Artificial intelligence assesses the severity of depression on the voice and human behavior

More than 300 million people in the world suffer from depression. A ever-growing number of patients with deep depression serves as an incentive to develop more advanced automatic diagnosis technologies. To this end, scientists from the Indian Institute of Technology Patna and the University of Canopy in Normandy have programmed artificial intelligence for analyzing human behavior and voice analysis.

The researchers created six directions for analyzing machine learning: head slope, duration and intensity of smile, touching to themselves, look and speech. Then the data was then translated into a mathematical representation and a comparison of indicators on the scale of the primary assessment of PHQ-8 mental disorders was carried out.

The AI was trained on the basis of 189 clinical polls of people suffering from anxiety, depression and post-traumatic stress. The data of human behavior, its voice, including the decoding of transcripts during the survey, remained digital form. As a result, II learned to determine the depths of depression by 8% better than the currently existing BEYOND Verbal, Affectiva and Mit Spinout Cogito system.

Previously, developers from Facebook created