脑电信号的多尺度熵分析方法研究

发布时间：2018-01-18 00:04

本文关键词：脑电信号的多尺度熵分析方法研究　出处：《太原理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着脑科学的不断发展和进步,脑电信号已经成为临床和研究人员分析诊断脑疾病最常用到的手段之一。由于脑电信号非平稳性和非线性的特点,脑电信号的分析方法基本还处于研究阶段,临床上仍采用人工判别的方法进行脑疾病诊断。目前,针对脑电信号的非线性分析方法大部分在单个尺度提取特征,而忽略了脑电信号的多尺度特性。因此,研究分析脑电信号的多尺度熵分析方法具有一定的理论价值。本文通过分析传统多尺度熵方法的特点,引入方差和标准差两个新指标重构序列,并提出基于非均匀时间窗的多尺度熵方法。文中重点对多尺度化过程中所选取指标和时间窗进行了研究,对比分析了三种指标在多尺度熵方法中的应用,并根据传统多尺度熵方法中使用均匀时间窗的特点引入一种新的时间窗划分方法,并在癫痫数据、酗酒数据和精分数据上验证了新方法的有效性和可能性。研究主要工作如下:(1)引入方差和标准差两个指标多尺度化时间序列。分析了传统多尺度熵方法中使用平均值指标导致信息丢失的问题,引入方差和标准差两个指标对脑电信号进行分析。通过在三个数据集中的验证,发现使用新指标可以得到比平均值更好的结果。(2)研究样本熵和模糊熵在多尺度熵方法中的性能。在特征提取过程中,选择样本熵和模糊熵两种特征值进行研究,结果表明,两种特征都可以较好的用于脑电信号的分析,但是模糊熵的性能优于样本熵,样本熵的时间复杂度较模糊熵低。(3)提出非均匀时间窗的多尺度熵方法。针对传统多尺度熵方法中粗粒度方法准确率低、细粒度方法时间复杂度高的问题,根据基于空间趋势非均匀分段的思想,提出使用非均匀时间窗的多尺度熵方法,并结合平均值、方差和标准差三种指标以及样本熵和模糊熵两个特征在三个数据集中对比验证新方法的可行性和有效性,结果表明,非均匀时间窗方法可以得到较好的准确率且具有较低的时间复杂度。本文将方差和标准差两种指标引入传统的均匀时间窗多尺度熵方法中,并提出非均匀时间窗多尺度熵方法,最后在三个数据集上进行了验证。实验结果表明,使用引入的方差和标准差两种指标多尺度化脑电信号可以提取到更有效的特征;非均匀时间窗方法在脑电信号分析上具有较好的准确率。文中所引入方法可以为以后脑疾病诊断提供一定的参考价值。
[Abstract]:With the continuous development and progress of brain science, EEG has become one of the most commonly used methods for clinical and researchers to analyze and diagnose brain diseases, due to the non-stationary and nonlinear characteristics of EEG. The analysis method of EEG signal is still in the research stage, and the artificial discriminant method is still used in the diagnosis of brain disease. At present, most of the nonlinear analysis methods for EEG signal extract features at a single scale. Therefore, it has certain theoretical value to study the multi-scale entropy analysis method of EEG signal. This paper analyzes the characteristics of traditional multi-scale entropy method. Two new indices, variance and standard deviation, are introduced to reconstruct the sequence, and a multi-scale entropy method based on the inhomogeneous time window is proposed. In this paper, the selected indexes and time windows in the process of multi-scaling are studied. The application of three indexes in the multi-scale entropy method is compared and analyzed. According to the characteristics of the traditional multi-scale entropy method using the uniform time window, a new time window partition method is introduced and used in the epileptic data. The effectiveness and possibility of the new method are verified by alcohol abuse data and refined data. The main work of the study is as follows: 1). This paper introduces two indexes, variance and standard deviation, and analyzes the problem of information loss caused by the use of average index in the traditional multi-scale entropy method. Two indexes of variance and standard deviation are introduced to analyze the EEG signal. The results are verified in three data sets. It is found that the performance of sample entropy and fuzzy entropy in multi-scale entropy method can be obtained by using the new index, which is better than the average value. In the process of feature extraction, the performance of sample entropy and fuzzy entropy in multi-scale entropy method is studied. Two eigenvalues of sample entropy and fuzzy entropy are selected to study. The results show that both of them can be used in EEG signal analysis, but the performance of fuzzy entropy is better than sample entropy. The time complexity of sample entropy is lower than that of fuzzy entropy. (3) Multi-scale entropy method based on non-uniform time window is proposed. Aiming at the problem of low accuracy of coarse-grained method and high time complexity of fine-grained method in traditional multi-scale entropy method. According to the idea of non-uniform segmentation based on spatial trend, a multi-scale entropy method using non-uniform time window is proposed and combined with average value. Variance and standard deviation as well as sample entropy and fuzzy entropy are compared in the three data sets to verify the feasibility and effectiveness of the new method. The non-uniform time window method can get better accuracy and lower time complexity. In this paper, the variance and standard deviation are introduced into the traditional uniform time window multi-scale entropy method. A non-uniform time window multi-scale entropy method is proposed and validated on three data sets. The multi-scale EEG signal can be used to extract more effective features by using the introduced variances and standard deviations. The non-uniform time window method has a good accuracy in EEG signal analysis, and the method introduced in this paper can provide a certain reference value for the diagnosis of brain diseases in the future.
【学位授予单位】：太原理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R741.044;TN911.6

【相似文献】