STRESS DETECTION IN IT PROFESSIONALS

USING IMAGE PROCESSING AND MACHINE LEARNING

Anusree K Keerthy M
B-Tech Student (Computer Science and Business System) B-Tech Student (Computer Science and Business System)
Dhanalakshmi Srinivasan College of Engineering Dhanalakshmi Srinivasan College of Engineering
Anna University Anna University
Tamilnadu, India Tamilnadu, India
anusreek20@dsce.ac.in keerthym20@dsce.ac.in

Amal E B Abinav N P
B-Tech Student (Computer Science and Business System) B-Tech Student (Computer Science and Business System)
Dhanalakshmi Srinivasan College of Engineering Dhanalakshmi Srinivasan College of Engineering
Anna University Anna University
Tamilnadu, India Tamilnadu, India
amaleb20@dsce.ac.in abinavnp20@dsce.ac.in

Ms. R. Reshma
Asst. Professor (Computer Science Engineering)
Dhanalakshmi Srinivasan College of Engineering
Anna University
Tamilnadu, India
reshma.r@dsce.ac.in

I. INTRODUCTION
Abstract— The IT industry faces a significant In contemporary work environments, stress is a
challenge: stress, which can adversely impact pervasive issue affecting the well-being and productivity
productivity, mental health, and overall well-being of professionals, particularly within the IT industry. The
among professionals. Conventional stress assessment demanding nature of IT roles, coupled with tight
methods often rely on subjective self-reporting, deadlines and high-pressure situations, often leads to
making them susceptible to biases. To address this elevated stress levels among employees. Consequently,
issue, our work introduces a pioneering approach that effective stress detection and management systems are
utilizes image processing and machine learning
crucial for fostering a healthy and productive workplace
techniques for stress detection in IT professionals.
environment. Traditional stress assessment methods,
Our innovative system aims to automatically detect reliant on subjective self-reporting or sporadic medical
stress levels by analyzing facial expressions captured evaluations, may not suffice in providing timely insights
through images or videos. Leveraging cutting-edge or personalized interventions to mitigate stress-related
image processing, the system extracts pertinent facial risks.
features associated with stress, such as muscle tension,
eye movements, and changes in expression. Machine
learning algorithms, including artificial neural To address the shortcomings of conventional stress
networks (ANN) and facial recognition techniques, detection approaches, this paper proposes an innovative
then analyze these features to classify individuals' system utilizing advanced image processing and machine
stress levels. learning techniques. By leveraging cutting-edge
technology, our system aims to automate the detection of
Index Terms— Stress detection, IT professionals, stress levels in IT professionals, offering real-time
image processing, machine learning, real-time monitoring and personalized interventions to mitigate
monitoring, socio-economic impact, mental health, stressors. The integration of image processing allows for
well-being, productivity, physiological signals, facial the analysis of facial expressions and other physiological
expressions, stress management systems, workplace signals associated with stress, providing a more
environment, health risks.
comprehensive understanding of employees' well-being.
 Furthermore, the socio-economic impact of stress affective states, including neutral, relaxed, and
on individuals and organizations underscores the stressed/anxious states, the study establishes a
importance of effective stress management strategies. By comprehensive experimental protocol. The analysis
implementing robust stress detection systems, employers primarily focuses on non-voluntary and semi-voluntary
can proactively identify and address stressors, thereby facial cues to objectively estimate emotional
fostering a healthier and more productive work representations. This approach offers a valuable
environment. This paper explores the intersection of contribution to stress detection by providing a more
image processing, machine learning, and stress nuanced understanding of emotional states based on
management to develop innovative solutions tailored to facial expressions .
the unique challenges faced by IT professionals in
managing stress.
Paper 2: Detection of Stress Using Image Processing and
Machine Learning Techniques:
Moreover, the World Health Organization's
recognition of stress as a prevalent mental health issue This paper introduces a real-time, non-intrusive
highlights the urgency of developing effective stress system for capturing video data and detecting the
management solutions. Stress not only affects individual emotional status of individuals by analyzing facial
employees but also impacts organizational performance, expressions. Utilizing image processing and machine
learning techniques, the system identifies emotional
leading to decreased productivity, increased absenteeism,
states in each video frame and makes stress level
and higher healthcare costs. Therefore, investing in
determinations over sequential hours of captured video.
advanced stress detection technologies is not only
The system employs advanced techniques to train models
beneficial for employees' well-being but also for the and analyze differences in predicting stress-related
overall success and sustainability of businesses in the IT features, enhancing the accuracy and efficiency of stress
sector and beyond. detection.

