TITLE:
An Application of Kalman Filtering and Artificial Neural Network with K-NN Position Detection Technique
AUTHORS:
Hakan Koyuncu, Baki Koyuncu
KEYWORDS:
RFID, RSSI, WSN, ANN, Kalman Filtering, MLP, SDE, AP
JOURNAL NAME:
Wireless Sensor Network,
Vol.9 No.8,
August
10,
2017
ABSTRACT: RFID technology is one of the important technologies to determine the object
locations. Distances are calculated with respect to calibration curves of RSSI
amplitudes. The aim of this study is to determine the 2D position of mobile
objects in the indoor environment. The importance of the work is to show
that localization by using Artificial Neural Network plus Kalman Filtering is
more accurate than using classical KNN method. An indoor wireless sensing
network is established with strategically stationed RFID transmitter nodes and
a mobile object with a RFID receiver node. A fingerprint map is generated
and K-Nearest Neighbourhood algorithm (KNN) is deployed to calculate the
object locations. Fingerprint coordinates and RSS values received at these coordinates
are deployed to set up an Artificial Neural Network (ANN). This network
is used to determine the unknown object locations by using RSS values
received at these locations. The accuracy of object localization is found to be
better with ANN technique than KNN technique. Object coordinates, determined
with ANN technique, are subjected to Kalman filtering. The results
show that localization accuracies are improved and localization error distances
are reduced by 46% with the deployment of ANN + Kalman Filtering.