Coughing is one of the most frequent presenting symptoms of many diseases affecting the airways and the lungs of both humans and animals. In piggeries, the continuous on-line monitoring of cough sound can be used to build an intelligent alarm system for the early detection of diseases. In a first study, with experiments under laboratory conditions, algorithms have been developed to detect cough sounds and to classify the animals whether they were ill or not. In this study, the algorithm was tested in field conditions. Pig cough sounds were registered on 44 pigs, Landrace × Large White × Duroc crosses, 150 days old, with a mean weight of 60 kg. The sounds were recorded by an operator holding a microphone at approximately 20-50 cm from the pig's head. From these sound files, feature vectors were extracted using a filter bank approach combined with an amplitude demodulation. These feature vectors were compared to a reference set by means of a dynamic time warping algorithm. This approach leads to a two-class classification: cough sounds and other sounds. The classification resulted in correct classification of 85.5% for the cough sounds and 86.6% correct classification of the other sounds.
|Titolo:||Field test of algorithm for automatic cough detection in pig houses|
|Settore Scientifico Disciplinare:||Settore AGR/10 - Costruzioni Rurali e Territorio Agroforestale|
|Data di pubblicazione:||2008|
|Digital Object Identifier (DOI):||10.1016/j.compag.2007.08.016|
|Appare nelle tipologie:||01 - Articolo su periodico|