Localização de fontes sonoras em recintos fechados aplicada à televigilância médica

Abstract

Due to the worldwide people ageing phenomenon, the health care at home (home care), a more humanized and cost saving method, for elderly people and long term care, is nowadays more and more usual. The problem is to assure prompt help when needed. A surveillance system can be the solution to improve caregiver awareness and patient safeness. This work investigates the possibility of indoor sounds direction of arrival detection, selection and determination, attempting to find out signals of a patient fall or collapse. The sounds after an acoustic event are recorded using a microphone array. If they pass the selection criteria, they are processed by a Direction of Arrival (DOA) detection algorithm. Five DOA algorithms were implemented and evaluated empirically. Three of them are already known, nominally, Generalized Cross Correlation (GCC), Generalized Cross Correlation - Phase Transform (GCC-PHAT), Degenerate Unmixing and Estimation Technique (DUET), and two are proposed at this work, Atack Border Detector and Spectrogram Technique. Performances were evaluated by taking the home care perspective into account.