Kinect-Powered App Diagnoses Depression with 90 Percent Accuracy

Researchers from the University of Southern California’s Institute for Creative Technologies have unveiled a new Kinect-powered tool that is capable of independently administering, recording, and scoring a psychological assessment that results in a depression diagnosis with a 90 percent accuracy.

The tool, called SimSensei, is currently diagnosing based only on verbal responses collected by an animated avatar that is programmed to administer a mental health questionnaire. However, on top of those verbal responses, researchers are now turning their attention to understanding and evaluating the non-verbal signs for depression.

The system uses facial recognition technology and depth-sensing cameras within Microsoft’s Kinect to interpret the the user’s body language and respond accordingly. Researchers say that depression is diagnosed entirely on a yes/no checklist and neglects to account for the potential diagnostic information hidden in body language. Researchers are training the tool to track exhibited non-verbal signs of depression – such as slower-than-usual blinking, distinctive facial expressions, certain upper body movements, and fewer and more restrained smiles — to help increase the accuracy of the diagnostic tool.

An automated evaluation for depression has a wealth of potential benefits in healthcare considering that it would be free. This information could be used to monitor patients from traditionally high risk populations, or those from geographically remote regions, such as rural Alaska, where mental health is a prominent issue and face-to-face evaluations are often unavailable.

The tool also holds potential for implications for faster and more routine screenings for PTSD for returning soldiers and veterans.

↑ Back to top

Founding Sponsors

Platinum Sponsors