基于双树复小波变换的非平稳时间序列去趋势波动分析方法

多重分形去趋势波动分析是研究非平稳时间序列非均匀性和奇异性的有效工具, 针对该方法中趋势项难以确定的问题, 提出一种基于双树复小波变换的方法, 实现了非平稳信号的多重分形自适应去趋势波动分析. 利用双树复小波变换提取信号的多尺度趋势和波动信息, 通过小波系数的希尔伯特变换确定每个时间尺度不重叠子区间的长度, 使多重分形分析具有信号自适应性及较高的计算效率. 以具有解析形式分形特征的倍增级联信号和分数布朗运动时间序列为例验证本文方法的有效性, 所得结果与解析解相吻合. 与传统的多项式去趋势多重分形方法相比, 本文方法根据信号自身特点自适应地确定信号的趋势和不重叠等长度子区间长度, 所得结果更加精确. 对倍增级联信号时间序列取不同的长度, 验证了算法的稳定性. 分别与基于极大重叠离散小波变换和离散小波变换多重分形方法进行比较, 表明本文方法具有更精确的结果和更快的运算速度.     

一种基于重标极差方法的动力学结构突变检测新方法

提出了一种针对相关性时间序列动力学结构突变检测的新方法——滑动移除重标极差分析.数值试验表明,新方法能够准确的检测出理想序列中的动力学结构突变,其检测结果对于子序列的长度依赖性较小,而且具有很强的抗噪能力,弥补了滑动去趋势波动分析的不足. 关键词： 标度指数 滑动移除重标极差分析 滑动去趋势波动分析 动力学结构突变     

滑动去趋势波动分析与近似熵在动力学结构突变检测中的性能比较

近似熵（ApEn）被认为是一种有效的动力学结构突变检测方法. 将一种新的动力学结构检测方法——滑动去趋势波动分析（MDFA）与ApEn的检测结果进行了比较,检验了新方法的性能. 结果表明,新方法的检测结果几乎不依赖于子序列的长度,而ApEn虽然能在一定程度上识别系统的动力学结构突变,但其检测结果依赖于子序列长度,且不能准确地检测出突变点的位置. 因此,相对于ApEn方法而言,MDFA方法更适合于动力学结构突变检测,其优越性是显而易见的. 关键词： 滑动去趋势波动分析 近似熵 动力学结构突变     

基于符号序列描述的一类分段光滑系统中分岔现象与混沌分析

从拓扑序列出发，提出了描述DC/DC变换器一类分段光滑系统中的分岔现象和混沌行为的符号序列方法，根据最大子序列的性态判别分岔的类型，以及检测边界碰撞分岔的发生.例如,当发生倍周期分岔时，最大子序列保持不变；当发生边界碰撞分岔时，最大子序列发生变化；混沌态则没有最大子序列.研究表明，占空比是表征DC/DC变换器一类分段光滑系统动力学行为的一个最本质的量，“饱和非线性”是引起边界碰撞分岔产生的根本原因. 关键词： 符号序列 分岔 混沌 分段光滑系统     

非线性动力学方法在气候突变检测中的应用

快速、准确的检测气候突变, 对于我们认识气候系统的变化和对未来气候系统演变趋势的预测有着重要的现实意义和社会经济价值. 本文主要回顾了近年来非线性突变检测技术的主要研究进展及其在实际观测资料中的应用, 其中包括基于气候系统长程相关性的检测方法, 如滑动去趋势波动分析方法、滑动移除去趋势波动分析方法、滑动移除重标极差方法和指纹法等; 以及基于时间序列复杂性的检测方法, 如近似熵方法, Fisher信息和小波Fisher信息等. 此外, 本文还指出发展针对空间场的突变检测技术是未来一个可能的发展方向. 由于空间场所包含的气候系统的演变信息远高于单点时间序列, 空间场的突变检测技术将会使得对气候突变的检测时间大大缩短, 从而使得人们有足够的时间去采取行动, 以便为适应气候突变所带来的新挑战做好准备.     

三维成像中的二值时空编码照明方法

提出了一种用于三维成像的二值时空编码照明方法.将投影平面的每一行分成由若干个像素组成的区间.利用区间内像素的空间坐标和时间坐标对区间进行编码.在测量时,通过对拍摄的图像序列分析,恢复区间的编码.在得到所有区域编码后,将相邻的给定个数的区域编码组成代码子序列.然后在设计的代码序列中进行代码子序列匹配,得到场景表面、摄像机像面及投影平面三者之间的对应点.最后,采用三角测量原理得到被测物体的面形.实验结果表明,这种方法在得到高密度距离像的同时可以得到物体的纹理,而不需要额外拍摄图像.测量结果有较高的精度和稳健性.     

松质骨超声背散射信号的谱信息熵判断方法

在松质骨超声背散射评价的实际应用中,如何快速准确地判断接收到的信号中是否包含有效背散射信号是一个重要问题。提出一种基于谱信息熵的背散射信号判断方法。对在体采集的984例成人跟骨处临床数据,采用该方法判断背散射信号的有效性,将判断结果与经验判断结果进行对照分析,并分析该方法中的信号区间长度和谱信息熵分段数对判断结果的影响。结果表明,当信号区间长度为13μs,谱信息熵分段数为15~20时,该方法可以获得最佳的判断结果(准确度〉95%,灵敏度〉99%,特异度〉87%),并且计算时间极短(1.5 ms)。因此采用谱信息熵方法判断背散射信号的有效性,可以满足超声骨质评价中对准确性和实时性的要求。      