In recent years, advancements in image processing

and machine learning have revolutionized various Paper 3: Machine Learning Techniques for Stress
industries, offering unprecedented opportunities for Prediction in Working Employees:
innovation and optimization. By harnessing these In this paper, machine learning techniques are applied
technologies specifically for stress detection in IT to analyze stress patterns in working adults and identify
professionals, organizations can stay at the forefront of key factors influencing stress levels. Through meticulous
workplace wellness initiatives. Additionally, the ability to data cleaning and preprocessing, various machine
provide real-time feedback and personalized interventions learning algorithms are trained to predict stress levels
based on employees' stress levels can significantly effectively. By narrowing down the determinants of
enhance job satisfaction, retention rates, and overall stress, this approach offers valuable insights for
organizational resilience. developing targeted interventions and stress management
strategies in workplace settings.
In summary, this paper seeks to bridge the gap
between technological innovation and employee well-
III. EXISTING SYSTEM
being by proposing an integrated approach to stress
detection in the IT industry. Through the convergence of
image processing, machine learning, and stress In the existing system, stress detection relies on
management principles, our system aims to empower digital signal processing, incorporating various
organizations to create healthier, more supportive work physiological indicators such as Galvanic skin response,
environments conducive to employee success and blood volume, pupil dilation, and skin temperature.
organizational growth. Additionally, other approaches utilize a combination of
physiological signals and visual features, such as eye
closure and head movement, to monitor stress levels in
II. LITERATURE SURVEY individuals during work activities. However, these
Paper 1: Stress and anxiety detection using facial cues methods often involve intrusive measurements and may
from videos: be less comfortable for real-world application. Each
sensor's data is compared against a predefined stress
This study pioneers a framework for detecting and index, serving as a threshold value to determine the
analyzing stress emotional states through video-recorded individual's stress level.
facial cues. By inducing systematic variability in
 DIS-ADVANTAGES OF EXISTING SYSTEM
Preventative Stress Management Solutions: The Stress
The existing system for stress detection relies Detection System empowers employees to cope with
heavily on physiological signals, which, while their stress issues by offering preventative stress
informative, present several drawbacks. One significant management solutions. By identifying stress early and
limitation is the non-stationary nature of these signals providing interventions, the system helps employees
over time, making their analysis challenging. This mitigate stressors and maintain well-being.
variability can lead to inconsistencies in stress detection
accuracy, as the signals may fluctuate unpredictably. Regular Image Capture: Images of employees are
Moreover, the current approach involves directly captured at regular intervals, allowing for continuous
extracting features from these signals to determine stress monitoring of stress levels over time. This proactive
levels. However, this simplistic method overlooks the approach enables timely interventions and adjustments to
complex interplay between physiological responses and stress management strategies, promoting long-term
stress, potentially resulting in inaccurate assessments. resilience and well-being.

Another notable disadvantage is the lack of Enhanced Security: The Stress Detection System
universal patterns in stress expression among ensures enhanced security by capturing images of
individuals. People may exhibit diverse behaviors and authenticated users and monitoring stress levels
expressions under stress, making it difficult to establish exclusively within the authorized user base. This
a one-size-fits-all approach to stress detection. This approach safeguards sensitive user data and maintains
variability poses a considerable challenge in developing privacy, mitigating concerns regarding unauthorized
algorithms capable of reliably identifying stress across access or data breaches.
different individuals.
Additionally, the algorithms employed in the existing
system, such as Bayesian Network and J48, may have V. SYSTEM ARCHITECTURE
limitations in capturing the nuances of stress responses.
These algorithms may not fully account for the intricacies
of individual stress expressions, leading to suboptimal
performance in stress detection tasks. Overall, while the
existing system provides valuable insights into stress
detection, addressing these drawbacks is essential for
improving its accuracy and effectiveness in real-world
applications.
IV. PROPOSED SYSTEM

