Cough detection using fuzzy classification

Abstract

In this paper a biomedical signal interpretation problem from the field of artificial ventilation is presented: cough detection. The problem was analyzed and solved by application of fuzzy classification methods using data from volunteers breathing through a mouthpiece. In a first step a fuzzy classification system is designed manually. Then, a fuzzy classification system is generated automatically utilizing the fuzzy-neuro method Fuzzy RuleNet. In a third step a noise elimination cycle for Fuzzy RuleNet is introduced to eliminate neurons i.e. rules, created by noise and artifacts as encountered in this and similar applications. The comparison of the manually and the two automatically generated systems shows that the automated method Fuzzy RuleNet is able to generate a fuzzy classification system with performance and complexity comparable to the manually generated system.