Objective monitoring of cough sound for extended period is an important step toward a better understanding of this symptom. Because ambulatory cough monitoring systems are not commercially available, we prepared own monitoring system, which is able to distinguish between voluntary cough sound and speech in healthy volunteers. 20-min sound records were obtained using portable digital voice recorder. Characteristics of the sound events have been calculated in time and frequency domains and by a nonlinear analysis. Based on selected parameters, classification tree was constructed for the classification of cough and non-cough sound events. We validated the usefulness of our algorithm developed against manual counts of cough obtained by a trained observer. The median sensitivity value was 100% (the interquartile range was 98-100) and the median specificity was 95% (the interquartile range was 90-97). In conclusion, we developed an algorithm to distinguish between voluntary cough sound and speech with a high degree of accuracy.