Core Technology and Application Development Based on M2M Networking
As getting elder, the senior citizens have special assistance to do their daily routine. The monitoring of the elder, especially family, relative, or friends is quite important to prevent them from undesirables’ things. In case of monitoring user, the ability of person recognition is quite important, especially if there are multiple people in the room. The ability of recognition people can help to notify if abnormal situation happened to the user, in this case is elder’s people. Based on that situation, we present EOY (Eye On You), a more ubiquitous and mobile PID solution via data fusion. The EOY aim address the person identification problem, which means it has better quality in person activity monitoring. As single-technology approaches usually work poorly for such needs, we propose a data fusion approach, which integrates a 3D depth camera with wearable devices equipped with inertial sensors. The scenario considered in this work is a group of people stand in front of a movable depth camera and perform some actions. Each person has a wearable device attached to his/her wrist, and each wearable device has been registered with a unique ID.
Role: Research Asisstant
Ministry of Science and Technology, Taiwan
The detail of the EOY as follows, it contains three fusion algorithms: SM (stop-and-move) algorithm, AC (acceleration correlation) algorithm, and 3AC (3-acceleration-correlation) algorithm. The SM algorithm compares temporal correlation between skeleton data and inertial data. By contrast, to further eliminate the negative effect of asynchronous timing between the camera and wearable devices, we enable the AC algorithm and the 3AC algorithm to consider not only temporal but also spatial correlation between skeleton data and inertial data. The above fusion algorithms quantify correlation to determine similarity scores. With the similarity score of each combination of skeleton data and inertial data, we further propose a pairing scheme that matches skeletons with wearable devices in an one-to-one manner. Via such similarity-based pairing, we are able to conduct PID in a more mobile way.