A Holistic Approach to Stress Identification: Integrating Questionnaires and Physiological Signals through Machine Learning
الموضوعات :
Mrunal Fatangare
1
,
Hemlata Ohal
2
1 - School of Engineering and Technology, Dr. Vishwanath Karad MIT World Peace University, Pune, India
2 - School of Engineering and Technology, Dr. Vishwanath Karad MIT World Peace University, Pune, India
الکلمات المفتاحية: Stress Identification, PPG, Age-Specific Assessment, Data Preprocessing, SVM, Feature Extraction, Classification Techniques, KNN, Stress Assessment, Well-being,
ملخص المقالة :
This research project presents a comprehensive methodology for stress identification by combining subjective self-report data and objective physiological signals. The proposed system employs a carefully designed questionnaire, tailored to different age groups, to enhance accuracy in stress assessment. Subjects respond to the questionnaire, providing valuable insights into their emotional well-being. Subsequently, physiological data is collected using an infrared (IR) sensor positioned beneath the wrist, close to the artery. The pulse data obtained is meticulously converted into a CSV file, allowing for efficient preprocessing. The preprocessing phase ensures the integrity of the data, preparing it for machine learning (ML) analysis. The study harnesses ML techniques, specifically SVM (Support Vector Machines) & KNN (K-Nearest Neighbors), to classify stress levels based on the pre-processed data. Through feature extraction, relevant patterns are identified, contributing to the accurate characterization of stress states. This integrative approach offers a robust framework for stress assessment, taking into account both subjective and physiological dimensions.
Results demonstrate promising accuracy levels: Support Vector Machine (SVM) Reached a level of precision of 0.98 (+/- 0.20), Decision Tree showed 0.93 (+/- 0.30), and K-Nearest Neighbors (KNN) reached 0.88 (+/- 0.44). It also implements the voting classifier for improved performance of 98.6% of accuracy. These findings underscore the effectiveness of the proposed methodology in accurately identifying stress levels. Integrating subjective insights with objective physiological data not only enhances stress identification but also offers a comprehension of the intricate correlation between mental states and physiological reactions. This comprehensive strategy holds substantial implications across diverse domains such as healthcare, psychology, and human-computer interaction.
[1] Alberdi, A. Aztiria, and A. Basarab, “Towards an automatic early stress recognition system for office environments based on multimodal measurements: A review,” Feb. 01, 2016, Academic Press Inc. doi: 10.1016/j.jbi.2015.11.007.
[2] L. Saturday and H. Selye, “Stress and the General Adaptation Syndrome,” Br Med J, pp. 1383–1392, Jun. 1950.
[3] S. Gedam and S. Paul, “A Review on Mental Stress Detection Using Wearable Sensors and Machine Learning Techniques,” 2021, Institute of Electrical and Electronics Engineers Inc. doi: 10.1109/ACCESS.2021.3085502.
[4] S. A. Singh, K. Gupta, M. Rajeshwari, and T. Janumala, “Detection of Stress Using Biosensors,” 2018. [Online]. Available:www.sciencedirect.comwww.materialstoday.com/proceedings [5] S. Cohen, T. Kamarck, and R. Mermelstein, “A Global Measure of Perceived Stress,” 1983.
[6] M. Masuda and T. H. Holmes, “The Social Readjustment Rating Scale: A Cross-Cultural Study Of Japanese And Americans *p,” Pergamon Press, 1967.
[7] P. F. Lovibond and S. H. Lovibond, “The structure of negative Emotion States: Comparision of the depression anxiety Stress Scales (DASS) with the Beck Depression and Anxiety Inventories,” BehaP. Res. Ther, vol. 33, no. 3, pp. 335–343, 1995.
[8] E. Askari, “Fear Recognition Using Early Biologically Inspired Features Model,” Journal of Information Systems and Telecommunication (JIST), vol. 12, no. 45, pp. 12–19, Mar. 2024, doi: 10.61186/jist.39051.12.45.12.
[9] Jia Uddin, “An Autoencoder based Emotional Stress Detection approach using Electroencephalography Signals,” Journal of Information Systems and Telecommunication, vol. 11, no. 1, pp. 24–30, Mar. 2023.
[10] N. Jadhav, “Hierarchical Weighted Framework for Emotional Distress Detection using Personalized Affective Cues,” Journal of Information Systems and Telecommunication (JIST), vol. 10, no. 38, pp. 89–101, Apr. 2022, doi: 10.52547/jist.16499.10.38.89.
[11] B. S. Mcewen, A. E. Mirsky, and M. M. Hatch, “Physiology and Neurobiology of Stress and Adaptation: Central Role of the Brain,” the American Physiological Society, pp. 873–904, 2007, doi: 10.1152/physrev.00041.2006.-The.
[12] G. P. Chrousos, “Stress and disorders of the stress system,” Jul. 2009. doi: 10.1038/nrendo.2009.106.
[13] M. Kivimäki, M. Virtanen, M. Elovainio, A. Kouvonen, A. Väänänen, and J. Vahtera, “Work stress in the etiology of coronary heart disease - A meta-analysis,” Scand J Work Environ Health, vol. 32, no. 6, pp. 431–442, 2006, doi: 10.5271/sjweh.1049.
[14] M. Malik et al., “Heart rate variability: Standards of measurement, physiological interpretation, and clinical use,” Circulation, vol. 93, no. 5, pp. 1043–1065, Mar. 1996, doi: 10.1161/01.cir.93.5.1043.
[15] J. F. Thayer and R. D. Lane, “A model of neurovisceral integration in emotion regulation and dysregulation,” 2000. [Online]. Available: www.elsevier.com/locate/jad.
[16] A. K. Jain, R. P. Duin, J. Mao, and S. Member, “Statistical Pattern Recognition: A Review,” IEEE Trans Pattern Anal Mach Intell, vol. 22, no. 1, pp. 4–37, Jan. 2000.
[17] Isabelle Guyon and Andr´e Elisseeff, “An Introduction to Variable and Feature Selection,” Journal of Machine Learning Research, vol. 1, pp. 1157–1182, 2003, doi: 10.1162/153244303322753616.
[18] G. Chandrashekar and F. Sahin, “A survey on feature selection methods,” Computers and Electrical Engineering, vol. 40, no. 1, pp. 16–28, Jan. 2014, doi: 10.1016/j.compeleceng.2013.11.024.
