【作者】 张道明；

【导师】 谭国俊； 郝继飞；

【作者基本信息】 中国矿业大学 ， 电力电子与电力传动， 2008， 博士

【摘要】 心电图(electrocardiogram,简称ECG)是心脏电活动在体表的综合反映,对心电信号进行分析进而诊断心脏疾病具有重要研究价值。目前,心电信号分析与辅助诊断的研究内容主要涉及四个方面,即数据滤波、特征点识别、病情辅助诊断及信号压缩。针对这几方面国内外学者提出许多分析方法,但还存在着诸多不足,如滤波效果不甚理想、特征点的定位精度有待提高、信号压缩方法有待改进等。另外,在辅助诊断方面大多数学者更倾向于时频分析,而心脏作为一个混沌系统,从非线性动力学的角度分析可以获得更多的信息,在这一方面还有深入研究的空间。作为实现临床心电检测的心电自动分析仪,目前大多体积较大、价格高昂,很难为普通家庭所接受。因此,为推广我国心血管疾病的预防与诊疗工作,研制一套价格低廉的便携式心电检测系统并将其推向市场,是一项十分有意义的工作。基于以上理论研究的需要和现实市场的需求,论文针对心电检测系统开发、心电信号预处理及特征点识别、心电辅助诊断和心电信号压缩四个方面开展研究。心电检测系统开发的研究:针对目前心电图机价格昂贵不易普及的问题,研制了一种便携式心电信号记录仪,与其他同类产品相比,该记录仪功能更加多样、操作更加灵活,有三种操作方式可选,可满足不同场合的要求;采样参数可调,在待机状态下电量消耗极小。此外,论文独创性地开发了基于智能手机的心电数据分析与管理系统,实现了数据的采集、处理、存储和显示,以及用户信息的数据库管理和数据远程传输,该系统简便易用、利于推广。心电信号预处理及特征点识别的研究:根据心电分析的实际需要,设计了基于形态学滤波方法和小波滤波方法的滤波器。通过分析表明,形态学滤波器在滤除基线漂移时有其独到的好处,在处理高频信号时则会产生截断误差;而小波滤波器在处理高频干扰时效果显著,对于基线漂移的处理则会损失心电信号P、T波的能量。而后,对两种典型的R波识别方法差分阈值法和小波模极大值法进行讨论,分析了它们的不足。针对以上问题,论文提出了一种基于经验模式分解的R波识别算法,结合数学形态学、小波变换和Hilbert包络算法,实现了高精度的检测,同时在运算过程中可实现心电信号高效去噪功能,与形态学滤波器和小波滤波器相比,在不损失信号质量的前提下可获得更高的信噪比。心电辅助诊断的研究:验证了非线性动力学方法在心电辅助诊断领域的可行性,通过四种方法重点研究了短时期不同病情的心电信号及其心律变异特性:1)通过计算六种病变信号与正常信号的最大Lyapunov指数,指出心脏运动呈现一种弱混沌状态,对诊断病情有特殊意义,为进一步研究指出方向。2)提出了短时心律变异信号的去趋势波动多重分形半谱分析方法,对正常心律、偶发性心律变异病人和室上性心律变异病人的心律变异信号进行计算,得出多重分形参数α的统计值分布随着病情的变化存在着一定的变化趋势,通过盒状图分析,分类精度可达75%以上。3)对严重室性心律失常病症,分析了其动力学变化特征,提出熵和分形的多参数分析方法,进行了室性心律的实时检测与分类,获得了较高的精度。4)首次将语音信号分析方法中的非线性归正技术引入到心电异常波的诊断当中,通过对比贴近度,对异常波的检测结果令人满意,应用到实际心电工作站中也获得了理想的结果。心电信号压缩的研究:探讨了心电信号压缩的必要性,对有损压缩嵌入式零树编码算法提出了改进方法,与已有算法相比可达到更高的压缩比;对无损压缩,根据心电信号的弱混沌特性,提出了基于混沌预测的无损压缩方法,并通过实验证明了该方法的有效性。更多还原

