基于角速度信号的呼吸参数提取研究及应用
发布时间：2019-07-06 17:31
【摘要】：呼吸生理参数是人们了解自身状况的重要参考指标,随着健康意识的深入人心,人们更期望长期的呼吸检测,这对呼吸信号的采集装置提出更高的要求。惯性传感器作为一种新型技术,可用于智能穿戴、人体传感网络和呼吸检测等领域,近年来得到学术界广泛的研究和探索。但目前,还未出现基于单通道的呼吸类型分类的研究。惯性传感器包括加速度计和角速度计(陀螺仪),基于角速度计在获取呼吸信号质量上的优势,本课题选用角速度计获取的呼吸信号,即呼吸角速度信号做研究。通过搭建平台,对从单通道呼吸角速度信号提取呼吸参数的可行性进行了分析,并研究了单通道呼吸角速度信号在提取呼吸相位和呼吸信号分类上的应用。研究的主要工作可分为分析和应用两个部分。在分析方面,本文主要分了两个部分:信号采集和呼吸参数的提取分析。其中,在信号的采集方面使用单个惯性传感器放置在胸骨上切迹以获取呼吸角速度信号,单通道设备方便了呼吸信号的采集,胸骨上切迹位置保证了单通道呼吸数据的鲁棒性。呼吸参数的提取分析部分则选用了呼吸频率和呼吸相位作为参数进行分析,选用呼吸二氧化碳浓度信号作为参考呼吸信号进行参数对比,结果表明呼吸频率位于置信区间内且相移中位误差低于0.5秒。在应用方面,首先用呼吸角速度信号提取了呼吸相位,通过上位机界面软件系统的设计将呼吸角速度转换为易于识别的呼吸角度信号从而方便呼吸相位的提取,此外还介绍了肺功能康复治疗仪的设计,该仪器是呼吸相位的应用点之一,也是本论文的工作量之一。其次对呼吸信号进行了分类,选用了七类常见的呼吸异常信号和正常呼吸信号,基于前人的研究基础,使用支持向量机技术设计分类器进行模式识别。在分类器的设计过程中结合了多种技术来获取较优的特征值,特征值包括:均值、方差、能量、过阈值呼吸数和符号聚集近似值,为了提高特征值的有效性使用了小波技术和窗口分割的技术。使用十折交叉验证的方式获得的分类准确率最高达到91.25%,验证了单通道呼吸角速度信号在分类呼吸类型应用上的可行性。综上,本课题的研究结果表明:惯性传感器采集的单通道呼吸角速度信号能够代替传统的呼吸检测仪器获取呼吸频率,并且能获取较准确的呼吸相位;该信号能够用于提取呼吸相位并在常见呼吸类型的分类上有较好的分类效果。本文的工作提出了呼吸检测的新方式,为长期呼吸检测提供了可靠的思路,可应用于呼吸相关疾病的早期预警。
[Abstract]:Respiratory physiological parameters are an important reference index for people to understand their own conditions. with the deepening of health awareness, people expect long-term respiratory detection, which puts forward higher requirements for respiratory signal acquisition devices. As a new technology, inertial sensor can be used in intelligent wear, human body sensor network and respiratory detection. In recent years, inertial sensor has been widely studied and explored in academic circles. However, at present, there is no research on the classification of respiratory types based on single channel. Inertial sensors include accelerometer and angular velocimeter (gyroscope). Based on the advantages of angular velocimeter in obtaining respiratory signal quality, the respiratory signal obtained by angular velocimeter, that is, respiratory angular velocity signal, is selected for research in this paper. By building a platform, the feasibility of extracting respiratory parameters from single channel respiratory angular velocity signals is analyzed, and the application of single channel respiratory angular velocity signals in extracting respiratory phase and classification of respiratory signals is studied. The main work of the study can be divided into two parts: analysis and application. In the aspect of analysis, this paper is divided into two parts: signal acquisition and respiratory parameter extraction and analysis. In the aspect of signal acquisition, a single inertial sensor is used to place a notch on the sternum to obtain the respiratory angular velocity signal. The single-channel equipment facilitates the acquisition of respiratory signal, and the location of the notch on the sternum ensures the robustness of the single-channel respiratory data. In the part of respiratory parameter extraction and analysis, respiratory frequency and respiratory phase were selected as parameters, and respiratory carbon dioxide concentration signal was used as reference respiratory signal for parameter comparison. The results showed that respiratory frequency was in confidence interval and phase shift median error was less than 0.5 seconds. In the aspect of application, the respiratory phase is extracted by respiratory angular velocity signal, and the respiratory angular velocity is converted into easily identifiable respiratory angle signal through the design of upper computer interface software system, so as to facilitate the extraction of respiratory phase. in addition, the design of pulmonary function rehabilitation therapy instrument is also introduced, which is one of the application points of respiratory phase and one of the workload of this paper. Secondly, the respiratory signals are classified, and seven kinds of respiratory abnormal signals and normal respiratory signals are selected. Based on the previous research basis, support vector machine (SVM) technology is used to design classifiers for pattern recognition. In the design process of the classifier, a variety of techniques are combined to obtain the better eigenvalues, including mean, variance, energy, over-threshold respiration and symbolic aggregation approximations. in order to improve the effectiveness of eigenvalues, wavelet technique and window segmentation technique are used. The highest classification accuracy is 91.5% by using ten fold cross verification, which verifies the feasibility of the application of single channel respiratory angular velocity signal in the classification of respiratory types. In summary, the results of this paper show that the single channel respiratory angular velocity signal collected by inertial sensor can replace the traditional respiratory detection instrument to obtain respiratory frequency, and can obtain more accurate respiratory phase, and the signal can be used to extract respiratory phase and has a good classification effect on the classification of common respiratory types. In this paper, a new method of respiratory detection is proposed, which provides a reliable idea for long-term respiratory detection and can be used for early warning of respiratory related diseases.
