HMM-Based Indoor Localization Using Smart Watches' BLE Signals

Abstract

This paper describes the implementation of indoor localization technology using Bluetooth modules and beacons featuring Bluetooth Low Energy (BLE) by smart watches. We implemented a Hidden Markov Model (HMM)-based fingerprinting method using various data from recognized BLE signals. For fingerprinting, we obtained both a received signal strength indication and a signal observation frequency that were obtained from nearby beacons. The location was estimated from both a presurveilled profile and an exponential fit model. When using an exponential fit model with the signal observation frequency, we were able to achieve approximately 80% accuracy, even with little data. In addition, when using the HMM-based fingerprinting and a transition model based on the probability of users' movement, the accuracy of our location prediction increased by up to 15%.