Researchers from New York University have used artificial intelligence (AI) to objectively detect Post traumatic stress disorder PTSD simply by listening to the sound of someone’s voice.
NYU and SRI spent five years creating a voice analysis program that not only understands human speech, but also detects PTSD signifiers and emotions.
The AI listened for minor variables and auditory markers that would be hardly noticeable to the human ear. The team claimed that the algorithm is able to diagnose PTSD with 89% accuracy.
For tests, the team interviewed and recorded 129 war-zone exposed veterans and collected 40,000 speech samples to study. They then used the audio to teach the AI with vocal changes associated with diagnoses of PTSD, for instance, a slower and dull tone was an indicator of PTSD, as well as a shorter tonal range with less enthusiasm.
AI can detect slight changes in the voice such as the tension of throat muscles and whether the tongue touches the lips all probable indicators of a PTSD diagnosis.
“Speech is an attractive candidate for use in an automated diagnostic system, perhaps as part of a future PTSD smartphone app, because it can be measured cheaply, remotely, and nonintrusively,” said Adam Brown, lead author of the study.