基于多传感器的人体生理状态判别技术的研究

发布时间：2018-11-27 11:26

【摘要】：近些年来,随着生物传感器技术和可穿戴技术迅速的发展,越来越多用于监测人的生命健康活动的可穿戴产品正不断进入人们的生活。在可穿戴设备上采集和记录人体生理健康数据,如脉搏、呼吸、体温等,已经变得现实可行。然而,在连续长时间范围内的人体的生理信号数据量过大,不利于用户分析、观察和提取有价值信息。因此,论文研究的目的就是实现一种从连续而大量的人体生理数据中提取出简洁而有效的关于人体生理状态变化信息的判别技术。利用此技术可以将人体的生理状态分为两类,一类是普通状态,即人体处于静息下的状态；另一类是事件状态,即人体经历活动、外力刺激或情绪变化等状态。该方法利用脉搏、呼吸、体温三种信号的各自判别机制对人体在相同时间内的生理状态进行二分类判别,并将分类结果以可视化等级分布图的形式回馈给用户,用户可以根据自身状态等级的高低对相应时间内的生理数据进行选择性关注。论文对呼吸信号和体温信号采用的是设置阈值的方式,通过判断从呼吸波中提取出的呼吸频率和体温传感器采集到的人体体温的温度是否超过正常的阈值范围来将人体的生理状态进行二分类判别。而对于脉搏信号,采用基于离散小波变换的方法来去除掉信号中掺杂的高频噪声和基线漂移,从脉搏波的时域中提取周期和主波高度这两个参数作为支持向量机(SVM)的输入特征向量,通过有监督学习的训练方法来构建二分类模型,从脉搏的角度分类出人的生理状态是处于普通状态还是事件状态。本文通过运动、睡眠、喝酒三组实验,对SVM的分类性能进行了统计分析和评价,验证了SVM对人体生理状态判别具有良好的效果。通过将三种信号的判别结果进行融合显示,利用可视化软件将人体状态等级随时间的分布情况展示给用户,这样提供给用户一个观察自身生理状态变化的整体而简洁的视角。
[Abstract]:In recent years, with the rapid development of biosensor technology and wearable technology, more and more wearable products used to monitor human life and health activities are coming into people's lives. It has become feasible to collect and record physical and health data on wearable devices, such as pulse, respiration, body temperature and so on. However, the amount of physiological signals in a continuous long time range is too large, which is not conducive to user analysis, observation and extraction of valuable information. Therefore, the purpose of this paper is to realize a simple and effective discriminant technique for extracting simple and effective information about the changes of human physiological state from the continuous and large amount of human physiological data. By using this technique, the physiological state of human body can be divided into two categories, one is the normal state, that is, the human body is in the state of resting, the other is the state of events, that is, the state of human body experiencing activity, external force stimulation or emotional change. In this method, the physiological state of human body at the same time is judged by using the discriminant mechanism of pulse, respiration and body temperature, and the classification result is given back to the user in the form of visual grade distribution map. The user can selectively pay attention to the physiological data in the corresponding time according to the level of their own state. In this paper, the method of setting threshold is used for respiratory signal and body temperature signal. By judging whether the respiration frequency extracted from the respiration wave and the temperature of the body temperature collected by the temperature sensor exceed the normal threshold range, the physiological state of the human body can be classified into two categories. For pulse signal, the method based on discrete wavelet transform is used to remove the high frequency noise and baseline drift. Two parameters, period and main wave height, are extracted from the time domain of pulse wave as input eigenvector of support vector machine (SVM), and two classification models are constructed by supervised learning training method. The physiological state of a person is classified from the point of view of pulse whether he is in a normal state or an event state. In this paper, the classification performance of SVM was statistically analyzed and evaluated through three groups of experiments: exercise, sleep and drinking, and it was proved that SVM has a good effect on the identification of physiological state of human body. The results of the three signals are fused and displayed, and the distribution of the human state level with time is displayed to the user by using the visualization software, which provides the user with an overall and concise perspective to observe the changes of his own physiological state.
【学位授予单位】：东北大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP212;TN911.7

【相似文献】