[19] Y. Saeys, I. Inza, and P. Larrañaga, “A review of feature selection techniques in bioinformatics,” Oct. 01, 2007, Oxford University Press. doi: 10.1093/bioinformatics/btm344.
[20] S. T. Charles and L. L. Carstensen, “Social and emotional aging,” Annu Rev Psychol, vol. 61, pp. 383–409, Jan. 2010, doi: 10.1146/annurev.psych.093008.100448.
[21] Ohal, Hemlata, et al. "Electroencephalogram Based Stress Detection Using Machine Learning." International Conference on Computational Intelligence in Data Science. Cham: Springer Nature Switzerland, 2024, doi: 10.1007/978-3-031-69986-3_38.
[22] R. Mukkamala et al., “Toward Ubiquitous Blood Pressure Monitoring via Pulse Transit Time: Theory and Practice,” IEEE Trans Biomed Eng, vol. 62, no. 8, pp. 1879–1901, Aug. 2015, doi: 10.1109/TBME.2015.2441951.
[23] A. Shahid, K. Wilkinson, S. Marcu, and C. M. Shapiro, “Perceived Stress Questionnaire (PSQ),” in STOP, THAT and One Hundred Other Sleep Scales, New York, NY: Springer New York, 2011, pp. 273–274. doi: 10.1007/978-1-4419-9893-4_64.
[24] ISMA, “Stress Questionnaire,” 2017. [Online]. Available: www.isma.org.uk.
[25] J. Sharma, N. Gupta, N. Khandelwal, and M. Gautam, “A Multimodal Approach for Stress Detection through Questionnaire and Emotion Analysis.,” in 2024 15th International Conference on Computing Communication and Networking Technologies, ICCCNT 2024, Institute of Electrical and Electronics Engineers Inc., 2024. doi: 10.1109/ICCCNT61001.2024.10725040.
[26] C. Brouwers et al., “Positive affect dimensions and their association with inflammatory biomarkers in patients with chronic heart failure,” Biol Psychol, vol. 92, no. 2, pp. 220–226, Feb. 2013, doi: 10.1016/j.biopsycho.2012.10.002.
[27] A. Baharum, S. H. Tanalol, C. X. Jian, M. Omar, N. A. M. Noor, and N. M. M. Yusop, “Stress catcher application for mobile stress monitoring using questionnaire-based,” Indonesian Journal of Electrical Engineering and Computer Science, vol. 16, no. 2, pp. 917–924, 2019, doi: 10.11591/ijeecs.v16.i2.pp917-924.
[28] M. Abd Al-Alim, R. Mubarak, N. M. Salem, and I. Sadek, “A machine-learning approach for stress detection using wearable sensors in free-living environments,” Comput Biol Med, vol. 179, Sep. 2024, doi: 10.1016/j.compbiomed.2024.108918.
[29] P. Mukherjee and A. Halder Roy, “A deep learning-based approach for distinguishing different stress levels of human brain using EEG and pulse rate,” Comput Methods Biomech Biomed Engin, vol. 27, no. 16, pp. 2303–2324, Dec. 2024, doi: 10.1080/10255842.2024.2275547.
[30] J. Allen, “Photoplethysmography and its application in clinical physiological measurement,” Mar. 01, 2007, Institute of Physics Publishing. doi: 10.1088/0967-3334/28/3/R01.
[31] C. Setz, B. Arnrich, J. Schumm, R. La Marca, G. Tröster, and U. Ehlert, “Discriminating stress from cognitive load using a wearable eda device,” IEEE Transactions on Information Technology in Biomedicine, vol. 14, no. 2, pp. 410–417, Mar. 2010, doi: 10.1109/TITB.2009.2036164.
[32] Fatangare, Mrunal, and Sukhada Bhingarkar. "Investigating an Efficient Filter to Implement Automated Nadi Pariksha by Analysing Time and Frequency Domain Features with Machine Learning Approach." International Research Journal of Multidisciplinary Technovation 7.2 (2025): 223-244, doi: 10.54392/irjmt25216.
[33] A. Singh, K. Singh, A. Kumar, A. Shrivastava, and S. Kumar, “Machine Learning Algorithms for Detecting Mental Stress in College Students,” Dec. 2024, doi: 10.1109/I2CT61223.2024.10544243.
[34] S. Sriramprakash, V. D. Prasanna, and O. V. R. Murthy, “Stress Detection in Working People,” in Procedia Computer Science, Elsevier B.V., 2017, pp. 359–366. doi: 10.1016/j.procs.2017.09.090.
[35] S. Khalid, “A Survey of Feature Selection and Feature Extraction Techniques in Machine Learning,” 2014. doi: 10.1109/SAI.2014.6918213 [Online]. Available: www.conference.thesai.org.
[36] Awasthi, Kushagra, Pranav Nanda, and K. V. Suma. "Performance analysis of Machine Learning techniques for classification of stress levels using PPG signals." 2020 IEEE International Conference on Electronics, Computing and Communication Technologies (CONECCT). IEEE, 2020. doi: 10.1109/CONECCT50063.2020.9198481.
[37] T. M. Cover and P. E. Hart, “Approximate formulas for the information transmitted bv a discrete communication channel,” 1952. Doi: 10.1109/TIT.1967.1053945.
[38] H. S. Kim, M. Kim, J. Kim, K. Park, D. S. Yoon, and J. Jo, “A Study on the Subjective Questionnaire-based Stress Assessment using k-means Clustering,” in International Conference on ICT Convergence, IEEE Computer Society, 2022, pp. 2131–2133. doi: 10.1109/ICTC55196.2022.9953015.
