CHAPTER 1

INTRODUCTION
Mental stress is a major issue nowadays, especially among youngsters. The age that
was considered once most carefree is now under a large amount of stress. Stress increase
nowadays leads to many problems like depression, suicide, heart attack, and stroke. Stress
has become an embedded part of our daily life and is a noticeable concept in public health.
Recently, stress has become an integral part of professional life, especially in today’s fiercely
competitive economy. In the workplace, an individual has to continuously face several
situations, such as work overload, job insecurity, lack of job satisfaction, and the pressure to
stay up-to-date. The continuous presence of stress can lead to several negative health effects,
such as high blood pressure, lack of sleep, susceptibility to infections, and cardiovascular
disease. All these situations result in mental stress, which has become the leading cause of
many diseases. Such adverse effects not only affect the employees’ health and well-being, but
also affect workplace productivity and overall profit.

Fig. 1 shows the basic conceptual model of stress. The significance of sleep, physical
activity, number of working hours and change in heart rate with regard to stress levels are
basic factors which affects the stress level.

Figure1.1: The conceptual model of STRESS

Stress is our body’s reaction to pressure. It is a compressed feeling. Various

circumstances or life occasions can cause pressure. Everyone manages pressure in an
unexpected way. Their capacity to adapt to it relies upon their hereditary, early life occasions,
character, and social what's more, monetary conditions. Nonetheless, an excessive amount of
pressure can cause adverse consequences prompting sadness. As World Health Organization
(WHO) says, Stress might be a mental state disadvantage moving the lifetime of one in four
citizens. Human pressure results in mental besides as socio-monetary issues, absence of
straightforwardness in work, poor working connections, gloom, and in the long run obligation
to self-destruction in serious cases. It might in fact leave them in an extremely durable phase
of consistent concern and fear. In long haul it can influence physical and psychological
wellnesssecerely. Therefore, individuals are unsettled regardless of flourishing. The pressure
can be brought about by profound, physical, or even mental.

One can isolate pressure into two kinds, Short-term, and the other Long-term or
ongoing pressure. Momentary pressure is situational when the circumstance changes, the
pressure is gone or over. Long haul pressure is a drawn-out issue. That is the reason it is
hazardous. Now and again persistent pressure is generally genetic or qualities related
moreover. It can make to end anyone’s life. Unreasonable pressure drives one ridiculously.
Tolerating pressure is the initial step to lead a calm blissful life. It is vital to recognize and
oversee pressure before it turns intense simple cures are accessible at a reasonable expense.
CHAPTER 2

LITERATURE SURVEY

2.1 LITERATURE REVIEW

In 2017, N. Widanti, B. Sumanto, P. Rosa and M.F. Miftahudin proposed work in

which stress detector classifies a stressed individual from a normal one by acquiring his/her
physiological signals through appropriate sensors such as Electrocardiogram (ECG),
Galvanic Skin Response (GSR) etc,. These signals are pre-processed to extract the desired
features which depicts the stress level in working individuals. Support Vector Machine
(SVM) and K-Nearest Neighbour (KNN) are investigated to classify these extracted feature
set. The result indicates feature vector with best features having a strong influence in stress
identification. An attempt is made to determine the best feature set that results in maximum
classification accuracy. Proposed techniques are applied on benchmark SWELL-KW dataset
and state-of-art results are obtained.

In 2017, Nisha Raichur, Nidhi Lonakadi and Priyanka Mural proposed work in
which a real-time non-intrusive videos are captured, which detects the emotional status of a
person by analysing the facial expression. They detect an individual emotion in each video
frame and the decision on the stress level is made in sequential hours of the video captured.
They employ a technique that allows them to train a model and analyze differences in
predicting the features. Theano is a python framework which aims at improving both the
execution time and development time of the linear regression model which is used here as a
deep learning algorithm. The experimental results show that the developed system is well on
data with the generic model of all ages.

In 2017, Huijie Lin, Jia Jia and Jiezhon Qiu find that users stress state is closely
related to that of his/her friends in social media, and they employ a large-scale dataset from
real-world social platforms to systematically study the correlation of users' stress states and
social interactions. They first define a set of stress-related textual, visual, and social attributes
from various aspects, and then propose a novel hybrid model - a factor graph model
combined with Convolutional Neural Network to leverage tweet content and social
interaction information for stress detection. Experimental results show that the proposed
model can improve the detection performance by 6-9 percent in F1-score. By further
analyzing the social interaction data, we also discover several intriguing phenomena, i.e., the
number of social structures of sparse connections (i.e., with no delta connections) of stressed
users is around 14 percent higher than that of non-stressed users, indicating that the social
structure of stressed users' friends tend to be less connected and less complicated than that of
non-stressed users.

