Telehealth, COPD, Cough, Unobtrusive monitoring, Long-term acoustic monitoring, Classification, Computational linguistics. Natural language processing, P98-98.5, Electronic computers. Computer science, QA75.5-76.95

Abstract For management of chronic respiratory diseases, unobtrusive longitudinal monitoring of cough has been proposed. Such a monitoring system was developed using a classifier trained on an initial observation period. After this initial period, a personalized system is available being optimized for the patient and the particular acoustic environment. Long-term deployment of the system requires that the extracted features and learned model characterizing the coughs (and its environment) are time-invariant. This is studied by an example using annotation of two largely different epochs. The results suggest that time-invariance of the cough sound is sufficiently guaranteed for practical deployment, but that changing acoustic environmental conditions may be a factor to reckon with. Cues for detecting changing situations are discussed.