[39] V. Kalai Vani and F. Ghouse, “Exploring Predictive Models for Stress Detection: A Machine Learning Approach,” Institute of Electrical and Electronics Engineers (IEEE), Sep. 2024, pp. 477–483. doi: 10.1109/icipcn63822.2024.00084.
http://jist.acecr.org ISSN 2322-1437 / EISSN:2345-2773 |
Journal of Information Systems and Telecommunication
|
A Holistic Approach to Stress Identification: Integrating Questionnaires and Physiological Signals through Machine Learning |
Mrunal Fatangare1, Hemlata Ohal1*
|
1.School of Engineering and Technology, Dr. Vishwanath Karad MIT World Peace University, Pune, India
|
Received: 14 Oct 2024/ Revised: 04 May 2025/ Accepted: 07 Jun 2025 |
|
Abstract
This research project presents a comprehensive methodology for stress identification by combining subjective self-report data and objective physiological signals. The proposed system employs a carefully designed questionnaire, tailored to different age groups, to enhance accuracy in stress assessment. Subjects respond to the questionnaire, providing valuable insights into their emotional well-being. Subsequently, physiological data is collected using an infrared (IR) sensor positioned beneath the wrist, close to the artery. The pulse data obtained is meticulously converted into a CSV file, allowing for efficient preprocessing. The preprocessing phase ensures the integrity of the data, preparing it for machine learning (ML) analysis. The study harnesses ML techniques, specifically SVM (Support Vector Machines) & KNN (K-Nearest Neighbors), to classify stress levels based on the pre-processed data. Through feature extraction, relevant patterns are identified, contributing to the accurate characterization of stress states. This integrative approach offers a robust framework for stress assessment, taking into account both subjective and physiological dimensions.
Results demonstrate promising accuracy levels: Support Vector Machine (SVM) Reached a level of precision of 0.98 (+/- 0.20), Decision Tree showed 0.93 (+/- 0.30), and K-Nearest Neighbors (KNN) reached 0.88 (+/- 0.44). It also implements the voting classifier for improved performance of 98.6% of accuracy. These findings underscore the effectiveness of the proposed methodology in accurately identifying stress levels. Integrating subjective insights with objective physiological data not only enhances stress identification but also offers a comprehension of the intricate correlation between mental states and physiological reactions. This comprehensive strategy holds substantial implications across diverse domains such as healthcare, psychology, and human-computer interaction.
Keywords: Stress Identification; PPG, Age-Specific Assessment; Data Preprocessing; SVM; Feature Extraction; Classification Techniques; KNN; Stress Assessment; Well-being.
1- Introduction
In today's fast-paced society, stress has become an unavoidable aspect of daily existence, impacting individuals of all age ranges. Acknowledging the crucial influence of stress on mental health and overall welfare, there is an increasing need for effective stress detection systems [1]. This research endeavors to present an innovative and integrative approach to stress assessment by combining self-reported data through a comprehensive questionnaire and physiological measurements using pulse data collected via an infrared (IR) sensor [2].
Stress, a complex physiological and psychological phenomenon, is often characterized by feelings of fear, anxiety, and helplessness [3]. Stressor is an event or situation due to which stress is generated. Stress may lead to damage of usual physical functions and as well as developing few pathological conditions [3].
The causes of stress are shown in figure 1. The stress is also of three different types, good stress, bad stress and neutral stress. Good stress, also called eustress, has a positive impact on your physiological conditions. It may improve your performance. One may get motivated in a few situations maybe like getting married or awarded. Bad stress, also called distress, may degrade one’s performance. It has a bad impact on the physiological conditions. Examples are getting diagnosed with some disease, losing job, falling under some calamity. The neutral stress is also called neustress. It is associated with events which are not good, not bad for someone. Examples are storms in other countries, sudden increase in birthrate of your country.
If you know the stress symptoms, then treating stress becomes easy. Those can broadly be classified as cognitive, emotional, physical, and behavioral symptoms. The treatment of the stress can start only after stress detection. These symptoms are very important in the case of detection. These symptoms do change the physiological parameters of the individual.
Figure 1. Causes of stress [3], [4]
It is well known fact that stress is a “flight – or – fight” kind of response. It changes the physical and physiological parameters in the human body and that’s why stress stands out to be a very important thing to be cured. Figure 2 shows various symptoms related to stress. One needs to understand that one present situation may be very stressful for one individual and the same situation may be very challenging and motivating for another individual. And so, the response to the same situation is different from different individuals.
Figure 2. Stress Symptoms [3], [4]
In recent years, statistical data has highlighted the widespread prevalence of stress and its impact on individuals' well-being. This study aims to detect stress of an individual. As several individuals are suffering from stress, if it is detected well in advance and treated, then the impact of stress can be reduced. To ensure the accuracy and reliability of stress detection, our study employs a questionnaire tailored with the help of psychologists. The questionnaire delves into behavioral, emotional, and cognitive aspects related to test anxiety, procrastination, physical symptoms, and the overall impact of stress on academic and career pursuits. Respondents rate their experiences on a scale, providing a quantitative foundation for subsequent analysis. The physiological data collection involves the use of an IR sensor placed below the wrist, capturing pulse data through the underlying artery. Utilizing computational techniques such as SVM (Support Vector Machines) & KNN (K-Nearest Neighbors) in the realm of machine learning constitutes a fundamental aspect of our approach. These algorithms play a pivotal role in our classification tasks, assisting in identifying patterns and relationships within the pre-processed data. This facilitates the extraction of significant features that serve as indicators of stress, ultimately contributing to the development of a robust stress detection system. Our study innovates a different methodology to combine the psychological (through questionnaire response) and physiological (through the sensor placed on wrist pulse) parameters on one subject to detect the stress accurately for the treatment.
The remaining article is arranged as, section II discourses the literature in the same area, section III presents the proposed work and its methodology, section IV deals with the results and its discussion, whereas section V concludes this article.
2- Literature Survey
Our review of the literature is organized into distinct sections, each focusing on critical aspects of our study. These categories include subjective evaluation for stress identification, physiological indicators for stress detection, machine learning methodologies for stress classification, feature extraction and selection techniques in stress analysis, and age-specific approaches to stress assessment. Each division offers valuable insights into comprehending and managing stress from various angles, spanning subjective perceptions to objective physiological responses, and from computational strategies to developmental considerations. Through an in-depth exploration of these diverse facets of stress assessment, our literature review establishes a solid groundwork for our research, steering our investigation towards innovative strategies and interventions in the domain of stress management.
Cohen et al. developed a widely utilized measure of perceived stress, involving self-reporting on overall stress levels [5]. Masuda et al. devised the Social Readjustment Rating Scale to quantify stress based on significant life events [6]. Lovibond et al. introduced the Depression Anxiety Stress Scales to assess emotional states related to stress [7]. These subjective assessment tools encompass various dimensions of stress experiences, including cognitive, emotional, and behavioral aspects [8].
Researchers present a biologically inspired model for optimal fear detection through facial expression analysis. Utilizing a four-layer computational approach, the model demonstrates superior performance [8].The research by Uddin et al. [9] focuses on identifying human stress levels in industrial workers using electroencephalogram (EEG) signals. Using a hybrid feature analysis and a two-layered autoencoder neural network; achieves stress detection. The work introduced a novel multimodal hierarchical weighted framework for detecting emotional distress vocal, and verbal cues. The framework uses residual networks and CNNs for facial cues, LSTM and CNN for audio, and a BERT transformer for text [10].
