSR signal.
Results
After recording, for each stage, GSR data was segmented into 10-second segments; 11 features were extracted from each segment, including the following. Mean (μ) and standard deviation (σ) of GSR plots with N samples. The minimum and maximum values of the GSR part indicate the range of changes of the part. The median value of the segment in statistics and probability theory is one of the measures of central tendency. The median is a number that divides a statistical population or a probability distribution into two equal parts. The slope of the GSR plot estimates the change in skin resistance value over 10 seconds. The average crossing time from zero is calculated by averaging the crossing times of the GSR signal. In calculating this feature, the average value of the signal is zero. The average and standard deviation of the low and high-frequency content of the GSR segment can be calculated by two low-pass and high-pass filters with a cutoff frequency of 2.5 Hz or the Fourier transform of the segment. Since the galvanic response of the skin depends on the perspiration of the skin, which is a slow process like body temperature, the sampling frequency is considered to be 10 Hz, and the low and high-frequency bands are considered between 0-2.5 and 2.5-5 Hz.
After extracting the features from the 10-second segments recorded in four states of relaxation, low, medium, and high stress, a three-layer neural network classifier that includes two hidden layers with ten neurons in the first layer and five neurons in the second layer and an output layer with one neuron has been trained with 70% samples. The activity functions of this network are also considered sigmoid. The number of stimuli in each Stroop test was 40. Based on this, the recording time of each test with 2-, 3-, and 4-secondinter-stimulus intervals was 60, 90, and 120 seconds, respectively. Since these signals are segmented into 10-second pieces, the total number of pieces in four situations is 1452 pieces. A Stroop test was performed for both color and word recognition modes.
Lunberg-Marquardt algorithm, which is a combination of gradient descent and Newton algorithms in classical optimization, has been a training method that has been used to train the network with 60% data. The results of the10-fold validation of the neural network classifier using training and test data showed that this classifier was able to provide accuracy of 95.91% and 92.63% for training and test data. Based on these results, the sensitivity of the classifier to the state of rest is lower than in other states. This value was higher for the medium stress.
Conclusion
One of the promising non-invasive methods that has been widely used in the detection of stress and emotions is electrodermal activity (EDA). EDA has a tonic and phasic component called skin conductance level and skin conductance response (SCR) named GSR. The same study was collected from the result of the deconvolution process, namely the sample mean, the first absolute difference, and the normalized first difference as the input of the classification process using the extreme learning machine (ELM). The output of stress level classification was mild, moderate, and severe. Using cvxEDA is more accurate or smooth than CDA. However, both methods can separate SCR from primary skin conductance and show small peaks from SCR. The results of the classification process showed that both CDA and cvxEDA methods with 50 hidden layers in ELM had high accuracy in stress level classification, which was 95.56% and 94.45%, respectively. This study developed a stress level classification method using ELM and SCR statistical features. The results showed that EDA can classify the stress level of the Stroop test with a high accuracy of 94%. This model can help people take control of their mental health while working too much, which can lead to anxiety and depression due to untreated stress.
Ethical Considerations
Compliance with ethical guidelines
The study was approved by Baqiyatallah University of Medical Sciences, with ethical code: IR.BMSU.REC.1399.519. The names of the individuals were stored in code form and archived confidentially. Written consent was obtained from the individuals participating in the study.
Authors' contributions
Boshra Hatef: Conceptualization, methodology, formal analysis, writing, reviewing and editing of manuscript, supervision, and project administration. Gila Pirzad Jahromi: Conceptualization, investigation, writing, original draft and writing, reviewing and editing of manuscript. Saeed Ramezanpour: Design and construction of the instrument, classification of data and writing, the original draft of the manuscript.
Funding
There is no financing support.
Acknowledgments
The authors thank the laboratory of the Neuroscience Center of the Baqiyatallah University of Medical Sciences.
Conflicts of interest
There was no conflict of interest.
Stress is the biological reaction of the human body and is the result of pressure, emotional, mental, or physical challenges such as unexpected and threatening events. Stress can become a serious problem in people's lives and destroys a person's performance, behavior, responsibility, and ability to think. Stress hormones create a sense of fight or flight; thus, we can react to dangerous situations. Despite the fact that stress is useful in some situations and is a sign of health, high stress can have negative effects in the long term. Therefore, identifying the state of stress in everyday life can be a preventive solution to control stressful factors and prevent bad stress from persisting and its complications. Today, various tools are used to measure the sympathetic axis of stress, such as measuring heart rate in smartwatches, measuring galvanic skin resistance (GSR) in lie measuring devices, or more complex tools that use a combination of them and even breathing rate. GSR is defined as a change in the electrical properties of the skin. The signal can be used to capture autonomic nerve responses as a parameter of sweat gland function. The Stroop Color and Word Test (SCWT) is a neuropsychological test widely used for experimental and clinical purposes. It evaluates the ability to control cognitive interference. Meanwhile, wearable artificial intelligence (AI) has emerged as a valuable tool for this purpose. The study hypothesized that the GSR could classify the Stroop test as a stress condition for grades from easy to hard.