In 2018, U. S. Reddy, A. V. Thota and A. Dharun would like to apply machine

learning techniques to analyze stress patterns in working adults and to narrow down the
factors that strongly determine the stress levels. Towards this, data from the OSMI mental
health survey 2017 responses of working professionals within the tech-industry was
considered. Various Machine Learning techniques were applied to train our model after due
data cleaning and preprocessing. The accuracy of the above models was obtained and studied
comparatively. Boosting had the highest accuracy among the models implemented. By using
Decision Trees, prominent features that influence stress were identified as gender, family
history and availability of health benefits in the workplace. With these results, industries can
now narrow down their approach to reduce stress and create a much comfortable workplace
for their employees.

In 2018, Saskia Koldijk, Mark A. Neerincx and Wessel Kraaij developed an

automatic classifiers to infer working conditions and stress related mental states from a
multimodal set of sensor data (computer logging, facial expressions, posture and physiology).
They address two methodological and applied machine learning challenges: 1) Detecting
work stress using several (physically) unobtrusive sensors, and 2) Taking into account
individual differences. A comparison of several classification approaches showed that, for
our SWELL-KW dataset, neutral and stressful working conditions can be distinguished with
90 percent accuracy by means of SVM. Posture yields most valuable information, followed
by facial expressions. Furthermore, we found that the subjective variable`mental effort' can
be better predicted from sensor data than, e.g., `perceived stress'. A comparison of several
regression approaches showed that mental effort can be predicted best by a decision tree
(correlation of 0.82). Facial expressions yield most valuable information, followed by
posture. They find that especially for estimating mental states it makes sense to address
individual differences. When they train models on particular subgroups of similar users, (in
almost all cases) a specialized model performs equally well or better than a generic model.
 In 2018, B. Padmaja, V. V. Rama Prasad and K. V. N. Sunitha provides an
effective method for the detection of cognitive stress levels using data provided from a
physical activity tracker device developed by FITBIT. The main motive of this system was to
use a machine learning approach in stress detection using sensor technology. Individually, the
effect of each stressor was evaluated using logistic regression and then a combined model
was built and assessed using variants of ordinal logistic regression models including logit,
probit, and complementary log-log. This system was used and evaluated in areal-time
environment by taking data from adults working in IT and other sectors in India. The novelty
of this work lies in the fact that a stress detection system should be as non-invasive as
possible for the user.

In 2019, Ravinder Ahujaa and Alisha Bangab proposed work in which they are
calculating the mental stress of students one week before the exam and during the usage of
the internet. Their objective is to analyze stress in the college students at different points in
his life. The effect that exam pressure or recruitments stress has on the student which often
goes unnoticed. They will perform an analysis on how these factors affect the mind of a
student and will also correlate this stress with the time spent on the internet. The dataset was
taken from Jaypee Institute of Information Technology and it consisted of 206 student’s data.
Four classification algorithms Linear Regression, Naïve Bayes, Random Forest, and SVM is
applied and sensitivity, specificity, and accuracy are used as a performance parameter. The
accuracy and performance of data are further enhanced by applying 10-Fold CrossValidation.
The highest accuracy recorded was by Support Vector Machine (85.71%).

2.2 EXISTING SYSTEM

Stress management systems play a major role to notice the stress levels that disrupts
our socio-economic mode. This demand counseling to be provided for the stressed people
cope up against stress. Stress turning away is not possible however preventive actions help to
beat the stress. Currently, solely medical and physiological consultants will verify whether or
not one is beneath depressed state (stressed) or not. One in every of the normal methodology
to notice stress is predicated on form. Nowadays as IT industries are setting a new peek in the
market by bringing new technologies and products in the market. Though there are many
organizations that provide mental health related schemes for their employees but the issue is
far from control form.
2.3 PROBLEM STATEMENT

Stress is called as an initial stage of depression. stress can be related to finance, work,
relationships etc. In corporate world employees are unaware of stress leading conditions
while working. It is always observed mostly in IT employees chronic stress is often ignored.
Companies use to give a survey form to the employees to fill and then use to predict stress
based on that form. It was not only time consuming but needed whole lot of efforts as forms
where distributed manually. There is lack of auto detection of employee’s stress or emotions
in real-time. Thus we need to develop an efficient real-time detection of tensity or stress or
emotions to manage the employee’s health.

2.4 PROPOSED SYSTEM

We propose a machine learning and visual processing model to identify overworked

IT employees. Our technology is an improved version of older stress detection systems that
did not include live detection or personal counseling. Stress detection methods that don't
include real-time monitoring or individual counseling are being updated in this project. A
survey is used to collect data on employees' mental stress levels in order to provide effective
stress management solutions. In order to get the most out of your employees, this project will
look at stress management and how to create a healthy, spontaneous work environment.

