Related Experiment Video
Updated: Jul 11, 2025
12:51
Evaluation of Commercial-Off-The-Shelf Wrist Wearables to Estimate Stress on Students
Published on: June 16, 2018
7.5K
Using Wearable Devices and Speech Data for Personalized Machine Learning in Early Detection of Mental Disorders:
Ramon E Diaz-Ramos1, Isabella Noriega2, Luis A Trejo3
1Department of Computing Science, University of Alberta, Edmonton, AB, Canada.
JMIR Research Protocols
|November 13, 2023
Summary
This study develops a tool using wearable, speech, and self-report data for early detection of depression, anxiety, and stress. Machine learning models analyze longitudinal data for personalized mental health monitoring.
Area of Science:
- Digital health and machine learning applications in mental healthcare.
- Multimodal data analysis for psychological well-being assessment.
Background:
- Early identification of mental health symptoms is critical for effective treatment and reducing long-term disability.
- A tool leveraging longitudinal activity, speech, and self-reported mood data can aid individuals in recognizing early warning signs of mental health conditions.
Purpose of the Study:
- To evaluate machine learning models for mental health indicator detection using multimodal data (wearables, speech, self-reports).
- To investigate the impact of longitudinal data, speech characteristics, and multilingual data on model accuracy.
- To compare personalized versus population-level machine learning models for mental health assessment.
Main Methods:
- Development of a mobile application collecting voice recordings, physiological and activity data from wearables, and self-reported mental state (Depression, Anxiety, and Stress Scale).
- Longitudinal data collection from participants aged 18-35, fluent in English or Spanish, to train machine learning models.
- Analysis of speech patterns during personal context discussions versus neutral text reading to identify mental health indicators.
Main Results:
- The study is ongoing, with data collection scheduled for completion in November 2023.
- Recruitment targets at least 50 participants from universities in Canada and Mexico.
- Inclusion criteria ensure a specific demographic and exclude individuals with communication or neurological disorders.
Conclusions:
- The research aims to advance personalized machine learning for mental health by creating a predictive dataset for depression, anxiety, and stress.
- A framework for early detection of mental health symptoms will be deployed.
- The long-term objective is to establish a noninvasive, objective method for continuous mental health data collection and symptom detection.