Physiological signals offer objective measures of stress responses. McEwen et al. discussed the physiology and neurobiology of stress, emphasizing the brain's role in stress responses [11]. Chrousos et al. focused on stress system dysregulation leading to stress-related disorders [12]. Kivimäki et al. conducted a meta-analysis linking work stress to cardiovascular disease, demonstrating the physiological impact of chronic stress [13]. The Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology established heart rate variability standards, a physiological stress marker [14]. Thayer et al. proposed a neurovisceral integration model explaining how the autonomic nervous system regulates emotions and physiological responses to stress [15].
Machine learning algorithms are vital for classifying stress using physiological data. Jain et al. reviewed statistical pattern recognition techniques forming the basis of many machine learning algorithms [16].
Feature extraction and selection are crucial for identifying patterns in physiological stress data. Guyon et al. introduced variable and feature selection methods like filter, wrapper, and embedded approaches [17]. Uddin Chandrashekar et al. surveyed feature selection methods, comparing them based on computational complexity and effectiveness in stress research [18]. Saeys et al. reviewed feature selection techniques in bioinformatics, applicable to physiological data analysis in stress research [19].
Charles et al. reviewed social and emotional ageing processes, highlighting age-related changes in stress reactivity and regulation [20]. Ohal et al. in previous work have explored other physiological signal - to detect the stress from the signal slowing characteristics of electroencephalogram signals [21].
3- Proposed Methodology
This research paper introduces a methodology for stress detection by amalgamating subjective assessments via questionnaires and physiological data acquired through a photoplethysmogram (PPG) sensor. The objective is to enhance stress detection accuracy by customizing the data collection process for distinct age groups. The proposed methodology encompasses participant consent, age-specific questionnaire administration, pulse data collection utilizing an IR sensor, and subsequent data processing and analysis employing machine learning algorithms such as SVM and KNN. The system architecture, illustrated in the accompanying figure 3, demonstrates the sequence of data flow and processing stages integral to the proposed methodology.
This architectural depiction elucidates the integration and processing of subjective assessments alongside physiological data to achieve precise stress level detection.
Figure 3. System Architecture Diagram
Stress is a pervasive health concern impacting individuals across diverse age groups. This research aims to construct an effective stress detection system by integrating subjective responses and physiological data, specifically, pulse data obtained through a PPG sensor [22].
3-1- Questionnaire Design
While designing the questionnaire, authors have consulted the psychologist, therapist as well as a few research papers which gave a thoughtful insight for the questionnaire design [23], [24], [25] . The Age-Specific Questionnaire: Devise and administer a questionnaire tailored to different age groups to gather subjective stress assessments. The questionnaire seeks to capture varied perspectives on stress experiences among participants. The questionnaire comprises 40 questions aimed at gauging levels of test anxiety. The questionnaire aims to assess the prevalence and impact of test anxiety across various aspects of an individual's academic and personal life. It consists of 40 questions divided into different categories, covering aspects such as the frequency of test anxiety experiences, coping mechanisms, academic performance, and overall well-being.
A detailed overview of the questionnaire, including its categories and sample questions, is provided in Table 1.
Table 1. Overview of Questionnaire
Category | Sample Questions |
Frequency of Test Anxiety | a. Do you encounter feelings of fear, anxiety, or helplessness before or during an examination? b. Do you often experience a sense of dread or impending doom before a test? |
Impact on Academic Performance | a. Have you performed poorly on a test in the past and fear of repeating the Performance? b. Does your test anxiety affect your overall academic or career goals? |
Coping Mechanisms | a. Have you tried relaxation techniques or mindfulness practices to cope with test anxiety? b. Do you frequently compare your abilities to others when preparing for a test? |
Physical Symptoms and Health Impact | a. Do you feel nauseous, sweaty, or experience a rapid heartbeat, shortness of breath, or dizziness during an examination? b. Are you concerned that your test anxiety may negatively impact your long-term memory recall during exams? |
Social and Emotional Impact | a. Do you feel embarrassed or ashamed when discussing test anxiety with others? b. Have you experienced a decrease in self-esteem due to test-related stress? |
Behavioral Patterns | a. Have you ever resorted to cheating or unethical behavior during a test due to extreme test anxiety? b. Are you concerned that your test anxiety may negatively impact your future career opportunities for advancement? |
3-2- Questionnaire Data Collection
Participant Consent: Obtain informed consent from participants, clearly explaining the study's objectives, data collection techniques, and confidentiality protocols. Stress the voluntary nature of participation and the freedom to withdraw at any point.
While collecting data, age, gender and occupation is asked to the participants.
3-3- Questionnaire Responses
a. Participants are required to indicate the frequency of experiencing various symptoms or behaviors associated with test anxiety using a scale ranging from 'Never' to 'Very Often'.
b. This scoring system in Table 2 quantifies test anxiety levels, aiding in a thorough analysis of stress experiences among participants from diverse age groups.
c. The responses received for this questionnaire clearly show that for the same situation different reactions/ responses are there, which proves the theory behind stress.
Table 2. Details of Scoring based on Categories
Response | Scoring points |
Never & 'Rarely | Combined, these responses equate to 0 points. |
Sometimes | This response amounts to 1 point |
Often' & 'Very Often | Combined, these responses equate to 2 points. |
Based on the total score derived from the responses, individuals can be classified into three groups:
a. Not Stressed: A score below 15 suggests that the individual is not experiencing significant levels of test anxiety and can proceed with their activities calmly.
b. Mild to Moderate Stress: Falling between 15 and 30, this score indicates a moderate level of test anxiety. Though not severely affected, the individual may benefit from practicing relaxation techniques and stress management strategies.
c. High Stress, Guidance Required: A score exceeding 30 indicates high levels of test anxiety, necessitating guidance and intervention to address the substantial impact of stress on the individual's well-being and academic performance. Figure 4 describes the strategy used to identify the stress level.