2.5 OBJECTIVES

The main objective of our work is to propose a reliable, convenient and accurate stress and
emotions detection system of employees working infront of computers. Therefore our project
has the following specific objectives:

1. To collect the data of different stress showing emotion images.

2. To pre-process the collected dataset.
3. To develop an efficient Machine learning model for detection of stress and emotions
accurately.
4. To evaluate performance analysis of the proposed system for validation.
CHAPTER 3
REQUIREMENTS SPECIFICATION
3.1 FUNCTIONAL REQUIREMENTS
Functional requirements are the software capabilities which are to be present in order
for the user to get the service from the system.

Our proposed system is having the applications in the field of IT and other employees
who working infront of computers for detecting stress. The following are the functional
requirements of proposed system:

1. Image Acquisition: The proposed system is fed with the images of different stress
faces.
2. Database Operations: Various types of employer’s images are maintained in the
database.

3.2 NON-FUNCTIONAL REQUIREMENTS

Non-Functional Requirements provides criteria that are used to decide the operations of the
system. The following are the non-functional requirement:

1. Scalability: The proposed system can extended to any number of IT fields.

2. Reliability: Our proposed system is reliable. There will be no false data and lesser
loss of information as it ensures increased data utility.
3. Flexibility: The proposed system is more flexible to the user as it uses simple
techniques and also less complexity so that everyone can understand easily.
4. Performance: Our proposed system accurately detects than the legacy system.
5. Cost: Lesser cost compare to existing technique.

3.3 HARDWARE AND SOFTWARE REQUIREMENTS

Minimal Hardware Requirements

 System : Pentium i3 Processor.

 Hard Disk : 500 GB.
 Monitor : 15’’ LED
 Input Devices : Keyboard, Mouse
  Ram : 4 GB

Minimal Software Requirements

 Operating system : Windows 10.

 Coding Language : Python 3.8
 Web Framework : Flask
CHAPTER 4

METHODOLOGY
As the title suggests, here the stress is detected using the facial recognition. The
dataset used for training is fer2013 dataset. It contains the greyscale images of seven different
emotions. The dataset has 35,887 images. The emotions are Happy, Angry, Sad, Disgust,
Surprise, Fear and Neutral. The dataset is trained using Convolutional Neural Networks. Here
five convolutional layers are used to train the dataset. These layers were implemented using
Sequential Model. The entire training process had completed with 100 epochs with a batch
size of 64. Adam optimizer is used to compile the model. The trained model is then saved in
json file. A camera is used to capture the near front sight of the person while they are
working in front of the computer.

Captured video is divided into sections of equivalent length and set of similar number
of image frames are extracted from each part correspondingly and are examined using some
image processing techniques. The stress detection module scans the binary image from the
extreme left top to record the co-ordinates of the eyebrow and lips. The image detection
includes the calculation of the variation in the place of the eyebrow and lip movements from
its mean position. The displacement of eyebrow from its place is considered by examining
the image for the eyebrow co-ordinates. The displacement of lips from its place is considered
by examining the image for the lip co-ordinates. Emotion is detected for that particular frame.
Emotions like Scared, Sad and Angry are classified as Stressed. This is deployed in a web
application using Flask framework. Block diagram of proposed system is shown in the below
figure 4.1.
Fig 4.1: Block Diagram of proposed System
CHAPTER 5

CONCLUSION
We design a system to predict stress in the person by monitoring captured video in
real-time. The image is captured from the frames. The captured images are used to detect the
stress of the user based on some standard conversion and image processing mechanisms
using Machine Learning algorithms which generates the results that are more efficient. Also
we do performance analysis of our developed system.
 REFERENCES
1. N. Widanti, B. Sumanto, P. Rosa and M.F. Miftahudin, "Stress level detection using
heart rate blood pressure and gsr and stress therapy by utilizing infrared", Industrial
Instrumentation and Control (ICIC) 2017 International Conference on, 2017.
2. Nisha Raichur, Nidhi Lonakadi and Priyanka Mural, Detection of Stress Using Image
Processing and Machine Learning Techniques, vol. 9, no. 3S, July 2017.
3. Huijie Lin, Jia Jia and Jiezhon Qiu, "Detecting stress based on social interactions in
social networks", IEEE Transactions on Knowledge and Data Engineering, 2017
4. U. S. Reddy, A. V. Thota and A. Dharun, "Machine Learning Techniques for Stress
Prediction in Working Employees", 2018 IEEE International Conference on
Computational Intelligence and Computing Research (ICCIC), 2018.
5. Saskia Koldijk, Mark A. Neerincx and Wessel Kraaij, "Detecting Work Stress in
offices by combining unobtrusive sensors", IEEE Transactions on affective
computing, 2018.
6. B. Padmaja, V. V. Rama Prasad and K. V. N. Sunitha, "A Machine Learning
Approach for Stress Detection using a wireless Physical Activity Tracker",
International Journal of Machine Learning and Computing, vol. 8, no. 1, February
2018
7. Ravinder Ahujaa and Alisha Bangab, "Mental Stress Detection in University Students
using Machine Learning Algorithms", International Conference on Pervasive
Computing Advances and Applications, 2019.