【学位授予单位】：深圳大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN911.7;R443.6
本文编号：2511207
[Abstract]:Respiratory physiological parameters are an important reference index for people to understand their own conditions. with the deepening of health awareness, people expect long-term respiratory detection, which puts forward higher requirements for respiratory signal acquisition devices. As a new technology, inertial sensor can be used in intelligent wear, human body sensor network and respiratory detection. In recent years, inertial sensor has been widely studied and explored in academic circles. However, at present, there is no research on the classification of respiratory types based on single channel. Inertial sensors include accelerometer and angular velocimeter (gyroscope). Based on the advantages of angular velocimeter in obtaining respiratory signal quality, the respiratory signal obtained by angular velocimeter, that is, respiratory angular velocity signal, is selected for research in this paper. By building a platform, the feasibility of extracting respiratory parameters from single channel respiratory angular velocity signals is analyzed, and the application of single channel respiratory angular velocity signals in extracting respiratory phase and classification of respiratory signals is studied. The main work of the study can be divided into two parts: analysis and application. In the aspect of analysis, this paper is divided into two parts: signal acquisition and respiratory parameter extraction and analysis. In the aspect of signal acquisition, a single inertial sensor is used to place a notch on the sternum to obtain the respiratory angular velocity signal. The single-channel equipment facilitates the acquisition of respiratory signal, and the location of the notch on the sternum ensures the robustness of the single-channel respiratory data. In the part of respiratory parameter extraction and analysis, respiratory frequency and respiratory phase were selected as parameters, and respiratory carbon dioxide concentration signal was used as reference respiratory signal for parameter comparison. The results showed that respiratory frequency was in confidence interval and phase shift median error was less than 0.5 seconds. In the aspect of application, the respiratory phase is extracted by respiratory angular velocity signal, and the respiratory angular velocity is converted into easily identifiable respiratory angle signal through the design of upper computer interface software system, so as to facilitate the extraction of respiratory phase. in addition, the design of pulmonary function rehabilitation therapy instrument is also introduced, which is one of the application points of respiratory phase and one of the workload of this paper. Secondly, the respiratory signals are classified, and seven kinds of respiratory abnormal signals and normal respiratory signals are selected. Based on the previous research basis, support vector machine (SVM) technology is used to design classifiers for pattern recognition. In the design process of the classifier, a variety of techniques are combined to obtain the better eigenvalues, including mean, variance, energy, over-threshold respiration and symbolic aggregation approximations. in order to improve the effectiveness of eigenvalues, wavelet technique and window segmentation technique are used. The highest classification accuracy is 91.5% by using ten fold cross verification, which verifies the feasibility of the application of single channel respiratory angular velocity signal in the classification of respiratory types. In summary, the results of this paper show that the single channel respiratory angular velocity signal collected by inertial sensor can replace the traditional respiratory detection instrument to obtain respiratory frequency, and can obtain more accurate respiratory phase, and the signal can be used to extract respiratory phase and has a good classification effect on the classification of common respiratory types. In this paper, a new method of respiratory detection is proposed, which provides a reliable idea for long-term respiratory detection and can be used for early warning of respiratory related diseases.
【学位授予单位】：深圳大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN911.7;R443.6
【相似文献】
相关期刊论文 前2条
1 金立群,苏龙;蚊蛹形态学研究及其分类[J];白求恩医科大学学报;1990年02期
2 ;[J];;年期
相关硕士学位论文 前1条
1 杨环宇;基于角速度信号的呼吸参数提取研究及应用[D];深圳大学;2017年
，本文编号：2511207
本文链接：https://www.wllwen.com/linchuangyixuelunwen/2511207.html
最近更新
教材专著