Figure 4. Stress level identification and label generation
The labels for the pulse data csv are generated from the questionnaire responses. These labels are not stressed, moderately stressed, and highly stressed. "High Stress, Guidance Required" for individuals with test anxiety scores surpassing 30, "Mild to Moderate Stress" for scores between 15 to 30 and "Not Stressed" for scores of 15 or below. This process allows the decision tree to categorize participants into various stress levels based on their test anxiety scores, offering a clear visual depiction of the decision-making process and facilitating comprehension of how data is categorized. By analyzing the responses and tallying the total score, individuals can gain insight into their level of test anxiety and take appropriate measures to manage it effectively[26] [27].
3-4- Wrist Pulse Data Collection
Many researchers have tried guessing the stress level using wearable devices. In these devices the sensors are the most important part [28], [29].
a. PPG Sensor Placement: Proper placement of the Photoplethysmography (PPG) sensor underneath the wrist is vital for precise pulse data acquisition. This sensor functions by detecting alterations in blood volume through the emission of light into the skin and subsequent measurement of the reflected or transmitted light. Focusing on the artery, commonly the radial artery located in the wrist, guarantees accurate measurement of pulse rate and waveform attributes [30]. For visual guidance on sensor placement, please refer to Figure. 5.
Figure 5. PPG Sensor Placement on Wrist
b. Pulse Data Collection: After positioning the PPG sensor correctly, it continuously monitors variations in blood volume linked to each heartbeat. Subsequently, these measurements are captured at consistent intervals and archived in a CSV (Comma-Separated Values) file format. Each entry within the CSV file usually includes a time-series dataset alongside the corresponding pulse rate value, thereby constructing a time-series dataset suitable for subsequent analysis [31].
3-5- Data Preprocessing
a. Interpolation: To handle missing or irregularly sampled pulse data points, interpolation techniques are utilized. These gaps may occur due to sensor malfunctions, motion artefacts, or technical glitches. Interpolation methods like linear or spline interpolation estimate the absent values by considering adjacent data points, thereby ensuring a continuous and evenly sampled dataset suitable for analysis.
b. Savitzky-Golay Filter: The application of the Savitzky-Golay filter aims to smooth and remove noise from the pulse data. Known for its effectiveness in eliminating high-frequency noise while retaining essential signal features like the underlying pulse waveform, this filter enhances data quality by reducing artifacts. Consequently, the accuracy of subsequent analytical procedures is improved [32]. For a visual representation of the data before and after preprocessing, please refer to Figure. 6. This graph represents the filtered data for subject 1. After applying the this filter, the data becomes smoother, with reduced noise while preserving key features like peaks and edges. This Savitzky-Golay filter also has maintained the shape and height of signals as shown in graph of Figure 6.
Figure 6. Graph of Original & Pre-processed Data
3-6- Feature Extraction
Time Domain Features: Time domain analysis entails examining signal characteristics along the time axis. In pulse signal analysis, these features offer insights into the temporal aspects of the heartbeat waveform, directly derived from pulse data. These features encompass metrics like average pulse rate, variance, median, maximum and minimum values, range, and Root Mean Square of Successive Differences (RMSSD). They offer an understanding of the overall characteristics of the pulse signal, depicting alterations in heart rate and variability across time.
List of Features Extracted:
a. Average Pulse Rate: The mean of pulse intervals over a specific time, indicating heart rate centrality.
b. Variance: Measures pulse interval dispersion around the mean, reflecting heart rate variability.
c. Median: The middle value of pulse intervals, providing a robust centrality measure.
d. Maximum and Minimum Values: Identifies peak and trough pulse intervals, showing heart rate fluctuation range.
e. Range: Difference between maximum and minimum pulse intervals, indicating overall heart rate variation.
f. Root Mean Square of Successive Differences (RMSSD): Quantifies variation in successive pulse intervals, reflecting short-term heart rate variability.
Frequency Domain Features: Frequency domain analysis examines signal frequency components and distribution. In pulse signal analysis, these features reveal heart rate variability's spectral characteristics, reflecting autonomic nervous system activity [33].
Frequency domain features are calculated to assess the spectral attributes of the pulse signal. These features encompass details concerning the energy distribution across various frequency bands, encompassing metrics of heart rate variability (HRV). The analysis in the frequency domain offers an understanding of the activity of the autonomic nervous system and can unveil patterns linked with stress and physiological arousal [34].
List of Features Extracted:
a. Power Spectral Density (PSD): Estimation via Fast Fourier Transform (FFT) or Autoregressive (AR) modelling decomposes pulse signals into frequency bands (VLF, LF, HF).
b. Total Power: Overall power across frequency bands, reflecting heart rate variability.
c. LF/ HF Ratio: Ratio of LF band power to HF band power, indicating sympathovagal balance.
d. Normalized LF and HF Power: LF and HF power in normalized units, facilitating comparison across individuals or conditions.
e. Peak Frequency: Identifies the spectrum's highest power frequency, showing the dominant heart rate oscillatory component.
Integrating time and frequency domain features offers a comprehensive understanding of heart rate dynamics under various conditions. These serve as valuable markers for assessing cardiovascular health, stress responses, and autonomic nervous system function[35].
For a visual representation of the Plot of Selected Features please refer to Figure 7.
Figure 7. Plot of Selected Features
3-7- Feature Selection
During the feature selection phase, the aim is to pinpoint the most pertinent and informative features crucial for stress detection from the pool of extracted features.
This entails scrutinizing the correlation among features, evaluating their discriminative capability, and cherry-picking a subset of features that effectively distinguish between stress and non-stress scenarios. This approach aids in curtailing dimensionality and computational intricacies, thereby bolstering the efficacy of the classification model [35].
3-8- Labeling the data
The pulse feature data of every subject is stored in the CSV format. The data is then labeled as per the result obtained during the questionnaire responses stage. These labels are not stressed, moderately stressed, and highly stressed.
Summarizing the proposed methodology, emphasizing its potential for accurate stress detection by integrating subjective assessments with physiological data. Highlighting the significance of the age-specific approach in improving system performance. This proposed methodology establishes the groundwork for a comprehensive stress detection system, combining subjective and physiological data collection, preprocessing, feature extraction, feature selection, and classification methodologies to establish a comprehensive framework for a nuanced understanding of stress across various age groups. By harnessing sophisticated analytical techniques and machine learning algorithms, this approach furnishes a robust and precise mechanism for identifying stress patterns from physiological signals
4- Result and Discussion
4-1- Classification Techniques
1. Support Vector Machine (SVM): The Support Vector Machine (SVM) stands as a supervised learning technique employed for classification purposes. Concerning stress detection, SVM scrutinizes the chosen features and creates a hyperplane that segregates stress and non-stress occurrences within the feature space. The objective of SVM is to widen the margin between diverse classes while mitigating classification errors, thus furnishing a sturdy and efficient approach for stress classification [36].
