基于小波分解的鼾音信号功率谱估计方法研究  

Research of Power Spectrum Estimation Method in Snoring Sound Based on Wavelet Decomposition

作　　者：刘勇 袁立飞 袁丽峰 LIU Yong;YUAN Lifei;YUAN Lifeng(Hebei Eye Hospital,Xingtai 054000)

机构地区：[1]河北省眼科医院,邢台054000

出　　处：《计算机与数字工程》2022年第3期650-655,共6页Computer & Digital Engineering

基　　金：河北省重点研发计划(编号:182777191)资助。

摘　　要：为对鼾音信号进行有效地功率谱估计,同时精确研究鼾音和阻塞位置的关系,提出了基于小波分解的鼾音功率谱估计方法。首先将鼾音信号进行小波分解,然后对提取的三个子带信号分别采用Welch法、多窗谱法、基于AR模型的自相关法和Burg法进行功率谱估计,最后分析了不同阻塞位置的鼾音功率谱估计特点。仿真结果表明,基于小波分解Welch法得到的鼾音功率谱估计分辨率和方差性能可以满足需要。不同阻塞位置的鼾音功率谱在2000Hz~4000Hz子带差别显著,在2000Hz~4000Hz子带中,上部阻塞鼾音信号的平均方差远小于下部阻塞鼾音信号。To estimate the power spectrum in snoring sound and research the relationship between snoring sound and obstructive position more accurately,an algorithm for estimating the power spectrum in snoring sound based on wavelet decomposition is presented.Snoring sound is decomposed by wavelet analysis firstly,then methods about Welch and multiple window spectrum and AR model autocorrelation and Burg are used to estimate the power spectrum of the three subband signal,finally the power spectrum estimation characteristics about snoring sound in different obstructive position are analyzed.The experimental results show that the resolution and variance performance of snoring sound power spectrum estimation used Welch method based on wavelet decomposition are effective.The power spectrum of snoring sound in different obstructive position varies greatly in the 2000Hz~4000Hz subband.The average variance of the upper blocking snoring sound is much smaller than that of the lower blocking snoring sound in this subband.

关 键 词：鼾音 小波分解 WELCH法 多窗谱法 自相关法 Burg法 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]