基于粒子群优化-支持向量机的睡眠呼吸暂停检测

Sleep Apnea Detection Based on Particle Swarm Optimization-Support Vector Machine

睡眠呼吸暂停(sleep apnea,SA)是一种睡眠障碍疾病,严重影响睡眠质量和身体健康。为降低睡眠呼吸障碍检测的复杂度并提高准确率,提出了一种粒子群优化-支持向量机(particle swarm optimization-support vector machine,PSO-SVM)方法,通过心电信号实现对SA的准确检测。首先,将心电信号分段,并从中提取心率变异性;其次,实现特征提取与选择,包含心电信号RR间期的均值、标准差、均值标准差、差值均方的平方、心率变异性的信号总功率、低频段功率、高频段功率、瞬时中位频率、边际谱熵和能量谱熵等;最后,通过PSO-SVM分类算法进行睡眠呼吸暂停检测。结果表明,筛选10个特征对SA进行检测,利用Apnea-ECG数据库通过PSO-SVM的检测准确率为94.0%,提升了现有方法的检测性能。

Sleep apnea(SA)is a sleep disorder disease,which seriously affects sleep quality and health.In order to reduce the complexity and improve the accuracy of sleep apnea detection,a particle swarm support vector machine(PSO-SVM)method was proposed to accurately detect SA through ECG.Firstly,the ECG signal was segmented and the heart rate variability was extracted from it.Secondly,feature extraction and selection were realized,including the mean,standard deviation,mean standard deviation,square of mean square of difference,total signal power of heart rate variability,low-frequency power,high-frequency power,instantaneous median frequency,marginal spectral entropy and energy spectral entropy of ECG RR interval.Finally,sleep apnea was detected by the PSO-SVM.The results show that this method selects 10 features to detect SA,and the detection accuracy through PSO-SVM using Apnea-ECG database is 94.0%,which improves the detection performance of the existing methods.