International Journal of Research in Engineering, Science and Management 121

Volume-3, Issue-1, January-2020

www.ijresm.com | ISSN (Online): 2581-5792

Stress Detection in IT Professionals by Image

Processing and Machine Learning
Ankita Patil1, Rucha Mangalekar2, Nikita Kupawdekar3, Viraj Chavan4, Sanket Patil5, Ajinkya Yadav6
1,2,3,4,5
Student, Department of Computer Science Engineering, D. Y. Patil College of Engineering & Technology,
Kolhapur, India
6
Assistant Professor, Department of Computer Science Engineering, D. Y. Patil College of Engineering &
Technology, Kolhapur, India

Abstract: The main motive of our project is to detect stress in System gets the ability to automatically learn and improve
the IT professionals using vivid Machine learning and Image from self-experiences without being explicitly programmed
processing techniques .Our system is an upgraded version of the using Machine learning which is an application of artificial
old stress detection systems which excluded the live detection and
intelligence (AI). Computer programs are developed by
the personal counseling but this system comprises of live detection
and periodic analysis of employees and detecting physical as well Machine Learning that can access data and use it to learn for
as mental stress levels in his/her by providing them with proper themselves. Explicit programming to perform the task based on
remedies for managing stress by providing survey form predictions or decisions builds a mathematical model based on
periodically. Our system mainly focuses on managing stress and "training data" by using Machine Learning. The extraction of
making the working environment healthy and spontaneous for the hidden data, association of image data and additional pattern
employees and to get the best out of them during working hours.
which are unclearly visible in image is done using Image
Keywords: Image processing, KNN classifier, One-hot encoding, Mining. It’s an interrelated field that involves, Image
Open CV, Supervised machine learning, Training dataset. Processing, Data Mining, Machine Learning and Datasets.
According to conservative estimates in medical books, 50-
1. Introduction 80% of all physical diseases are caused by stress. Stress is
Nowadays as IT industries are setting a new peek in the believed to be the principal cause in cardiovascular diseases.
market by bringing new technologies and products in the Stress can place one at higher risk for diabetes, ulcers, asthma,
market. In this study, the stress levels in employees are also migraine headaches, skin disorders, epilepsy, and sexual
noticed to raise the bar high. Though there are many dysfunction. Each of these diseases, and host of others, is
organizations who provide mental health related schemes for psychosomatic (i.e., either caused or exaggerated by mental
their employees but the issue is far from control. In this paper conditions such as stress) in nature. Stress has three prong
we try to go in the depth of this problem by trying to detect the effects:
stress patterns in the working employee in the companies we  Subjective effects of stress include feelings of guilt,
would like to apply image processing and machine learning shame, anxiety, aggression or frustration. Individuals
techniques to analyze stress patterns and to narrow down the also feel tired, tense, nervous, irritable, moody, or
factors that strongly determine the stress levels. lonely.
Machine Learning algorithms like KNN classifiers are  Visible changes in a person's behavior are represented
applied to classify stress. Image Processing is used at the initial by Behavioral effects of stress. Effects of behavioral
stage for detection, the employee’s image is clicked by the stress are seen such as increased accidents, use of
camera which serves as input. In order to get an enhanced image drugs or alcohol, laughter out of context, outlandish or
or to extract some useful information from it image processing argumentative behavior, very excitable moods, and/or
is used by converting image into digital form and performing eating or drinking to excess.
some operations on it. By taking input as an image from video  Diminishing mental ability, impaired judgment, rash
frames and output may be image or characteristics associated decisions, forgetfulness and/or hypersensitivity to
with that image. Image processing basically includes the criticism are some of the effects of Cognitive stress.
following three steps:
 Importing the image via image acquisition tools. 2. Need of work
 Analyzing and manipulating the image. Stress is called as an initial stage of depression. stress can be
 Output in which result is altered image or report that related to finance, work, relationships etc. In corporate world
is based on image analysis. employees are unaware of stress leading conditions while
 International Journal of Research in Engineering, Science and Management 122
Volume-3, Issue-1, January-2020
www.ijresm.com | ISSN (Online): 2581-5792

