Towards on-site instruction of cardio pulmonary resuscitation with wearable sensors

On serious accidental situations, such that there is a cardiac arrest patient in a city, giving quick cardiopulmonary resuscitation (CPR) by a bystander increases the resuscitation probability significantly. However, we cannot always expect that a person who can perform correct CPR exists on the location. With the aim to support CPR where an accident happens, we develop a system to teach the correct way of CPR using wearable sensors. First, we examined sensor data features to verify the correct way of CPR. Then, experiments are conducted with four subjects. The result showed that our method could validate correct CPR with more than about 93% accuracy. Also, we compare instruction method of "visual instruction", "voice instruction", and "voice and metronome instruction" from the view point of the effectiveness and the capability of situational awareness of a person performing CPR. The result implied that "voice instruction" is preferable for CPR support.