Support Vector Machine Algorithm in a Bluetooth Low Energy Passenger Detection System

Abstract

Internet of Things (IoT) and Bluetooth Low Energy (BLE) technologies are drawing attention in the trend to integrate electronic devices and services. One example of an IoT application is a smart fare collection/payment system. Because people carry transportation cards nowadays, it is troublesome to tap the card every time, but making payment automatic by eliminating the tapping operation lets a greater number of passengers take public transportation in a shorter time. The first step toward a smart transportation fare collection system is passenger detection. One way to detect a passenge is by received signal strength indication (RSSI). The basic idea consists of adding BLE transmitters (beacons) to the vehicles and comparing intensity (power) between signals transmitted and received. However, due to multi-path propagation and fading effects, RSSI is not sufficiently accurate to calculate an exact location. Therefore, instead of determining the exact position, it is preferable to only check if the passenger is inside or outside the vehicle. In this support vector machine estimation method, RSSI signals are filtered using a moving average filter. Algorithm estimation efficiency was 92.51% (on average) for one beacon, and close to 100% for five beacons.