Estimation of Human Movements from Body Acceleration Monitoring Using One Nearest Neighbor Method - Ubiquitous Health Care

We are developing an ubiquitous system for constant monitoring of human health that uses a 3D accelerometer and an electrocardiogram communicating wirelessly to a personal computer. The should provide an automatic alert whenever a person’s condition indicates a risk of heart attack or other danger. We collected acceleration data and bodily movement records from our staffs and from volunteer in order to determine from the movement data what sorts of movements, such as walking or riding a bicycle, a person is engaging in. Our system is able to infer the type of human movements from the acceleration data and use this in combination with ECG data to decide whether to issue an alert. Recently, we used a threshold method for the estimating what types of movements were being detected. Then we recently improved on our movement estimates by using normalization and the one nearest neighbor method.