Children Tracking System Based on ZigBee Wireless Network and Neural Network

Abstract

The safety of children is one of the fundamental concerns of parents. Recently, child kidnapping has increased by a large percentage, some children have been found, and some children have not found yet. This paper proposes an indoor localization system based on ZigBee wireless sensor network (WSN) and Backpropagation Artificial Neural Network (BP-ANN) to locate the child in an indoor environment. Several ANN topologies were investigated to select the best one with minimum tracking or localization error. The Received Signal Strength Indicator (RSSI) was collected from four ZigBee XBee S2C anchor nodes by the mobile node carried by the child in an indoor area of 32m × 32m. The RSSI was collected from 127 test points inside the tested area. The measured RSSI was used to train, test, and validate the performance of BP-ANN to determine the two dimensions (2D) of the target child’s location. Different topologies of ANN have been examined for training, testing, and validation which are 5-5, 10-10, 15-15, and 20-20 neurons in the hidden layer. The findings indicate that the 20-20 ANN topology can achieve higher accuracy than other topologies. Additionally, 20-20 topology localization errors were 1.0, 1.157, and 1.356 m for training, testing, and validating ANN performance.