基于听觉子带关联维数的SAHS鼾声筛查

摘要

提出一种依据鼾声录音筛查睡眠呼吸暂停低通气综合征严重程度的新方法,使用关联维数计算方法分析鼾声的非线性性质.根据听觉心理感知尺度在整个频带上划分子带,通过计算听觉子带关联维数向量,刻画睡眠呼吸症状严重程度的分布状况.提取各类鼾声的子带关联维数特征,以期训练出高斯混合模型,判断SAHS患者病情的严重程度.该方法与黄金标准多导睡眠监测诊断相比,其诊断患病严重程度的正确率为90％,相关系数为0.96 (P＜0.001).实验表明该方法能有效筛查鼾症严重程度,对辅助医疗和居家监护等有积极作用.%This paper proposes a method for screening sleep apnea hypopnea syndrome (SAHS) by recording snore sound.Based on the equivalent rectangular bandwidth (ERB) scale used in psychoacoustics,the ERB correlation dimension (ECD) of snore is used to analyze and classify snores of different severity levels.For the training group,snore episodes were manually segmented and ECD features of snores were extracted to establish a Gaussian mixture model (GMM).The nocturnal snoring sound of the test group was validated to detect SAHS snores.The apnea hypopnea index (AHI) was estimated,to determine severity of SAHS.AHI is an average number of apneic events per hour of sleep.Compared with the polysomnography diagnosis that is a gold-standard,accuracy of SAHS severity reached 90％.The proposed method is useful in accessory examination of SAHS severity and for home uses.