基于小波变换的心电信号分析研究及其FPGA实现

发布时间：2018-05-07 19:43

本文选题：心电信号 + 小波变换　；参考：《吉林大学》2014年硕士论文

【摘要】：心脏病突发性强但病情隐蔽，死亡率一直居高不下，已然成为世界性的重大公共卫生问题。医学中往往通过心电图对心率失常进行分析诊断，因此心电信号自动分析诊断系统的研究开发具有很高的临床价值。由于心电信号是一种极易受环境影响的非线性的微弱信号，而且常夹杂各种噪声干扰，增加了分析诊断的难度，因此提高心电自动分析检测系统的精度和准确性，，是心血管疾病的诊断的前提。本文主要研究了心电信号的去噪和特征波检测两方面的关键技术，利用FPGA器件处理信号时所具有的实时性、可靠性、高效性、可并行处理等优势，加快了运算处理的速度，并在Matlab软件分析的基础上设计了基于小波变换的从心电信号预处理到R波检测的一整套检测系统。预处理算法：本文基于db4提升小波变换提出了一种面向FPGA硬件的去噪方案，首先选择了合适的小波基和分解层数，对提升小波变换后各层上的小波系数采用软阈值的方法进行处理，进一步基于Cyclone II系列EP2C35F672C8核心芯片，利用MATLAB/DSP Builder、Quartus II等软件以及VHDL语言设计了该算法的SOPC硬件电路，对系统中各个模块进行功能仿真，并面向FPGA硬件实现。验证结果表明本算法能够在保证原信号不失真的情况下，较好地去除基线漂移、工频干扰、肌电干扰等噪声。最后，通过对系统的实现时间和功耗进行分析得出本文提出的系统具实时性和低功耗性。特征波检测算法：本检测算法是在选择双正交样条小波作为小波基的基础上，根据Mallat算法和小波变换的奇异点检测原理，以及心电信号在奇异点处的Lipschitz指数说明奇异点与模极值对的关系，进而定位心电信号的R波，并提出了避免误检和漏检的策略，有效地提高了检测准确率。其次在小波变换的基础上设计了面向硬件FPGA实现的心电信号R波检测系统。检测系统的硬件实现主要分两步完成：首先对去噪后的心电信号进行小波变换，其实对小波变换的输出结果进行检测。本系统采用流水线操作实现4级的小波变换，即采用级联的方式连接每一层的小波变换，提高了数据运算效率，其中各层小波变换可作为基本运算单元。把小波变换第三、四尺度上的输出结果进行检测，根据检测策略在这两个尺度上寻找模极值对的过零点，即可定位R波。此外，本文还利用48组心电数据对该检测算法进行波形仿真验证，实验结果表明本文提出的R波检测算法具有较高的检测率和稳定性，有一定的实用价值。
[Abstract]:Heart disease has become a major public health problem in the world. Electrocardiogram (ECG) is often used to analyze and diagnose arrhythmias in medicine, so the research and development of ECG automatic analysis and diagnosis system has high clinical value. The ECG signal is a kind of nonlinear weak signal easily affected by the environment, and is often mixed with various kinds of noise interference, which makes the analysis and diagnosis more difficult, so the accuracy and accuracy of the ECG automatic analysis and detection system are improved. Is a prerequisite for the diagnosis of cardiovascular disease. In this paper, the key technologies of ECG signal denoising and characteristic wave detection are studied. The advantages of real-time, reliability, high efficiency and parallel processing of FPGA devices are used to accelerate the processing speed. Based on the analysis of Matlab software, a set of detection system from ECG signal preprocessing to R wave detection based on wavelet transform is designed. Preprocessing algorithm: based on db4 lifting wavelet transform, this paper proposes a denoising scheme for FPGA hardware. Firstly, the appropriate wavelet basis and decomposition layer number are selected. The wavelet coefficients on each layer after lifting wavelet transform are processed by the method of soft threshold. Further, based on Cyclone II series EP2C35F672C8 core chips, the SOPC hardware circuit of the algorithm is designed by using MATLAB/DSP Builder Quartus II and VHDL language. The function of each module in the system is simulated, and the FPGA hardware is implemented. The results show that the proposed algorithm can remove baseline drift, power frequency interference and electromyography interference without distortion of the original signal. Finally, by analyzing the implementation time and power consumption of the system, it is concluded that the proposed system is real-time and low power consumption. Characteristic wave detection algorithm: based on the selection of biorthogonal spline wavelet as wavelet basis, the detection algorithm is based on Mallat algorithm and singular point detection principle of wavelet transform. The Lipschitz exponents of ECG signals at singularity points illustrate the relationship between singularity points and mode-extremum pairs, and then locate R waves of ECG signals. The strategies of avoiding false detection and missing detection are put forward, which can effectively improve the accuracy of detection. Secondly, the R wave detection system of ECG signal based on wavelet transform is designed for hardware FPGA. The hardware implementation of the detection system is mainly divided into two steps: firstly, wavelet transform is applied to the de-noised ECG signal, but in fact, the output result of the wavelet transform is detected. The system adopts pipeline operation to realize the four-level wavelet transform, that is, the wavelet transform of each layer is connected in cascade mode, which improves the efficiency of data operation, and the wavelet transform of each layer can be used as the basic operation unit. The output results of the third and fourth scales of wavelet transform are detected. According to the detection strategy, the zero-crossing points of the mode-extremum pair are found on these two scales, and the R-wave can be located. In addition, 48 sets of ECG data are used to verify the waveform of the algorithm. The experimental results show that the proposed R-wave detection algorithm has a higher detection rate and stability, and has a certain practical value.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN911.6

【参考文献】