Loading...

Design and Implementation of Wearable Device for Stress Level Measurement

Mohammadi, Amir Mohammad | 2021

334 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 53695 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Fakharzadeh, Mohammad
  7. Abstract:
  8. An inseparable problem from human daily life is stress that causes problems such as heart disease and depression, so stress management and control is essential for the health of the individual and society. This thesis explores the possibility of stress detection using vital signs and machine learning algorithms. First, by examining the potential of unsupervised learning algorithms for stress detection, a general method is developed and the accuracy of the algorithm is evaluated with the ECG signals of a smart wristband made in Sharif University of Technology Biosen group as well as WESAD data set. The self-organizing map structure is created based on stress-related features and final result is divided into three levels no stress, low stress and high stress with the Fuzzy c-means algorithm. Then, in the next step, 65 features were extracted by supervised learning method from WESAD data set, and 43 of them were selected for machine learning by Kruskal-Wallis test. Finally, accuracy of 93.92 ± 2.47%, sensitivity of 94.94 ± 3.06% and specificity of 92.12 ± 5.32% were obtained by combining ECG and EDA signals. In the next step, the board needed to receive the skin conduction signal (EDA) was designed and built. Thirteen participants with a mean age of 26.7 ± 8.3 years were exposed to stress by SCWT and arithmetic task and EDA signals were collected. After processing the signals, the accuracy was 81.33 ± 4.42%, the sensitivity was 80.12 ± 7.39% and the specificity was 82.41 ± 6.61%. Based on the results obtained from the WESAD dataset and practical measurements, the KNN algorithm was selected as the best method for stress diagnosis. The final model of the supervised algorithm can be used to detect stress through wearable devices that can collect data and transfer it to the server
  9. Keywords:
  10. Stress Distribution ; Machine Learning ; Electrocardiogram ; Electrodermal Activity ; Heart Activity

 Digital Object List

 Bookmark

No TOC