2. k-Nearest Neighbors (KNN): The k-Nearest Neighbors (KNN) algorithm is a non-parametric classification method that categorizes instances based on their resemblance to nearby data points. In stress detection applications, KNN assesses the proximity of a specific instance to its closest Neighbors within the feature space and assigns it to the predominant class among those Neighbors. Renowned for its simplicity and adaptability, KNN is well-suited for various classification tasks, including stress detection [37] [38].
3. Decision Trees (DT): The Decision Tree (DT) is a robust classification method that divides the feature space into smaller regions, each corresponding to a specific class label. Their transparency in decision-making is particularly advantageous in domains like healthcare and finance. DTs can effectively handle both numerical and categorical data without extensive preprocessing and can capture nonlinear relationships. However, they are prone to overfitting, which can be addressed through techniques such as pruning or ensemble methods. In summary, DTs offer a valuable combination of simplicity, interpretability, and performance, making them a useful tool in various classification tasks, including stress detection [39].
4-2- Results
The methodology proposed for stress identification, which integrates subjective self-report data and objective physiological signals, has yielded promising outcomes. After data collection, preprocessing, feature extraction, and classification using machine learning algorithms, the accuracy of stress level classification was evaluated. For a visual representation of the data for Normal & Stressed Subjects, please refer to Figure. 8.
The Support Vector Machine (SVM) classifier achieved an accuracy of 0.98 with a standard deviation of ±0.20, indicating precise identification of stress levels. Moreover, the Decision Tree classifier attained an accuracy of 0.93 (±0.30), while the K-Nearest Neighbors (KNN) algorithm demonstrated an accuracy of 0.88 (±0.44). Table 3 demonstrates the accuracy of these classifiers. These results underscore the effectiveness of the proposed methodology in accurately classifying stress levels based on both subjective and physiological data.
The high accuracy levels achieved by the SVM classifier underscore the robustness of the proposed methodology in stress identification. SVM's capability to establish a hyperplane effectively separating stress and non-stress instances within the feature space contributes to its superior performance. Though slightly less accurate, the decision tree classifier provides an intuitive representation of the decision-making process, aiding in the interpretation of stress classification outcomes.
Figure 8. Graph of Data for Normal & Stressed Subjects
Table 3. Classifier accuracy
Classifier | Accuracy |
SVM | 0.98 (+/- 0.20) |
Decision Tree | 0.93 (+/- 0.30) |
KNN | 0.88 (+/- 0.44) |
Voting Classifier | 0.986 |
In this study, a Voting Classifier was employed to enhance stress detection accuracy by combining SVM, decision tree and KNN machine learning models. The classifier aggregated predictions using both hard voting and soft voting techniques, leveraging the strengths of individual models. Experimental results demonstrated that the Voting Classifier outperformed individual classifiers, achieving improved accuracy of 98.6 %.
The existing work of researchers have been discussed in upcoming paragraph and have been listed in Table 4.
Table 4. Comparison with Existing Work
Author | Method | Performance |
Awasthi et al [36] | SVM | 86% |
Awasthi et al [36] | Decision tree | 82% |
Kim et al [38] | K-means clustering | 87% |
Proposed | Voting classifier | 98.6% |
Awasthi et al [36], have trained SVM and decision tree as machine learning models for classification of stress and classified it as high stress vs average stress. This work has explored and used Pulse Rate Variability (PRV) as features to classify the stress and have achieved the accuracy of 86% and 82% respectively. The work of Kim et al [38] have experimented for the number of clusters to use for classification of stress using elbow method and have come to with optimal value of k as 12. This research work after SVM, KNN and decision tree have implemented Voting classifier to achieve the improved the results of 98.6% of accuracy.
The integration of subjective self-report data, acquired through tailored questionnaires, with objective physiological signals enhances the comprehensiveness and accuracy of stress assessment. By capturing both cognitive and physiological responses to stressors, the proposed methodology offers a holistic understanding of individuals' stress experiences.
The age-specific approach in questionnaire design acknowledges the diverse manifestations of stress across different life stages, thereby improving the relevance and accuracy of stress assessment. Additionally, the utilization of physiological signals, such as pulse data obtained through infrared sensors, adds an objective dimension to stress detection, reducing reliance on self-reported data alone.
The findings of this study have significant implications for various fields, including healthcare, psychology, and human-computer interaction. By providing a nuanced understanding of stress dynamics and effective stress detection mechanisms, the proposed methodology can inform the development of tailored interventions and support systems for individuals experiencing stress.
However, it's important to acknowledge some limitations of the study. The sample size and demographic diversity of the participants may impact the generalizability of the results. Further research is necessary to validate the proposed methodology across different populations and settings.
In conclusion, the integration of subjective and physiological data through machine learning techniques offers a promising approach to stress identification. By leveraging the strengths of both subjective and objective measures, the proposed methodology contributes to advancing the field of stress management and well-being.
The future work of this research is to enhance machine learning algorithms, so that the accuracy and reliability of stress detection from PPG signals can be significantly improved. Additionally, combining PPG data with other physiological signals, like ECG and GSR, can provide a more comprehensive understanding of stress. The work can be integrated further into personalized stress management applications and clinical applications of PPG sensors that can aid in the early detection and management of stress-related health issues, contributing to better healthcare outcomes.
5- Conclusion
Acknowledging the extensive research presented in this paper, it's evident that the work represents a significant advancement in stress identification methodologies. The integration of subjective self-report data with objective physiological signals, supported by machine learning techniques, provides a nuanced and comprehensive approach to stress assessment. This multifaceted methodology holds significant promise across various domains such as healthcare, psychology, and human-computer interaction.
In conclusion, this research paper introduces an innovative framework for stress detection that addresses the urgent need for precise and reliable assessment tools in today's dynamic environments. By tailoring questionnaires to different age groups and pairing them with sophisticated physiological data collection methods, the proposed methodology enables a holistic understanding of stress experiences. The integration of advanced machine learning algorithms further enhances the accuracy and effectiveness of stress identification, representing a significant advancement in the field.