Methods
The method consisted of two parts: The first part: Designing and assembly of instrument: 1) Skin conductance sensors module (GSR), 2) STMf032k6t6 module, and 3) Ac (USB to Serial) module. The GSR module connects to an STM32 microcontroller via an output pin on port A1, which performs the analog-to-digital conversion. This conversion transforms the voltage into a digital number that is sent to the recording module. A USB-to-serial connection is utilized to communicate with the Rx base. Additionally, MATLAB software is employed to read GSR data. The GSR sensor records stress by detecting changes in electrical activity, which occur due to sweat gland activity. To accurately capture these changes, the electrodes must be highly sensitive and capable of transmitting the data to the recording device. Skin conductance is measured using easily applied skin electrodes. Essentially, the GSR sensor functions like an ohmmeter, as it measures the electrical conductivity between two points. Second part: Design the stress protocol, data recording, and learning machine to categorize stress levels. In this study, the Stroop test was used to detect Word and Color. The 21 healthy young volunteers sat in a chair in front of the monitor. The participant was asked to respond to the stimuli (color or word recognition) of the six-color Stroop test (blue, red, green, yellow, purple, and white) by left-clicking the computer mouse. The test consisted of six sets of 40 with intervals between 4 s, 3 s, and 2 s trials. GSR signals recorded at three levels of low, medium, and high stress are also included in the data. After recording and for each set, GSR data was segmented into 10-second segments; features were extracted from each segment, including the following: Mean, median, SD, Max, Min, Low and high-frequency band average, and SD, Slope, and Zero crossing of GSR signal.
Results
After recording, for each stage, GSR data was segmented into 10-second segments; 11 features were extracted from each segment, including the following. Mean (μ) and standard deviation (σ) of GSR plots with N samples. The minimum and maximum values of the GSR part indicate the range of changes of the part. The median value of the segment in statistics and probability theory is one of the measures of central tendency. The median is a number that divides a statistical population or a probability distribution into two equal parts. The slope of the GSR plot estimates the change in skin resistance value over 10 seconds. The average crossing time from zero is calculated by averaging the crossing times of the GSR signal. In calculating this feature, the average value of the signal is zero. The average and standard deviation of the low and high-frequency content of the GSR segment can be calculated by two low-pass and high-pass filters with a cutoff frequency of 2.5 Hz or the Fourier transform of the segment. Since the galvanic response of the skin depends on the perspiration of the skin, which is a slow process like body temperature, the sampling frequency is considered to be 10 Hz, and the low and high-frequency bands are considered between 0-2.5 and 2.5-5 Hz.
After extracting the features from the 10-second segments recorded in four states of relaxation, low, medium, and high stress, a three-layer neural network classifier that includes two hidden layers with ten neurons in the first layer and five neurons in the second layer and an output layer with one neuron has been trained with 70% samples. The activity functions of this network are also considered sigmoid. The number of stimuli in each Stroop test was 40. Based on this, the recording time of each test with 2-, 3-, and 4-secondinter-stimulus intervals was 60, 90, and 120 seconds, respectively. Since these signals are segmented into 10-second pieces, the total number of pieces in four situations is 1452 pieces. A Stroop test was performed for both color and word recognition modes.
Lunberg-Marquardt algorithm, which is a combination of gradient descent and Newton algorithms in classical optimization, has been a training method that has been used to train the network with 60% data. The results of the10-fold validation of the neural network classifier using training and test data showed that this classifier was able to provide accuracy of 95.91% and 92.63% for training and test data. Based on these results, the sensitivity of the classifier to the state of rest is lower than in other states. This value was higher for the medium stress.
Conclusion
One of the promising non-invasive methods that has been widely used in the detection of stress and emotions is electrodermal activity (EDA). EDA has a tonic and phasic component called skin conductance level and skin conductance response (SCR) named GSR. The same study was collected from the result of the deconvolution process, namely the sample mean, the first absolute difference, and the normalized first difference as the input of the classification process using the extreme learning machine (ELM). The output of stress level classification was mild, moderate, and severe. Using cvxEDA is more accurate or smooth than CDA. However, both methods can separate SCR from primary skin conductance and show small peaks from SCR. The results of the classification process showed that both CDA and cvxEDA methods with 50 hidden layers in ELM had high accuracy in stress level classification, which was 95.56% and 94.45%, respectively. This study developed a stress level classification method using ELM and SCR statistical features. The results showed that EDA can classify the stress level of the Stroop test with a high accuracy of 94%. This model can help people take control of their mental health while working too much, which can lead to anxiety and depression due to untreated stress.
Ethical Considerations
Compliance with ethical guidelines
The study was approved by Baqiyatallah University of Medical Sciences, with ethical code: IR.BMSU.REC.1399.519. The names of the individuals were stored in code form and archived confidentially. Written consent was obtained from the individuals participating in the study.
Authors' contributions
Boshra Hatef: Conceptualization, methodology, formal analysis, writing, reviewing and editing of manuscript, supervision, and project administration. Gila Pirzad Jahromi: Conceptualization, investigation, writing, original draft and writing, reviewing and editing of manuscript. Saeed Ramezanpour: Design and construction of the instrument, classification of data and writing, the original draft of the manuscript.
Funding
There is no financing support.
Acknowledgments
The authors thank the laboratory of the Neuroscience Center of the Baqiyatallah University of Medical Sciences.
Conflicts of interest
There was no conflict of interest.
Type of Study: Research |
Received: 2024/09/7 | Accepted: 2024/10/20 | Published: 2024/10/20
Received: 2024/09/7 | Accepted: 2024/10/20 | Published: 2024/10/20
Send email to the article author
| Rights and permissions | |
 |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |
