基于鼾声识别的睡眠呼吸暂停监测与体位干预系统设计

Design of Sleep Apnea Monitoring and Postural Intervention System Based on Snoring Recognition

摘要: 打鼾是阻塞性睡眠呼吸暂停综合征(OSAS)患者的常见症状，会降低打鼾者及其床伴的生活质量。为了缓解打鼾症状、通过体位治疗(PT)有效治疗轻中度OSAS，本文提出并实现了一种基于鼾声信号的睡眠呼吸暂停综合征(OSAS)检测与体位干预系统。该系统包含集成鼾声检测、呼吸暂停事件识别、睡眠周期预测与非侵入式体位干预模块，通过手机内置传感器为核心的卷积神经网络鼾声识别模型，实时捕捉和分析打鼾特征，检测到鼾声信号，进一步判断呼吸暂停事件后，体位干预系统会自动调节枕头的高度与形状，改变患者睡姿或颈部俯仰角度，减少鼾声及睡眠呼吸暂停事件的发生。我们对20名患者进行了临床实验，结果表明，使用新型体位干预系统，降低了患者的睡眠呼吸暂停指数(AHI)、睡眠呼吸暂停事件(AHE)及呼吸暂停发作的持续时间，并且体位干预系统的治疗有效率为80%。

Abstract: Snoring is a prevalent symptom in patients with obstructive sleep apnea syndrome (OSAS) and can significantly diminish the quality of life for both the snorer and their bed partner. To alleviate snoring symptoms and effectively treat mild to moderate OSAS through postural therapy (PT), this article proposes and implements a detection and intervention system based on snoring signals. The system encompasses integrated modules for snoring detection, apnea event recognition, sleep cycle prediction, and non-invasive posture intervention. Utilizing a convolutional neural network for snoring recognition, the system leverages the built-in sensor of a mobile phone to capture and analyze snoring characteristics in real time, thereby detecting snoring signals. After further determining the apnea event, the postural intervention system automatically adjusts the height and shape of the pillow, alters the patient’s sleeping posture, or modifies the neck pitch angle to reduce the frequency of snoring and sleep apnea events. Clinical experiments conducted on 20 patients demonstrated that the new postural intervention system effectively decreased the patients’ apnea-hypopnea index (AHI), the number of sleep apnea events (AHE), and the duration of apnea episodes, achieving a treatment effectiveness rate of 80%.