Psychoanalysts and clinical psychologists have been using vocal patterns of subjects/patients since decades to analyze their personality traits and help make headway in medical and clinical investigations.
And in several cases, through these speech patterns, they also successfully identified sociopathy and other psychological or psychosomatic conditions.
And recent breakthroughs in healthcare have combined this methodology with the concept of biological markers to produce a unique diagnostic tool called vocal biomarker, which can diagnose a disease, disorder or condition based entirely on vocal cues.
As part of an ongoing breakthrough, a team lead by New York University Langone Medical Center psychiatry department chairman Charles Marmar is developing an Artificial Intelligence (AI) system that uses machine learning to analyze various voice characteristics, including pitch, tone and rhythm from patients.
Marmar’s team so far has gathered a set of 30 voice characteristics apparently associated with Post Traumatic Stress Disorder (PTSD) and Traumatic Brain Injury (TBI) from a total of 40,000 features mined from the vocal patterns.
These features are fed processed by the AI and an algorithm picks out vocal patterns. As an example, people with cognitive problems may elongate certain frequencies of sound, or struggle with pronouncing phrases that require complex facial muscle movements. The AI tracks these patterns and diagnoses the underlying condition.
The role of vocal biomarkers in remote health monitoring is also being researched into. As an example, Mayo Clinic and Beyond Verbal are collaborating to discover vocal patterns of patients with coronary artery disease, a common form of heart disease. The reasoning behind this effort is that the hardening of arteries affects speech production.
Novelty
The role of acoustic biomarkers isn’t restricted to audible frequencies. Their prominence lies in detecting frequencies inaudible to human beings.
Certain inaudible frequencies in the above mentioned case were related to a 19-fold increase in the risk of coronary artery disease in the subjects.
The AI used for vocal biomarkers could be installed in a smartphone and used as a low-cost, quick screening tool or a remote health monitoring device that will indicate whether or not the patient is taking the prescribed medication.
Extent of Applications
Vocal biomarkers, with advancement in technology, could soon be used by psychologists or psychiatrists to instantly diagnose PTSD, postpartum depression in new mothers, certain cancers, heart conditions, and even dementia, Parkison’s disease.
Challenges
The challenge is now to make the new type of biomarker ubiquitous and universal by developing technology that can accommodate the advanced AI in smart devices.
Another challenge is the uncertainty surrounding the rumours of the results of these tests, in their current state of technology, being susceptible to fraudulent sounds.
Apart from these, scientists are also trying to work on developing firewalls to protect privacy and security breaches. However, most researchers claim that the software only identifies speech patterns, and cannot understand the meaning of the words the patients say.
Hi, my interesting area is Machine Learning. I am very inspired after looking this article. I would like to work on vocal biomarkers for Parkinson disease. As you said that a date set is collected that consists of 40, 000 features. Could you please provide that data for my work!
Kindly provide that data….
Thank You..