Communications of the ACM
How Machine Learning Can Help With Voice Disorders
Researchers from the Massachusetts Institute of Technology's Computer Science and Artificial Intelligence Laboratory and Massachusetts General Hospital have developed a system that diagnoses patients with muscle tension dysphonia.
Credit: Daryush Mehta/Massachusetts General Hospital
A new diagnostic approach using machine learning could help detect speech disorders exacerbated by vocal misuse.
Researchers from the Massachusetts Institute of Technology's (MIT) Computer Science and Artificial Intelligence Laboratory (CSAIL) and Massachusetts General Hospital developed a system that diagnoses patients with muscle tension dysphonia.
Current approaches to diagnosing physiological signals via machine learning often involve supervised learning, in which researchers must label data and provide outputs, but the CSAIL team opted to use unsupervised learning.
"People with vocal disorders aren't always misusing their voices, and people without disorders also occasionally misuse their voices," says MIT student Marzyeh Ghassemi. "The difficult task here was to build a learning algorithm that can determine what sort of vocal cord movements are prominent in subjects with a disorder."
To train the algorithm, patients diagnosed with voice disorders and a control group wore accelerometers that captured the motions of their vocal folds. By comparing more than 110 million glottal pulses, the team differentiated between patients and controls, and were able to measure the positive effects of voice therapy on patients with vocal disorders.
The team says the data could be used by doctors and scientists to study the underlying causes of vocal disorders and help patients employ healthier vocal behaviors.
From MIT News
View Full Article
Abstracts Copyright © 2016 Information Inc., Bethesda, Maryland, USA
No entries found