Improving the automatic identification of crackling respiratory sounds using fuzzy logic

Crackles are adventitious sounds that can be interpreted in many ways, depending on the type of signal in which they are located. In music applications, crackles are noisy phenomena usually associated with local degradations in the audio source. Another field of application is the biomedical one; for example, crackling sounds detected in the human respiratory sound are an evidence of cardiovascular and pulmonary diseases. Identification and recognition of crackles, by means of a full algorithmic approach, is not as accurate as human expertise, therefore the method presented in this paper includes a fuzzy logic process, in which the experts knowledge is transferred. The fuzzy inference subsystem analyzes the breath sound in order to extract the parameters required by the identification algorithm. The aim of this project is to implement a flexible and automatic method to recognize crackles in breath sounds.