摘要阻塞性睡眠呼吸暂停/低通气综合症（OSAHS）是睡眠呼吸障碍领域的一种常见病， 症状为打鼾和呼吸暂停。研究发现鼾声信号携带鼾声来源和上气道阻塞部位的重要信息， 因此可通过分析鼾声信号的声学特征来实现 OSAHS 诊断。采集到的整夜鼾声信号数据量 大，鼾声自动分段可以节约大量时间，有助于鼾声信号数据库的建立和后续的研究工作。 同时鼾声数据中不可避免地混杂普通呼吸声，这些呼吸声对于 OSAHS 诊断并无用处，区 分鼾声信号和普通呼吸声信号可减少提取出的片段数，提高工作效率。75937

本文依托国家自然科学基金面上项目（“基于声学分析的鼾症人群鼾声来源及上气 道阻塞部位识别”，No。61271410），研究了一种基于概率模型的鼾声自动分段方法。根 据改进的最小统计量控制递归平均算法计算得到每个时频点的有声概率，计算过程由两 次包含平滑和最小值搜索的迭代过程组成。将一帧内有声概率为 1 的频率点数占比与门 限值比较可区分有声帧和无声帧，将 0-400Hz 频带内有声概率为 1 的频率点数占比与门 限值比较可区分鼾声信号和普通呼吸声信号，最终实现自动分段与鼾声片段提取。基于 实测 OSAHS 患者鼾声数据的实验结果验证了本文所述方法的有效性。

毕业论文关键词 阻塞性睡眠呼吸暂停/低通气综合症 鼾声信号 自动检测 概率模型

毕 业 设 计 说 明 书 外 文 摘 要

Title The Design and Implementation of Automatic Segmentation of  Snoring Signals  Based on Model of Probability

Abstract Obstructive Sleep Apnea / Hypopnea Syndrome(OSAHS) is prevalent disorder among the community。 Snoring and apnea are typical symptoms for OSAHS。 Snore signals have been found to carry important information about the snore source and obstruction site in the upper airway, thus snoring signal characteristics can be used for diagnosis of OSAHS。 Due to a huge amount of snoring signals data collection lasting through the whole night, automatic segmentation of snoring signals can save a lot of time and help to establish the snoring signals database for further research。 Snoring signals are inevitably mixed with normal breathing signals, which are useless for diagnosis of OSAHS。 Thus distinction between snoring signals and normal breathing signals can reduce the number of segments and improve work efficiency。

Relying on the National Natural Science Foundation of China (61271410): " The recognition of snoring signal resource and upper airway obstruction site from snoring crowd based on the acoustic analysis", this paper elaborate a method of automatic segmentation of snoring signals based on model of probability。 Firstly, sound presence probability of each time frequency point is calculated according to the improved minima controlled recursive averaging。 The calculation procedure comprises two iter-ations of smoothing and minimum tracking。Then, detection thresholds of two parameters is determined to achieve automatic detection and segmentation of snoring signals。 One of the parameters is the proportion of frequency point where sound presence probability is 1 throughout the entire frequency band。 This parameters is used for detect sound。 The other is the proportion of frequency point where sound presence probability is 1 within 0-400Hz frequency band。 This parameters is used for distinguish snoring signals from normal breathing signals。 Audio data of snoring signals measured in patients OSAHS demonstrated the effectiveness of the methods described in this paper。

Keywords Obstructive  Sleep  Apnea  /  Hypopnea  Syndrome, Snoring  Signals, Automatic detection , Model of probability