基于FCM聚类算法的呼吸音分类识别研究  

Study on breath sound recognition based on FCM clustering algorithm

作　　者：徐学良 易鹏飞[2,3] 王海滨[2,3] 张凯[2,3] 秦国瑾 

机构地区：[1]二0三医院外一科,黑龙江齐齐哈尔161000 [2]西华大学电气信息学院,四川成都610039 [3]四川省信号与信息处理省重点实验室,四川成都610039

出　　处：《现代电子技术》2015年第12期16-19,共4页Modern Electronics Technique

基　　金：四川省重点实验室开放研究基金资助项目(szjj2013-014)

摘　　要：为了分类识别不同部位的呼吸音,提出一种基于呼吸气持续时间相对值参数,以及呼吸气间歇时间相对值参数提取的呼吸音分析方法。该方法包括呼吸音预处理、包络提取、基于FCM聚类算法的自适应阈值线的选取、特征值提取为主的4个部分。对50例3种不同呼吸音进行分类实验,分类准确率为94%,从而验证了该方法的有效性。在此基础上,对该方法在同一种呼吸音病理和健康研究方面做出了展望。In order to achieve classification identification of the respiratory sounds from different parts,a respiratory sound analysis method based on extracting the relative parameter of inhalation and exhalation duration,as well as the relative parameter of inhalation and exhalation intermittent time is proposed. This method includes respiratory sound preprocessing,envelope extraction,adaptive threshold line selection based on FCM clustering algorithm and feature extraction. In this paper,50 cases from three different respiratory sounds are classified. The classification accuracy is 94%. Thus,the validity of this method was verified. Furthermore,the prospect in research on pathology and healthcare of respiratory sounds is made for the future research.

关 键 词：呼吸音 呼吸音分类 MORLET小波 FCM聚类算法 

分 类 号：TN911.34[电子电信—通信与信息系统]