Daniel Berckmans

Automated recognition of spontaneous versus voluntary cough.

Cough or cough epochs may be an important and persistent symptom in many respiratory diseases requiring both a continuous and objective observation. The research presented in this paper is aimed at assessing a blind data-based classification between spontaneous and voluntary human cough on individual sound samples. Cough sounds were registered in the free acoustic field on 3 pathological and 9 healthy non-smoking subjects, all aged between 20 and 30. Each sound is represented by the normalized power spectral density (PSD). Different transformations of the cough PSD-vector are chosen as input features to the classification algorithm. An experimental error rate comparison between different neural and fuzzy classification networks is performed. All evaluated algorithms used the Euclidean metric. This resulted in a correct class-discrimination between spontaneous and voluntary cough for 96% of the cough database.

Fuzzy approach for improved recognition of citric acid induced piglet coughing from continuous registration

Abstract A natural acoustic indicator of animal welfare is the appearance (or absence) of coughing in the animal habitat. A sound-database of 5319 individual sounds including 2034 coughs was collected on six healthy piglets containing both animal vocalizations and background noises. Each of the test animals was repeatedly placed in a laboratory installation where coughing was induced by nebulization of citric acid. A two-class classification into ‘cough’ or ‘other’ was performed by the application of a distance function to a fast Fourier spectral sound analysis. This resulted in a positive cough recognition of 92%. For the whole sound-database however there was a misclassification of 21%. As spectral information up to 10000 Hz is available, an improved overall classification on the same database is obtained by applying the distance function to nine frequency ranges and combining the achieved distance-values in fuzzy rules. For each frequency range clustering threshold is determined by fuzzy c-means clustering.

Development of new control techniques for the ventilation and heating of livestock buildings

From statistical analysis of field data, normal mechanical ventilating systems using exhaust fans were found to give no better mean production results than natural open-ridge systems. To explain this, components of the mechanical ventilating system, namely the fan and the controller, were analysed in a laboratory test installation. The equipment tested revealed important shortcomings that explain why normal mechanical ventilating systems do not perform adequately in the field. By use of a steady-state analysis, the effects of the air flow rate on indoor temperature and on energy losses in a livestock building were analysed. A new controller has been developed, in an attempt to achieve improved control of air flow and hence more effective environmental control in livestock buildings.

Considering the influence of artificial environmental noise to study cough time–frequency features

Abstract In general the study of the cough mechanism and sound in both animal and human is performed by eliciting coughing in a reproducible way by nebulization of an irritating substance. Due to ventilation the controlled evaporation-protocol causes artificial noises from a mechanical origin. The resulting environmental low-frequency noises complicate cough time–frequency features. In order to optimize the study of the cough-sound the research described in this paper attempts on the one hand to characterize and model the environmental noises and on the other hand to evaluate the influence of the noise on the time–frequency representation for the intended cough sounds by comparing different de-noising approaches. Free field acoustic sound is continuously registered during 30 min citric acid cough-challenges on individual Belgian Landrace piglets and during respiratory infection experiments, with a duration of about 10 days, where room-ventilation was present.