To main content

One-Dimensional Convolutional Neural Networks on Motor Activity Measurements in Detection of Depression

One-Dimensional Convolutional Neural Networks on Motor Activity Measurements in Detection of Depression

Category
Part of a book/report
Abstract
Nowadays, it has become possible to measure different human activities using wearable devices. Besides measuring the number of daily steps or calories burned, these datasets have much more potential since different activity levels are also collected. Such data would be helpful in the field of psychology because it can relate to various mental health issues such as changes in mood and stress. In this paper, we present a machine learning approach to detect depression using a dataset with motor activity recordings of one group of people with depression and one group without, i.e., the condition group includes 23 unipolar and bipolar persons, and the control group includes 32 persons without depression. We use convolutional neural networks to classify the depressed and nondepressed patients. Moreover, different levels of depression were classified. Finally, we trained a model that predicts MontgomeryÅsberg Depression Rating Scale scores. We achieved an average F1-score of 0.70 for detecting the control and condition groups. The mean squared error for score prediction was approximately 4.0.
Client
  • Norges forskningsråd / 259293
Language
English
Author(s)
  • Frogner Joakim Ihle
  • Noori Farzan Majeed
  • Halvorsen Pål
  • Hicks Steven Alexander
  • Garcia-Ceja Enrique
  • Tørresen Jim
  • Riegler Michael
Affiliation
  • Simula Metropolitan Center for Digital Engineering
  • University of Oslo
  • University of Oslo
  • Oslomet - Oslo Metropolitan University
  • University of Oslo
  • SINTEF Digital / Software and Service Innovation
  • Kristiania University College
Year
2019
Publisher
ACM Digital Library
Book
Proceedings of the 4th International Workshop on Multimedia for Personal Health & Health Care
ISBN
978-1-4503-6914-5
Page(s)
9 - 15