【Abstract】 Electrocardiogram (ECG for short) is the synthetic reflection of the heart electricity on body surface, analyzing the electrocardiogram signal and then diagnosing heart’s diseases has a very important researching value. At present, there are four portions in the research of the ECG signal analyzing and auxiliary diagnosing, which are data filter, character points recognition, classification and recognition of pathological signals and signal compression. According to these portions, scholars bring forward a lot of analysis methods, there are insufficiencies yet. For example, the result of filter is not so fine, the precision of character point location should be heightened, the methods of the signal compression awaiting improved. In addition, most scholars are inclined to time-frequency analyzing on the auxiliary diagnosing, however, the heart is a chaos system, it may achieve more information with by non-linear dynamics analysis, and it has space to research deeply. The automatic analyzing instruments which achieve the clinical electrocardiogram are always too big and expensive, and it is hard for the common families to be accepted. So, for the sake of extending the diagnosis and treat of early cardiopathy, it is a very significant work to develop a cheap and portable electrocardiogram detecting system and put it in the market. Based on the necessary of theory research and the factual requirement of market, this thesis carries on the study in four main aspects which are development of the electrocardiogram detecting system, the preprocession of ECG signals and the character point recognition, the ECG auxiliary diagnosis and the data compression.Research on the development of the ECG detecting system: aiming at the high price of the electrocardiograph and its hardly popularization, the thesis develops a portable ECG signal recorder. Compared with the other products, this recorder has more functions and is very flexible. There are three operating modes to choice, and it can be used in different situations, the sampling parameter can be adapted, it consumes very low under the idle mode. Basing on the above groundwork, the thesis innovatively achieves the ECG analysis and management system on a smartphone, which realizes the data’s collection, processing, memory and display, even the database management of the users’information and remote transfers.Research on the preprocession of ECG signal and the character point recognition: based on the realistic requirement of the ECG analyzing, the thesis designs wavelet filter and morphological filter. It indicates after much analysis that the morphological filter has its particular advantage in correcting baseline drift, but it will bring truncation error when disposing high frequency signals; the wavelet filter has distinct effect on high frequency restraining, but will lose P and T wave energy when disposing the drift. Then, the thesis discusses about the two typical R wave detecting methods: differentiator threshold method and wavelet modulus maxima method, whose shortages are discussed. Aiming at the questions above, the thesis puts forward an R wave identifying algorithm which basing on empirical mode decomposition, combining the mathematics morphology, wavelet and the Hilbert algorithm. This algorithm can achieve a higher detecting precision and a batter effect for noises wiping in the process. Compared to the wavelet filter and the morphological filter, the new algorithm can obtain much higher SNR but not lost the signal quality.Research on the ECG auxiliary diagnosis: the thesis validates the feasibility using the non-linear dynamics in the ECG auxiliary diagnosis field. Through four methods, the thesis studies chiefly on the different ECG signals and their HRV characteristics of different illness in a short time. 1) by calculating the Lyapunov exponent of six states of illness, proves that the heart movement is a weak chaos which has the special meaning to the diagnose, but needs more reseach.2) puts forward the short-time multi-fractal detrended fluctuation half-spectrum analysis of HRV signals, through the calculating, the multi-fractal parameterαof normal arrhythmia, occurrent illness arrhythmia and supraventricular arrhythmia presents a trend, using box-plot, the classifying precision can achieve more than 75%. 3) for malignant ventricular arrhythmia, analyzes its dynamic character, puts forward the multi-parameter method of entropy and fractal to detect the onset and classify these illness, and gets a higher detecting precision. 4) firstly introduces the dynamic time warping from speech recognition to classify ECG abnormal wave, through the compare of close degree, the result of detecting is satisfying, and when used at practical workstation the effect is excellent.Research on the ECG signals compression: the thesis discusses the necessary of the ECG signals compression, for lossy compression, brings forward an improved zero tree coding algorithm, compared with other algorithm which can achieve higher compression ratio; for lossless compression, according to the weak chaos character of ECG signal, puts forward a based chaos forecasting method, and through the experiment proves the validity of this algorithm.更多还原