Open AccessPredictive Analysis for Real Time Stress Detection
Author(s) -
Srivathsa Bharadwaj K S
Publication year - 2021
Publication title -
international journal for research in applied science and engineering technology
Language(s) - English
Resource type - Journals
ISSN - 2321-9653
DOI - 10.22214/ijraset.2021.35177
Subject(s) - computer science , pulse rate , classifier (uml) , stress (linguistics) , pulse (music) , feeling , heart rate variability , speech recognition , interval (graph theory) , artificial intelligence , pattern recognition (psychology) , heart rate , psychology , mathematics , biology , telecommunications , social psychology , linguistics , philosophy , combinatorics , detector , blood pressure , endocrinology
Stress can be characterized as a feeling of either emotional or physical tension. Due to the biology of the human body, it releases some hormones when under stress. These hormones might cause tensed muscles, increase pulse rate or the heart rate, increase brain activity to make the brain more alert to the surrounding. Stress can be predicted well before it happens by constantly measuring the Heart Rate Variability (HRV) parameters obtained using the pulse sensor. In this project a supervised machine learning model is created using the data acquired from Physionet, once the data is acquired it is cleaned and the missing data is filled. This data set is later used to create a random forest classifier and is saved using pickle library. Once the model is created it is used to detect stress in real time. Pulse sensor amped is used to get the required pulse data in the form of a CSV file and a numpy array is created using inter beat interval information got from pulse sensor. Once a numpy array is created neurokit2 library is used to extract the HRV information of the R-R interval. Later these parameters are compared with the created model and checked to see if the subject is stressed, if the model detects the subject as stressed an alerting message is sent to the subject’s smartphone using Twilio.