Heart Attack Detection Using Body Posture and Facial Expression of Pain

It is not uncommon for a person to be alone when they have a heart attack. Getting help quickly can mean the difference between life and death. The pain a person feels during a heart attack may prevent them from seeking help in time, so automated methods are needed to detect and alert to such events. This article presents a proposal based on machine vision and deep learning to identify possible heart attacks. First, a typical human posture of a possible heart attack is identified using the upper body joints from skeletal studies. As similar non-infarct postures are possible, the posture analysis is integrated with the facial expression of pain to reduce the number of false positives. The proposed method has achieved 93.33% accuracy in detecting myocardial infarction.