A team of scientists from Dartmouth College of New Hampshire has developed a prototype of the MoodCapture app, which can help people diagnosed with depressive disorder. The software relies on facial recognition to detect a person’s emotions.
Whenever the user unlocks the smartphone, MoodCapture will record the facial expression of the owner of the gadget using a facial recognition sensor and the front camera. Then an algorithm based on artificial intelligence will analyze individual facial features and track the dynamics of the development of a mental disorder.
If MoodCapture determines that the user’s depression is getting worse, the software will offer the smartphone owner several solutions. For example, playing sports outdoors or socializing with family and friends. To do this, the AI was trained on 177 subjects diagnosed with major depressive disorder.
For three months, each person’s smartphone photographed their faces. In parallel, the subjects were asked to assess their psychological state. There were several items to choose from, such as “I feel depressed and hopeless.” The questions were designed according to a depression rating scale widely used in the treatment of psychological illness.
In total, the AI analyzed 125,000 photos. Artificial intelligence took into account even small changes such as gaze direction, eye movements and head position. In addition, the neural network determined the level of illumination of the room in which the person was located.
The application has learned to identify people experiencing depression with 75% accuracy. Within five years, the authors plan to increase this figure to 90%. Potentially, MoodCapture will make it possible to effectively assess the dynamics of the development of a mental disorder, quickly respond to the deterioration of a person’s condition and propose solutions to combat it.