Distinction of cough from other sounds produced by daily activities in the upper airways.

Objectives: The aim of this study was to validate the successfulness of our developed system for distinction between cough and other sounds which are present in daily human activities from the upper airways.

Background: To date, methods used for monitoring of cough sound were primarily subjective. A reliable measure of cough is needed so that the severity of cough in various patients and the effectiveness of treatment can be assessed.

Methods: Sounds of induced cough and sneezing, voluntary throat and nasopharynx clearing, forced ventilation and laughing, snoring, eructation, loud swallowing, and nasal blowing were studied. Characteristics of the sound events in 20 volunteers were calculated using the time-domain, spectral and non-linear analysis. The classification tree was constructed for classification between cough and non-cough sounds. We have validated the usefulness of our developed algorithm against subjective cough counts, which were performed by two trained observers.

Results: The value of sensitivity for distinction between cough and other sounds was 86% and the value of specificity was 91%. The value of sensitivity for distinction between voluntary and induced cough sounds was 96% and specificity was 43%. The value of sensitivity between cough sounds and voluntary throat clearing was 96% and specificity was 85%. The value of sensitivity between cough sounds and induced sneezing was 95% and specificity was 93%.

Conclusion: We have developed an algorithm for distinction between cough and other sounds with a relatively high degree of accuracy (Tab. 1, Fig. 5, Ref. 15).