More than 264 million people globally suffer from Major Depressive Disorder (MDD), with another 20 million experiencing schizophrenia. Combined, these two afflictions are among the most common preludes to suicide. Finding ways of recognizing symptoms of people with either is crucial to preventing a bad end, and a new system purports to do exactly that.
A speech inversion system analyzes acoustic signals (sounds) to specific vocal track variables, depicting the spatial movement and timing of speech gestures, according to a new study presented virtually during the 180th Meeting of the Acoustical Society of America on Tuesday, June 8.
And, the researchers think an analysis of speech coordination will reveal depression, which typically alters the way we talk. However, while the drive to identify and help people with major depression is admirable, mental illness (and even well-balanced minds) can present in surprisingly random ways, so it’s best to approach this technology with a grain of salt. No one wants their iPhone to limit who they can see or where they may go because of the way their voice sounds.
Scientists say depression is linked to slower speech gestures
Suicide is the second-leading cause of death in children and young adults aged 10 to 34 years. It’s also the second-leading cause of death for Black children aged 10 to 14 years, and the third-leading cause of death for adolescents who are Black and aged 15 to 19 years. The 180th Meeting of the Acoustical Society of America (which is virtual this year) saw Carol Espy-Wilson of the University of Maryland discuss how a person’s mental health may be empirically observed in speech gestures. Her keynote lecture happened on Tuesday, June 8, at 4:05 PM EDT.
Espy-Wilson claims speech coordination changes when a person falls into major depression, and suggests that a speech inversion system capable of mapping acoustic signals to vocal track variables and noting the timing between speech gestures and spatial movement could identify these changes. “Depression is accompanied by psychomotor slowing. As a result, they cannot think as fast, and their speaking rate is slowed with more longer pauses than if they are not depressed,” said Espy-Wilson, in an embargoed release shared with IE. “This results in a simpler coordination of the articulators. What this means is that there is less coproduction of neighboring sounds and more of the sounds are fully articulated, that is, the articulators reach their targets.”
A depression-detecting system could be loaded onto an app
Employing machine-learning techniques to provide data for a deep-learning model of mental health classification, the research aims to develop this technological depression detector into something like a smartphone app, which could help patients and psychiatric professionals gauge the mental health of patients. Espy-Wilson claims this could have a substantial impact on people with MDD, and, by logical consequence, the suicide rate. “Ideally, therapists will give the app to patients who suffer from MDD when they are in remission or only have mild depression,” she said, in the embargoed release. “That is, they are in a state where they are likely to use it regularly, so their mental health status can be tracked, and the appropriate people will be alerted if the app detects that the severity of the depression is increasing.”
“In that way, we hope there will be intervention before their depression increases to a level where they may consider suicide,” added Espy-Wilson. But one can imagine the gray inside the silver lining: imagine an insurance company requiring you to download an app to assess your voice for signs of pre-existing mental health issues, and suddenly facing a higher deductible, all from a false positive. What if instead of an insurance company or medical professional, the one tuning in to your voice for signs of MDD works for someone else. It’s a well-intentioned and early-stage idea, but, with mounting concerns about data privacy, there are also ethical stakes at hand.