However, a large variety of possible random ...

In this paper we propose a reliable classification strategy for distinguishing healthy people and patients with pulmonary emphysema based on lung lasix heart failure sounds. Symptom of emphysema that almost all the sounds of light include custom (ie, random) sounds. However, a large variety of possible random sounds and noises with auscultation makes high-precision detection difficult. To overcome these difficulties, our strategy is to accept two-tier classification approach based on identifying "certain abnormal breath." In the first stage, hidden Markov models and bihram model used for the acoustic features and acoustic segments of each period of abnormal breathing, respectively, to calculate two types of stochastic likelihood: high probability for a segment of the sequence to be abnormal breathing and the likelihood for normal breathing. In the second stage, patients are determined based on the detection of abnormal breathing are confident that when the difference between the two likelihood over the set threshold. Our strategy has achieved the highest classification rate of 88. 7% of healthy and sick among the three main strategies for classification, showing the validity of our approach. .