PROJECTE DE FI DE CARRERA Acoustic Alarm Detection Based on Matched Filter and Morphological Tools (Detecció d'Alarmes Acústiques Basada en Filtre Adaptat i Eines Morfològiques)

Newborns of a gestational age of 24-32 weeks are commonly very weak and have health problems. The facilities of a Neonatal Intensive Care Unit (NICU) provide the medical equipment needed for their care and in some cases, for their survival. The acoustic environment of a typical NICU is very rich and may contain a large number of different sounds, which come either from the equipment or from the human activities taking place in it. Alarm sounds triggered by biomedical equipment play a key role in providing healthcare in a NICU, but at the same time loud sounds or particular sounds may be harmful for the neurological development of preterm infants. The aim of this project is to develop a system that allows to detect these alarm sounds. Previously, several systems have been proposed for this task, all based on machine learning techniques. Our goal is to provide an alternative solution based on signal processing techniques. The system proposed in this work consists of a set of consecutive stages that convert the audio input into a binary data stream specifying each alarm detection in time. The experimental evaluations were performed with a database recorded in a real-world hospital environment (Hospital de Sant Joan de Déu). The performance of the detection system is assessed both at the frame level and at the alarm period level.