This study not only contributes to advancing our understanding of stress but also emphasizes the importance of adopting multidimensional approaches for comprehensive assessment and intervention strategies. By combining subjective insights with objective physiological measurements, the proposed framework offers unprecedented depth and detail in stress assessment. Additionally, the technical rigor demonstrated in data preprocessing, feature extraction, and classification techniques highlights the robustness and reliability of the proposed methodology.
Looking forward, this research lays a solid foundation for further exploration and application in real-world contexts. As stress continues to impact individuals' well-being across diverse demographics, the integration of subjective and objective measures is poised to transform stress management and prevention strategies. By leveraging state-of-the-art technologies and methodologies, this study not only advances current research but also holds significant potential for practical implementations to improve mental health and overall quality of life.
In summary, this research represents a groundbreaking contribution to the field of stress identification, offering a sophisticated and scientifically validated methodology that promises to reshape our understanding and management of stress in contemporary society.
6- Conclusions
Acknowledging the extensive research presented in this paper, it's evident that the work represents a significant advancement in stress identification methodologies. The integration of subjective self-report data with objective physiological signals, supported by machine learning techniques, provides a nuanced and comprehensive approach to stress assessment. This multifaceted methodology holds significant promise across various domains such as healthcare, psychology, and human-computer interaction.
In conclusion, this research paper introduces an innovative framework for stress detection that addresses the urgent need for precise and reliable assessment tools in today's dynamic environments. By tailoring questionnaires to different age groups and pairing them with sophisticated physiological data collection methods, the proposed methodology enables a holistic understanding of stress experiences. The integration of advanced machine learning algorithms further enhances the accuracy and effectiveness of stress identification, representing a significant advancement in the field.
This study not only contributes to advancing our understanding of stress but also emphasizes the importance of adopting multidimensional approaches for comprehensive assessment and intervention strategies. By combining subjective insights with objective physiological measurements, the proposed framework offers unprecedented depth and detail in stress assessment. Additionally, the technical rigor demonstrated in data preprocessing, feature extraction, and classification techniques highlights the robustness and reliability of the proposed methodology.
Looking forward, this research lays a solid foundation for further exploration and application in real-world contexts. As stress continues to impact individuals' well-being across diverse demographics, the integration of subjective and objective measures is poised to transform stress management and prevention strategies. By leveraging state-of-the-art technologies and methodologies, this study not only advances current research but also holds significant potential for practical implementations to improve mental health and overall quality of life.
In summary, this research represents a groundbreaking contribution to the field of stress identification, offering a sophisticated and scientifically validated methodology that promises to reshape our understanding and management of stress in contemporary society.
Acknowledgments
We, the authors, want to acknowledge the efforts made by a few medical practitioners, who helped us a lot in understanding stress-related issues. We also thanks to researchers whom we referred while writing the paper.
References
[1] Alberdi, A. Aztiria, and A. Basarab, “Towards an automatic early stress recognition system for office environments based on multimodal measurements: A review,” Feb. 01, 2016, Academic Press Inc. doi: 10.1016/j.jbi.2015.11.007.
[2] L. Saturday and H. Selye, “Stress and the General Adaptation Syndrome,” Br Med J, pp. 1383–1392, Jun. 1950.
[3] S. Gedam and S. Paul, “A Review on Mental Stress Detection Using Wearable Sensors and Machine Learning Techniques,” 2021, Institute of Electrical and Electronics Engineers Inc. doi: 10.1109/ACCESS.2021.3085502.
[4] S. A. Singh, K. Gupta, M. Rajeshwari, and T. Janumala, “Detection of Stress Using Biosensors,” 2018. [Online]. Available:www.sciencedirect.comwww.materialstoday.com/proceedings
[5] S. Cohen, T. Kamarck, and R. Mermelstein, “A Global Measure of Perceived Stress,” 1983.
[6] M. Masuda and T. H. Holmes, “The Social Readjustment Rating Scale: A Cross-Cultural Study Of Japanese And Americans *p,” Pergamon Press, 1967.
[7] P. F. Lovibond and S. H. Lovibond, “The structure of negative Emotion States: Comparision of the depression anxiety Stress Scales (DASS) with the Beck Depression and Anxiety Inventories,” BehaP. Res. Ther, vol. 33, no. 3, pp. 335–343, 1995.
[8] E. Askari, “Fear Recognition Using Early Biologically Inspired Features Model,” Journal of Information Systems and Telecommunication (JIST), vol. 12, no. 45, pp. 12–19, Mar. 2024, doi: 10.61186/jist.39051.12.45.12.
[9] Jia Uddin, “An Autoencoder based Emotional Stress Detection approach using Electroencephalography Signals,” Journal of Information Systems and Telecommunication, vol. 11, no. 1, pp. 24–30, Mar. 2023.
[10] N. Jadhav, “Hierarchical Weighted Framework for Emotional Distress Detection using Personalized Affective Cues,” Journal of Information Systems and Telecommunication (JIST), vol. 10, no. 38, pp. 89–101, Apr. 2022, doi: 10.52547/jist.16499.10.38.89.
[11] B. S. Mcewen, A. E. Mirsky, and M. M. Hatch, “Physiology and Neurobiology of Stress and Adaptation: Central Role of the Brain,” the American Physiological Society, pp. 873–904, 2007, doi: 10.1152/physrev.00041.2006.-The.
[12] G. P. Chrousos, “Stress and disorders of the stress system,” Jul. 2009. doi: 10.1038/nrendo.2009.106.
[13] M. Kivimäki, M. Virtanen, M. Elovainio, A. Kouvonen, A. Väänänen, and J. Vahtera, “Work stress in the etiology of coronary heart disease - A meta-analysis,” Scand J Work Environ Health, vol. 32, no. 6, pp. 431–442, 2006, doi: 10.5271/sjweh.1049.
[14] M. Malik et al., “Heart rate variability: Standards of measurement, physiological interpretation, and clinical use,” Circulation, vol. 93, no. 5, pp. 1043–1065, Mar. 1996, doi: 10.1161/01.cir.93.5.1043.
[15] J. F. Thayer and R. D. Lane, “A model of neurovisceral integration in emotion regulation and dysregulation,” 2000. [Online]. Available: www.elsevier.com/locate/jad
[16] A. K. Jain, R. P. Duin, J. Mao, and S. Member, “Statistical Pattern Recognition: A Review,” IEEE Trans Pattern Anal Mach Intell, vol. 22, no. 1, pp. 4–37, Jan. 2000.
