Mobile Indoor Localization with Bluetooth Beacons in a Pediatric Emergency Department Using Clustering, Rule-Based Classification and High-Level Heuristics

摘要

To mitigate anxiety, pain and dehydration in Pediatric Emergency Departments (PED), it is paramount to tailor educational, motivational and self-help content towards the current location inside the PED, since this reflects the current stage in their PED visit. However, accurately identifying the patient's indoor location in a real-world complex environment, such as a hospital, is still a challenging problem, with interference and attenuation from patients, staff, walls and various electromagnetic sources (e.g., imaging devices). We present an indoor localization methodology that achieve a best-effort localization accuracy given the available sensors, (low-quality) motion data and computational platforms. First, we utilize machine learning methods to find a suitable accuracy/granularity balance and then proceed by training a localization model. Then, we apply a set of heuristics based on motion data to eliminate false location estimates. We validated of our approach in a real-life busy and noisy PED with a 92% accuracy.