【作者】 陈明杰；

【导师】 杨浩；

【作者基本信息】 重庆大学 ， 电工理论与新技术， 2002， 硕士

【摘要】 心率变异性(Heart rate variability，HRV)是了解人体植物神经系统功能状态的一种方便的窗口，已成为心血管疾病临床研究的一个前沿热点。 本文在分析研究国内外有关心率变异信号实时采集及后期分析方法的基础上，设计完成了模拟心电信号接口放大器，使用Protel和Matlab对滤波电路进行仿真，研究了元器件参数变化对滤波器的影响，经实际检测符合设计要求。 编制了实时采集、显示、存储软件，该软件能设定采样频率和采样时间，最高频率可达1000Hz。利用数字滤波器实现了50Hz干扰信号的实时滤波。 采用传统的方法设计了低通数字滤波器，比较了用传统FFT和现代谱估计得到的谱的分辨能力，实际验证了现代谱估计功率谱的分辨率优于经典谱估计方法。 对23例正常人(坐姿)15分钟HRV信号分别采用FFT和现代谱估计得到的功率谱进行统计，发现两种方法所得VLF％、LF％、HF％基本吻合(相同性别相应年龄段)。发现随着年龄的增加(20-60岁)，两种方法所得的LF/HF、VLF/HF比值基本呈现增加的趋势，与临床认识吻合。 对19例正常人(卧姿)5分钟HRV信号分别采用FFT和现代谱估计得到的功率谱进行统计，发现两种方法所得VLF％、LF％有较大差异，而HF％基本吻合(相同性别相应年龄段)。推测由于5分钟HRV信号点数较少，FFT分析时分辨率不高，因此两种方法所得极低频VLF％和低频LF％有较大差异。现代谱估计所得的VLF/HF比值随年龄(＞20岁)增加基本呈现增加的趋势，与临床认识吻合。建议做频域分析时，应采集分析15分钟以上信号。 对4例女性心肌缺血病人卧姿5分钟HRV信号分别采用FFT和现代谱估计得到的功率谱进行统计，发现总功率TF和VLF/HF与正常人相应年龄段有较大差异。 对10例女性甲亢病人卧姿5分钟HRV信号分别采用FFT和现代谱估计得到的功率谱进行统计，证实甲亢病人VLF/HF比值比正常人相应年龄段高。 本文还对心率变异信号的维格纳分布作了研究，选用Choi-Williams窗，确定了滤波函数的参数σ=0.05，实际证明抑制交叉项的干扰效果较好。维格纳方法在三维立体图上展示心率变异信号的时频分布时，能揭示出随时间演变频率演变的信息，而这些信息通过单纯的频域分析是不能得到的。医生可以查找任一时段的不同谱峰分布，也可以查看任一频段对应的时域情况。对实际的心率变异信号进行了维 重庆大学硕士学位论文格纳分析，证实通过计算三维时频分布对频率轴的投影得到的各频段（高频段、低频段等）功率，与直接用FFT所得各频段功率，二者只相差一个比例系数，并且二者LF／HF的值基本相同。三维CWD立体图是医生观察时间与频率对应关系的有力工具。更多还原

【Abstract】 Heart rate variability(HRV) is a convenient window to study the autonomic nervous system of human body. It is a focus of studying cardiac disease in clinic.On the basis of the current study on real-time HRV collection and later processing home and abroad, ECG signal sampling interface amplifier is designed, It is simulated in Protel and Matlab. The further study on the effect of component parameter change for filter is done in the paper. It shows that the design is correct.Writing a software for real-time collecting .displaying. saving, this software can set sampling frequency and time ,the maximal frequency is lOOOHz. 50Hz interference is filtered by using digital notch filter.In this paper, digital Low pass filter is designed. AR model algorithm and FFT algorithm for resolving power spectrum is compared as well. It shows that AR model algorithm is better than FFT method.Analyzing 15 minute HRV(sit) of 23 healthy persons by using AR and FFT method. VLF%. LF%. HF% is tally with these 2 methods(same sex and same age segment).With the increasing of age(>20) , LF/HF. VLF/HF computed by using the 2 method is increasing , this is tallied with clinic.Analyzing 5 minute HRV(decubitus) of 19 healthy persons by using AR and FFT method, we found that VLF%. LF% by using 2 method is different, but HF% is tallied (same sex and same age segment). It is infered signal length is short, the spectrum resolving power is low by using FFT method, so the VLF% and LF% by using 2 method has great difference. The VLF/HF by using AR presents trend of increasing, this is tallied with clinic.Analyzing 5 minute HRV(decubitus) for 4 myocardial ifchemia sickness by using AR and FFT. We found that TF and VLF/LF by using the 2 methods are great different with healthy person. More than 15 minute signal should be sampled for analyzing spectrum.Analyzing 5 minute HRV(decubitus) for 10 hyperthyroibism sickness by using AR and FFT. We found that VLF/LF of hyper thyroibism sickness by using the 2 methodsinare high than healthy person.Study HRV WD, when using Choi-Williams window to restrain cross disturb, setting filter parameter a = 0.05. It is proved the effect is good. With 3D figure we can know HRV time-frequency distribution and the information of frequency change with time, while this information can not be gotten only in spectrum region. A doctor can find different spectrum distribution at each time segment, and can also look over frequency field corresponding time domain information. Analyzing real HRV, proving the every segment power(high frequency> low frequency) of projection in frequency axis is times with the every segment power by using FFT , the LF/HF by using 2 method is primarily sameness. 3D figure is a trenchancy tool for doctor to look over the relation between time and frequency.更多还原