Towards Predicting Risky Behavior Among Veterans with PTSD by Analyzing Gesture Patterns

Risky behavior including violence and aggression, self-injury, anger outburst, domestic violence along with self-injury, sexual abuse, rule-breaking, use of drugs and alcohol, suicide, etc. are alarming issues among US military veterans who return from combat zone deployment in Iraq and Afghanistan. Veterans are exposed to trauma in war zones which affect most of them with post-traumatic stress disorder (PTSD) or other mental health problems to some degree. Studies have shown that veterans have much higher rates of PTSD than civilians and are more likely to engage in risky behavior. One of the forms of displaying and engaging in risky behaviors is through gestures. We collaborated with veterans and social scientists to find the list of 13 gestures that are often used by veterans engaged in risky behaviors. In this research work, we have collected accelerometer data from subjects performing the gestures mentioned above and have tried to detect them using machine learning techniques. This paper describes identifying gesture clusters from the accelerometer coordinate data and development of a predictive model that can classify the gestures resulting in the prediction of risky behaviors among the veterans who suffer from PTSD.