Implementation of Dead Reckoning System Using Fingerprint and K-NN Algorithm for An Object Position and Posture Estimation

Limitation of GPS in indoor area become the main problem for indoor localization system. Dead reckoning system usually used inertia sensor or GPS for localization system. Initial location becomes a lack for the dead reckoning system due to the requirement of additional algorithm or additional device. Fingerprint method and K-Nearest Neighbor (K-NN) Algorithm can be applied as a dead reckoning system to estimate object position in the observation area. Posture estimation system is implemented in system by using IMU sensor to provide information that the object is standing, sitting or in laying posture. This paper proposed a system which can estimate the position of an object along with the object's posture. The implementation result shows that the average Mean Square Error (MSE) of proposed system using 3NN is 4.14 meters with 93.5% accuracy. Due to the computation is done in sever, estimation computing only takes time about 0.467 seconds. Object posture estimation shows 91% accuracy in 7 object's position.