【摘要】 本文提出用鼾声信号的频谱特性和完整上气道声学模型联合对阻塞性睡眠呼吸暂停低通气综合征患者的上气道阻塞部位做分类.用子带能量对数比能较好地区分上气道中两种不同的阻塞模式,还利用自回归滑动平均模型对上气道的生理结构进行估计.使用子带能量对数比和咽腔内声管的横截面积共同作为支持向量机的输入特征,对20名受试者包含有软腭游离缘平面以上和以下阻塞两类鼾声片段,共4 638个做分类.结果表明:分类的非加权平均正确率达85%,说明用本文的方法预测阻塞性睡眠呼吸暂停低通气综合征患者上气道阻塞部位的有效性.更多还原

【Abstract】 This paper proposes to classify the upper airway obstruction site of patients with Obstructive Sleep Apnea Hypopnea Syndrome(OSAHS)by using the spectral characteristics of the snoring signal and the complete upper airway acoustic model.In this paper,the sub-band energy logarithm ratio can better distinguish two different blocking modes in the upper airway.This paper also uses the Auto-Regressive Moving Average(ARMA)model to estimate the physiological structure of the upper airway.The sub-band energy log ratio and the pharyngeal acoustic tube cross-sectional area are used together as the input characteristics of the support vector machine.There are a total of 4 638 episodes of snoring from the 20 subjects including the soft sputum free edge plane up and below using in experiments.The results show that the unweighted average recall is 85%.The proposed method in determining the upper airway obstruction site in OSAHS is effective.更多还原