【作者】 杨凤霞；

【导师】 张爱华；

【作者基本信息】 兰州理工大学 ， 检测技术与自动化装置， 2006， 硕士

【摘要】 亚健康是指人的机体虽然没有明确的疾病,但呈现出活力降低,适应力呈不同程度减退的一种生理状态,是介于健康与疾病之间的一种生理功能降低的状态。它既可以向健康状态转化,又可以向坏的方向转化而进一步发展为各种疾病。亚健康已经成为当今危害人类健康的头号隐形杀手。由于没有器质性病变,通常不伴有明显的病理表现,现有的传统医疗检测设备,无法对机体的亚健康状态和导致功能低下的原因做出描述和判断。目前,亚健康状态的诊断主要靠一些生化指标和问卷调查,至今为止还未能提出同时具备无创、简便、快速、成本低、可重复采用、客观量化、敏感性和特异性高、能持续监测亚健康对个体的影响过程的测评方法。疲劳是导致亚健康的主要原因。本文从疲劳的角度入手,从理论上论证了基于脉搏信号分析对亚健康评估的可行性,并提出了一种通过脉搏信号分析来识别亚健康状态的方法。 中医学认为亚健康往往是人体阴阳失衡、脏腑功能失调的初始状态。脉搏信号中蕴涵着丰富的人体生理病理信息,是传递和窥视体内功能变化的窗口。因此,可以通过对脉搏信号的分析来对亚健康状态进行评估。 本文在综述国内外研究现状的基础上,系统地分析了中医脉象研究的工作流程和工作方法。用HK—2000C数字脉搏传感器构建了脉搏信号采集系统,设计了数据采集实验方案,采集了60余名大学生志愿者的脉搏信号,并从中选择处于中度以上亚健康状态的数据17组,健康状态数据13组进行分析。然后采用具有良好的时频局部化及自适应特性的小波分析法,研究了人体脉搏信号的去噪问题,取得了较好的效果。并且根据脉搏信号的产生机理、性质,提取了功率谱峰值、功率谱重心及其频率,AR模型系数,SER值,Renyi信息量等多个特征量。经对30例样本的分析识别检验,结果表明:功率谱峰值和峰值频率作为特征量能取得比较好的识别效果,识别正确率达到了86.7%,对17例亚健康状态仅有2例未能正确识别。功率谱重心和重心频率作为特征量也取得了80%的识别率,对AR模型系数的分类正确率也达到了76.7%,结果令人满意。Renyi信息量被引入到脉搏信号识别当中,取得了73.3%的识别率。SER值是比较常用的一种脉搏信号的频域特征,从分类效果来看SER值并不能很好的识别亚健康状态。文中还就线性判别式分析(Linear Discriminant Analysis,LDA)和支持向量机(support Vector Machine,SVM)的识别结果进行了比较。应用LDA取得了较好的识别结果。应用SVM也取得了一定的效果,但在其核函数选择以及参数调整方面均需要参考经验值而确定,在实际应用中推广性较差。最后介绍了亚健康状态识别系统的软件设计与实现。更多还原

【Abstract】 Sub-health, also called "the third state", is defined as a critical state between the health and diseases. It is a kind of physiological state that appears with vigor reducing, adaptive capacity failing in various degrees, although there isn’t any identified disease in organism. Sub-health could be transformed to the health if the state is dealt with aptly, and, contrariwise, to illness. Now, sub-health has been badly endangering the health of the residents. However, the state can’t be diagnosed by traditional medical equipment because all necessary physical and clinical indexes are tested negative. At present, the diagnosis of sub-health is lacking of inexpensive, handy, objective and quantitative method that can evaluate sub-health state effectively. Fatigue was the primary cause of sub-health. This dissertation theoretically demonstrates the feasibility to evaluate people’s sub-healthy state by quantifying fatigue, which is based on pulse analysis. And according to the needing of the scientific research, a new method for identifying the sub-health state from pulse wave is presented in this paper.According to the Chinese medicine understanding, sub-health is the initial state which the humors (qi, moisture, blood) are affected, and organ systems suffer from dysfunction. Human pulse contains a lot of useful information about what goes on inside the body. So sub-health could be evaluated by analyzing pulse wave.Based on the study status in quo, working method and working process of the pulse recognition has been introduced in this dissertation. Firstly, pulse collection system was built by using HK-2000C digital integrated pulse transducer, and data collection project was designed. Pulse waves of more than 60 voluntary undergraduates were measured from their left wrist. Then, 17 sub-healthy data and 13 healthy data were chosen for analysis. Secondly, wavelet analysis which has a good qualities both in time domain and frequency domain and is an ideal tool in analyzing unsteady signal, is used in pretreatment of human pulse and good result has been obtained. In addition, considering the origin, the mechanism and the framework of the pulse signals, peak value, peak frequency, center of gravity (cg), gravity frequency of power spectrum, AR model parameter, the value of SER and Renyi entropy were extracted. Moreover, we successfully use Linear Discriminant Analysis (LDA) to identify sub-health status from the pulse waves of 17 sub-healthy persons and 13 healthy persons. The recognition accuracy is up to 86.667% by using peak value and peak frequency of power spectrum as characteristics. Only two sub-healthy personsare misjudged. The recognition accuracy of 80% was attained by using eg and gravity frequency of power spectrum as characteristics. The accuracy of 76.7% by using AR model parameter and of 73.3% by using Renyi entropy was obtained yet. SER value, which is often used in pulse recognition, can’t identify the sub-health state effectively. Then, LDA and Support Vector Machine (SVM) were compared. The preferable result has been obtained by using LDA. Using SVM also gained good result in recognition, but there still are some problems in selection of kernel parameters that usually select by experience. Finally, the software design of sub-health recognition system has been introduced.更多还原