一步优化OSAHS鼾声分类算法  

One-step Optimization of OSAHS Snoring Classification Algorithm

作　　者：余佳琪 王冬霞[1] 马晓冬 张严 YU Jiaqi;WANG Dongxia;MA Xiaodong;ZHANG Yan(School of Electronic and Information Engineering,Liaoning University of Technology,Jinzhou 121001,Liaoning,China;Dalian Fanyi Technology Co.,Ltd.,Dalian 116033,Liaoning,China)

机构地区：[1]辽宁工业大学电子与信息工程学院,辽宁锦州121001 [2]大连凡益科技有限公司,辽宁大连116033

出　　处：《实验室研究与探索》2023年第7期136-140,181,共6页Research and Exploration In Laboratory

摘　　要：针对人工标注分类阻塞性睡眠呼吸暂停低通气综合征(OSAHS)患者鼾声准确率低等不足,提出一步切割聚类的鼾声分类算法(BVCL)。该算法基于贝叶斯信息准则(BIC)对单切割点检测进行优化,实现鼾声多切割点检测和聚类,并采用语音端点检测(VAD)技术对切割点进行快速筛选;通过改进线性判别分析(LDA)算法矩阵,构建分类回归树(CART)-LDA分类模型,实现OSAHS患者鼾声、正常鼾声和非鼾声的一步三分类,提高OSAHS患者的筛查效率,为便携式睡眠鼾声监测设备走向应用提供一种可能。以某医科大学临床夜间实录鼾声数据作为输入,仿真结果表明,该算法与鼾声两步分类法和一步分类法相比,降低算法复杂度的同时提升了分类准确率。To address the low accuracy of manual labeling and classification of snoring in patients with obstructive sleep apnea-hypopnea syndrome(OSAHS),a one-step cut clustering for snoring classification algorithm(BVCL)is proposed in this paper.The algorithm is based on Bayesian information criterion(BIC)to optimize the single cut point detection and realize the multi-cut point detection and clustering of snoring.Additionally,the voice endpoint detection(VAD)technology is employed to quickly screen cutting points.At the same time,the classification and regression tree(CART-LDA)classification model is constructed by improving the matrix of linear discriminant analysis(LDA)algorithm,the one-step three-step classification of snoring,normal snoring,and non-snoring in OSAHS patients are realized.It improves the screening efficiency of OSAHS patients,and provides a possibility for portable sleep snoring monitoring equipment to become practical.With the data of clinical nocturnal snore recorded in a medical university,the simulation results show that the proposed algorithm can reduce the complexity and improve the classification accuracy compared with the two-step classification and one-step classification.

关 键 词：鼾声分类 鼾声检测 切割聚类 一步优化分类 

分 类 号：TN912.3[电子电信—通信与信息系统]