Azarbarzin Ali, Moussavi Zahra
Department of Electrical and Computer Engineering, University of Manitoba, Winnipeg, MB, Canada R3T
Annu Int Conf IEEE Eng Med Biol Soc. 2011;2011:1749-52. doi: 10.1109/IEMBS.2011.6090500.
Snoring sounds is commonly known to be associated with obstructive sleep apnea (OSA). There are many studies trying to distinguish between the snoring sounds of non-OSA and those of OSA patients. However, OSA is only one of the conditions that affect the structure of upper airway. In this study, we investigated the effect of anthropometric parameters on the snoring sounds. Since snoring sounds are non-Gaussian signals by nature, we derived its Higher Order Statistical (HOS) features and investigated the statistical significance of the anthropometric parameters on each of these features. Data were collected from 40 patients with different levels of OSA. Tracheal respiratory sounds collected by a microphone placed over suprasternal notch, were recorded simultaneously with full-night Polysomnography (PSG) data during sleep. The snoring segments were identified semi-automatically from respiratory sounds using an unsupervised snore detection algorithm. The bispectrum of each SS segment was estimated. We calculated two common HOS measures, Skewness and Kurtosis, plus a new feature called Projected Median Bifrequency (PMBF) from the SS segments. Then, we investigated the statistical relationship between these features and anthropometric parameters such as height, Body Mass Index (BMI), age, gender, and Apnea-Hypopnea Index (AHI). The result showed that gender, BMI, height, and AHI are the parameters that do change the characteristics of snoring sounds significantly.
鼾声通常被认为与阻塞性睡眠呼吸暂停（OSA）有关。有许多研究试图区分非OSA患者的鼾声和OSA患者的鼾声。然而，OSA只是影响上呼吸道结构的病症之一。在本研究中，我们调查了人体测量参数对鼾声的影响。由于鼾声本质上是非高斯信号，我们推导了其高阶统计（HOS）特征，并研究了人体测量参数对这些特征中每一个的统计显著性。数据收集自40名不同程度OSA的患者。通过放置在胸骨上切迹上方的麦克风收集气管呼吸音，并在睡眠期间与全夜多导睡眠图（PSG）数据同时记录。使用无监督的鼾声检测算法从呼吸音中半自动识别出鼾声段。估计每个鼾声段的双谱。我们从鼾声段计算了两个常见的HOS度量，即偏度和峰度，以及一个名为投影中位数双频率（PMBF）的新特征。然后，我们研究了这些特征与身高、体重指数（BMI）、年龄、性别和呼吸暂停低通气指数（AHI）等人体测量参数之间的统计关系。结果表明，性别、BMI、身高和AHI是确实会显著改变鼾声特征的参数。