working. It is always observed mostly in It employees chronic Logistic Regression: Like all regression methods, the logistic
stress is often ignored. Companies use to give a survey form to regression is a predictive analysis. It is used in scenarios where
the employees to fill and then use to predict stress based on that one binary variable is dependent on one or more independent
form. It was not only time consuming but needed whole lot of variables. A logistic regression can be called as a statistical
efforts as forms where distributed manually. Stress Detection model that uses a binary dependent variable. In regression
System enables employees with coping up with their issues analysis, logistic regression [8] is estimating the parameters of
leading to stress by preventative stress management solutions a logistic model. Mathematically, a binary logistic model has a
which is concerned with eliminating stress and improving dependent variable with two possible value, which is
employee health. In our work we have designed a system which represented by an indicator variable, where the two values are
will capture images of the employee based on the regular labeled "0" and "1".
intervals and then the tradition survey forms will be given to the Mini-Batch Gradient Descent: A subset of training is
employee. This will reduce the manual efforts and time. This considered; it can make quick updates in the model parameters
organizational technique can be used to help improve employee and can also exploit the speed associated with vectorizing the
stress by diagnosis through our specially designed code. Depending upon the batch size, the updates can be made.
Questionnaires. Mini Batching is used to set the threshold value updated by
gradient which makes it a robust algorithm.
3. Objectives KNN Classifier: K-Nearest Neighbor (KNN) is used for
 To predict stress in a person by the symptoms classification as well as regression analysis. It is a supervised
calculated by monitoring. learning algorithm which is used for predicting if a person
 To analyze the stress levels in the employee. needs treatment or not. KNN classifies the dependent variable
based on how similar it is; independent variables are to a similar
 To provide solutions and remedies for the person to
instance from the already known data.
recover his/her stress.
Dataset: Dataset contains grid view of already stored dataset
consisting numerous properties[8], by Property Extraction
4. Methodologies
newly designed dataset appears which contains only numerical
Image Pre-processing: G (i, j) = α. F (i, j) + β, α > 0 and β input variables as a result of Principal Component Analysis
are called as gain and bias parameters, these are used to bright feature selection transforming to 6 principal components which
and contrast the image. Here G (i, j) is output image pixel and are Condition (No stress, Time pressure, Interruption), Stress,
F (i, j) is input image pixel. Physical Demand, Performance and Frustration.
Pixel transformation: Pixel transformation is a technique
used in image processing to obtain pixel values. This
transformation is used to make image generic and diverse. The
image is converted into Gray scale image that is a color image
is converted into black and white or in shades of grey.
Threshold of the image is found which is used to convert gray
scale image into binary form, If the pixel value is greater than
threshold pixel value is set to 1 otherwise 0.
One hot encoding: All the textual responses were given
numerical weights according to their significance. 'Yes' is taken
Fig. 2. Dataset without property extraction
to be 1, 'no' to be 0. The categorical data was converted into
numeric using label encoder. A decoder is used to decode into Raw dataset contains numerous unusable properties like
binary code. Though a one-hot state machine, does not need a Temporal Demand, Heart Rate, Effort, Mental Effort, Nasa
decoder as machine is in the nth state if and only if the nth bit TLX, Mental Demand etc.
is high.

Fig. 1. Values after one-hot encoding operation Fig. 3. Dataset after property extraction
 International Journal of Research in Engineering, Science and Management 123
Volume-3, Issue-1, January-2020
www.ijresm.com | ISSN (Online): 2581-5792

Essential properties namely Condition (No stress, Time provided for overcoming stress.
pressure, Interruption), Stress, Physical Demand, Performance
and Frustration from raw dataset are extracted to build a new 5. Conclusion
property extracted dataset. Stress Detection System is designed to predict stress in the
employees by monitoring captured images of authenticated
users which makes the system secure. The image capturing is
done automatically when the authenticate user is logged in
based on some time interval. The captured images are used to
detect the stress of the user based on some standard conversion
and image processing mechanisms. Then the system will
analyze the stress levels by using Machine Learning algorithms
which generates the results that are more efficient.

References
[1] G. Giannakakis, D. Manousos, F. Chiarugi, “Stress and anxiety detection
using facial cues from videos,” Biomedical Signal processing and
Control”, vol. 31, pp. 89-101, January 2017.
[2] T. Jick and R. Payne, “Stress at work,” Journal of Management Education,
vol. 5, no. 3, pp. 50-56, 1980.
Fig. 4. Architecture diagram of stress detection system [3] Nisha Raichur, Nidhi Lonakadi, Priyanka Mural, “Detection of Stress
Using Image Processing and Machine Learning Techniques”, vol.9, no.
The flow and working of the stress detection system is 3S, July 2017.
explained in Fig. 4, the system is divided into three modules the [4] Bhattacharyya, R., & Basu, S. (2018). Retrieved from ‘The Economic
Times’.
first module registers the employee, sends alerts and later [5] OSMI Mental Health in Tech Survey Dataset, 2017
provides survey forms. Second module works on image [6] U. S. Reddy, A. V. Thota and A. Dharun, "Machine Learning Techniques
capturing, conversion of image into coordinates and then for Stress Prediction in Working Employees," 2018 IEEE International
Conference on Computational Intelligence and Computing Research
mapping the image and getting a prediction about stress and the (ICCIC), Madurai, India, 2018, pp. 1-4.
third module converts data into binary values, based on which [7] https://www.kaggle.com/qiriro/stress
stress level is detected and based on the level solution is