An Echo State Network based pedestrian counting system using wireless sensor networks

Abstract

A pedestrian counter has a lot of applications like effective resource utilization, planning of service activities and ensuring safety and convenience. The design and implementation of a new intelligent pedestrian counter is presented in this paper. The counter is publicly usable, low cost, easily deployable and scalable. We used off-the-shelf components for our design and the overall cost is less than 200 euro. The counter works in distributed mode and has wireless communication facilities. The hardware platform consists of PIR sensor units that detect the pedestrian movements, the wireless sensor nodes that handle the sensor data acquisition and transmission and the base station computer that process the data. We have trained an echo state network and this recurrent neural network functions as the brain for the the pedestrian counter. A system model was constructed using Simulink for generating the training data for the neural network. A modular layered software framework was designed for processing the pedestrian counter data. The echo state network successfully learned the various motion patterns and the pedestrian counter gave a reasonably good performance of 80.4%. To improve the performance further, redesigning the systems using low cost Active IR distance sensor is suggested.