The proposed system utilizes machine learning

algorithms, such as KNN classifiers, to classify stress
levels. Image processing is employed at the initial stage
for detection, where the employee's image is provided by
the browser, serving as input. To enhance the image and
extract useful information, image processing techniques
are applied, converting the image into a digital format and
performing operations on it. The input image can yield an
VI. MODULES
altered image or characteristics associated with it, with
emotions displayed in rounded boxes. Stress levels are
 User
indicated by emotions such as Angry, Disgusted, Fearful,
and Sad.  Admin
 Data Preprocess
Advantages of Proposed System:  Machine Learning

Altered Image or Report Output: The system provides 1. User:

output in the form of an altered image or report based on The User module facilitates user registration,
image analysis. This allows for a visual representation of requiring a valid email and mobile number for
stress levels and associated emotions, aiding in easy communication purposes.
interpretation and decision-making.
 Upon registration, the admin can activate the user, VII. CONCLUSION
granting access to the system. Once logged in, users can
input images into the system, which then utilizes Python The Stress Detection System is meticulously designed to
libraries to extract features and determine the forecast stress levels in employees through the
appropriate emotion depicted in the image. The system continuous monitoring of authenticated users' captured
can detect multiple faces in a single image and indicate images, ensuring system security. Automated image
stress levels based on facial expressions such as sadness capture occurs at predetermined time intervals when
or anger. Live streaming capabilities enhance real-time authenticated users are logged in, providing a seamless
facial expression detection, with TensorFlow offering and non-intrusive stress monitoring experience. Utilizing
faster and more accurate results. After image processing, standard conversion and image processing mechanisms,
the dataset is loaded for KNN classification, assessing the captured images are meticulously analyzed to discern
accuracy and precision scores. the user's stress levels. Subsequently, the system employs
advanced Machine Learning algorithms to process this
2. Admin: data, generating highly efficient and accurate results. By
The Admin module allows administrators to log in seamlessly integrating image processing and Machine
with their credentials and manage user activations. Learning techniques, the Stress Detection System offers a
Admins can activate users, granting them access to the comprehensive solution for proactive stress management
application. They have the authority to dynamically set in workplace environments.
training and testing data for the project within the code.
Admins can view all users' detected results and emotions
through a user-friendly interface. By clicking hyperlinks, VIII. REFERENCE
admins can access detailed emotion detection results and
view KNN classification outcomes. The dataset is [1] G. Giannakakis, D. Manousos, F. Chiarugi,
presented in Excel format, allowing authorized personnel “Stress and anxiety detection using facial cues
to expand it according to predetermined criteria. from videos,” Biomedical Signal processing and
Control”, vol. 31, pp. 89- 101, January 2017.
3. Data Preprocess:
The Data Preprocess module involves handling [2] Nisha Raichur, Nidhi Lonakadi, Priyanka
the dataset, which initially contains a grid view of stored Mural, “Detection of Stress Using Image
data with various properties. Through Property Extraction, Processing and Machine Learning Techniques”,
a new dataset is generated, containing only numerical vol.9, no. 3S, July 2017.
input variables derived from Principal Component
Analysis feature selection. This transformation results in [3] U. S. Reddy, A. V. Thota and A. Dharun,
six principal components, including Condition (e.g., No "Machine Learning Techniques for Stress
stress, Time pressure), Stress, Physical Demand, Prediction in Working Employees," 2018 IEEE
Performance, and Frustration. International Conference on Computational
Intelligence and Computing Research (ICCIC),
4. Machine Learning: Madurai, India, 2018, pp. 1-4.
The Machine Learning module employs the K-
Nearest Neighbor (KNN) algorithm for classification and [4] T. Jick and R. Payne, “Stress at work,”
regression analysis. KNN is a supervised learning Journal of Management Education, vol. 5, no. 3,
algorithm utilized for predicting whether a person requires pp. 50-56, 1980.
treatment based on similarity to instances from known
data. It operates by estimating the parameters of a KNN [5] Bhattacharyya, R., & Basu, S. (2018).
model, with a binary dependent variable having two Retrieved from ‘The Economic Times’.
possible values represented by an indicator variable (e.g.,
"0" and "1"). KNN classification involves estimating the [6] OSMI Mental Health in Tech Survey
parameters of the model and analyzing accuracy in Dataset, 2017
predicting stress levels based on input variables.
[7] https://www.kaggle.com/qiriro/stre