[17] Isabelle Guyon and Andr´e Elisseeff, “An Introduction to Variable and Feature Selection,” Journal of Machine Learning Research, vol. 1, pp. 1157–1182, 2003, doi: 10.1162/153244303322753616.
[18] G. Chandrashekar and F. Sahin, “A survey on feature selection methods,” Computers and Electrical Engineering, vol. 40, no. 1, pp. 16–28, Jan. 2014, doi: 10.1016/j.compeleceng.2013.11.024.
[19] Y. Saeys, I. Inza, and P. Larrañaga, “A review of feature selection techniques in bioinformatics,” Oct. 01, 2007, Oxford University Press. doi: 10.1093/bioinformatics/btm344.
[20] S. T. Charles and L. L. Carstensen, “Social and emotional aging,” Annu Rev Psychol, vol. 61, pp. 383–409, Jan. 2010, doi: 10.1146/annurev.psych.093008.100448.
[21] Ohal, Hemlata, et al. "Electroencephalogram Based Stress Detection Using Machine Learning." International Conference on Computational Intelligence in Data Science. Cham: Springer Nature Switzerland, 2024, doi: 10.1007/978-3-031-69986-3_38
[22] R. Mukkamala et al., “Toward Ubiquitous Blood Pressure Monitoring via Pulse Transit Time: Theory and Practice,” IEEE Trans Biomed Eng, vol. 62, no. 8, pp. 1879–1901, Aug. 2015, doi: 10.1109/TBME.2015.2441951.
[23] A. Shahid, K. Wilkinson, S. Marcu, and C. M. Shapiro, “Perceived Stress Questionnaire (PSQ),” in STOP, THAT and One Hundred Other Sleep Scales, New York, NY: Springer New York, 2011, pp. 273–274. doi: 10.1007/978-1-4419-9893-4_64.
[24] ISMA, “Stress Questionnaire,” 2017. [Online]. Available: www.isma.org.uk
[25] J. Sharma, N. Gupta, N. Khandelwal, and M. Gautam, “A Multimodal Approach for Stress Detection through Questionnaire and Emotion Analysis.,” in 2024 15th International Conference on Computing Communication and Networking Technologies, ICCCNT 2024, Institute of Electrical and Electronics Engineers Inc., 2024. doi: 10.1109/ICCCNT61001.2024.10725040.
[26] C. Brouwers et al., “Positive affect dimensions and their association with inflammatory biomarkers in patients with chronic heart failure,” Biol Psychol, vol. 92, no. 2, pp. 220–226, Feb. 2013, doi: 10.1016/j.biopsycho.2012.10.002.
[27] A. Baharum, S. H. Tanalol, C. X. Jian, M. Omar, N. A. M. Noor, and N. M. M. Yusop, “Stress catcher application for mobile stress monitoring using questionnaire-based,” Indonesian Journal of Electrical Engineering and Computer Science, vol. 16, no. 2, pp. 917–924, 2019, doi: 10.11591/ijeecs.v16.i2.pp917-924.
[28] M. Abd Al-Alim, R. Mubarak, N. M. Salem, and I. Sadek, “A machine-learning approach for stress detection using wearable sensors in free-living environments,” Comput Biol Med, vol. 179, Sep. 2024, doi: 10.1016/j.compbiomed.2024.108918.
[29] P. Mukherjee and A. Halder Roy, “A deep learning-based approach for distinguishing different stress levels of human brain using EEG and pulse rate,” Comput Methods Biomech Biomed Engin, vol. 27, no. 16, pp. 2303–2324, Dec. 2024, doi: 10.1080/10255842.2024.2275547.
[30] J. Allen, “Photoplethysmography and its application in clinical physiological measurement,” Mar. 01, 2007, Institute of Physics Publishing. doi: 10.1088/0967-3334/28/3/R01.
[31] C. Setz, B. Arnrich, J. Schumm, R. La Marca, G. Tröster, and U. Ehlert, “Discriminating stress from cognitive load using a wearable eda device,” IEEE Transactions on Information Technology in Biomedicine, vol. 14, no. 2, pp. 410–417, Mar. 2010, doi: 10.1109/TITB.2009.2036164.
[32] Fatangare, Mrunal, and Sukhada Bhingarkar. "Investigating an Efficient Filter to Implement Automated Nadi Pariksha by Analysing Time and Frequency Domain Features with Machine Learning Approach." International Research Journal of Multidisciplinary Technovation 7.2 (2025): 223-244, doi: 10.54392/irjmt25216
[33] A. Singh, K. Singh, A. Kumar, A. Shrivastava, and S. Kumar, “Machine Learning Algorithms for Detecting Mental Stress in College Students,” Dec. 2024, doi: 10.1109/I2CT61223.2024.10544243.
[34] S. Sriramprakash, V. D. Prasanna, and O. V. R. Murthy, “Stress Detection in Working People,” in Procedia Computer Science, Elsevier B.V., 2017, pp. 359–366. doi: 10.1016/j.procs.2017.09.090.
[35] S. Khalid, “A Survey of Feature Selection and Feature Extraction Techniques in Machine Learning,” 2014. doi: 10.1109/SAI.2014.6918213 [Online]. Available: www.conference.thesai.org
[36] Awasthi, Kushagra, Pranav Nanda, and K. V. Suma. "Performance analysis of Machine Learning techniques for classification of stress levels using PPG signals." 2020 IEEE International Conference on Electronics, Computing and Communication Technologies (CONECCT). IEEE, 2020. doi: 10.1109/CONECCT50063.2020.9198481
[37] T. M. Cover and P. E. Hart, “Approximate formulas for the information transmitted bv a discrete communication channel,” 1952. Doi: 10.1109/TIT.1967.1053945
[38] H. S. Kim, M. Kim, J. Kim, K. Park, D. S. Yoon, and J. Jo, “A Study on the Subjective Questionnaire-based Stress Assessment using k-means Clustering,” in International Conference on ICT Convergence, IEEE Computer Society, 2022, pp. 2131–2133. doi: 10.1109/ICTC55196.2022.9953015.
[39] V. Kalai Vani and F. Ghouse, “Exploring Predictive Models for Stress Detection: A Machine Learning Approach,” Institute of Electrical and Electronics Engineers (IEEE), Sep. 2024, pp. 477–483. doi: 10.1109/icipcn63822.2024.00084.
* Hemlata Ohal
hemlata.ohal@mitwpu.edu.in