不同滤波方法在去趋势波动分析中去噪的应用比较

研究了连续噪声和尖峰噪声对去趋势波动分析的影响,发现噪声的存在使得双对数曲线在尺度较小时发生了"转折"现象.针对这一问题,文中采用三种不同滤波方法对理想时间序列进行了实验,结果表明,多级Vondrak滤波得到的高频序列与真实噪声序列无论是在强度还是在演变趋势上都展现出惊人的一致性,低频滤波序列的去趋势波动分析结果与真实信号十分接近,多级Vondrak滤波基本上能够消除由于噪声所引起的"转折"现象,而且这一研究结果对于滤波周期阈值的依赖性并不太大.多点滑动加权平均滤波虽然能够在一定程度上减轻噪声对于去趋势波动的影响,但不能从根本上消除由于噪声所引起的"转折"现象.快速傅里叶滤波在选择合适的滤波周期阈值时,能够基本消除噪声对去趋势波动分析的影响,但是由于其滤波结果对于滤波周期阈值的依赖较大,在实际应用中滤波周期阈值的选取比较困难.因此,多级Vondrak滤波是消除噪声对去趋势波动分析结果影响的一种有效的途径. 关键词： 多级Vondrak滤波 去趋势波动分析 多点滑动加权平均滤波 快速傅里叶滤波     

基于复杂度分析logistic映射和Lorenz模型的研究

采用三次粗粒化方法得到了logistic映射和Lorenz模型的符号序列，运用动态非线性时间序 列分析方法——Lemper-Ziv复杂度，分别对两组符号序列进行了对比分析.对于logistic映 射，其复杂度反映了时间序列的演化；Lorenz模型三个分量的复杂度序列都具有混沌性质， 即由许多振幅非常接近而长度完全不同的循环所组成，反映了Lorenz模型内在的准周期特性 .进一步研究发现，当取不同的窗口长度时，复杂度序列的特征基本相同，并且复杂度反映 了时间序列的时空特性.因此，可以借助复杂度的计算来反演观测资料的动力学结构. 关键词： 三次粗粒化 Lemper-Ziv复杂度 logistic映射 Lorenz模型     

利用复杂网络研究中国温度序列的拓扑性质

依据粗粒化方法,将中国1961—2002年逐日平均温度序列转化为由5个特征字符{R,r,e,d,D}构成的温度符号序列.以符号序列中的125种3字串组成的温度波动模态为网络的节点(即连续4d的温度波动组合),并按照时间顺序连边,构建有向加权的温度波动网络,进而将温度波动模态间的相互作用等综合信息蕴含于网络的拓扑结构之中.对随机序列和Lorenz系统的混沌序列分别构建随机和混沌波动网络.计算三种网络的度与度分布、聚类系数、最短路径长度等动 关键词： 气候变化 气候复杂网络 拓扑结构     

Interpretation of heart rate variability via detrended fluctuation analysis and alphabeta filter

Detrended fluctuation analysis (DFA), suitable for the analysis of nonstationary time series, has confirmed the existence of persistent long-range correlations in healthy heart rate variability data. In this paper, we present the incorporation of the alphabeta filter to DFA to determine patterns in the power-law behavior that can be found in these correlations. Well-known simulated scenarios and real data involving normal and pathological circumstances were used to evaluate this process. The results presented here suggest the existence of evolving patterns, not always following a uniform power-law behavior, that cannot be described by scaling exponents estimated using a linear procedure over two predefined ranges. Instead, the power law is observed to have a continuous variation with segment length. We also show that the study of these patterns, avoiding initial assumptions about the nature of the data, may confer advantages to DFA by revealing more clearly abnormal physiological conditions detected in congestive heart failure patients related to the existence of dominant characteristic scales.     

Two general models that generate long range correlation

In this paper we study two models that generate sequences with LRC (long range correlation). For the IFT (inverse Fourier transform) model, our conclusion is the low frequency part leads to LRC, while the high frequency part tends to eliminate it. Therefore, a typical method to generate a sequence with LRC is multiplying the spectrum of a white noise sequence by a decaying function. A special case is analyzed: the linear combination of a smooth curve and a white noise sequence, in which the DFA plot consists of two line segments. For the patch model, our conclusion is long subsequences leads to LRC, while short subsequences tend to eliminate it. Therefore, we can generate a sequence with LRC by using a fat-tailed PDF (probability distribution function) of the length of the subsequences. A special case is also analyzed: if a patch model with long subsequences is mixed with a white noise sequence, the DFA plot will consist of two line segments. We have checked known models and actual data, and found they are all consistent with this study.     

基于离散萤火虫算法的近红外波长优选方法研究

近红外光谱数据量大,需要进行压缩,以降低建立光谱校正模型的计算复杂度,提高模型精度和稳健性。为此,提出了一种基于离散萤火虫算法(discrete firefly algorithm)的近红外光谱波长变量筛选方法。首先采用蒙特卡罗方法剔除异常值,并应用Kennard-Stone法进行校正样本的选择。对通用萤火虫算法进行离散化处理,改进了吸引度的自适应公式,在移动公式中增加了牵引权重,以适应离散化处理的影响和优化算法,并在离散萤火虫算法中加入精英保留策略,加快算法的收敛速度。实验中找到DFA算法中的各项参数中的最佳值。通过离散萤火虫算法优选波长变量,建立发酵液中丁二酸含量的近红外光谱偏最小二乘回归(partial least squares regression)校正模型。与标准遗传算法(genetic algorithm)优选波长方法进行了比较。结果显示,基于离散萤火虫算法的波长优选方法所建立的PLS校正模型,其校正集的相关系数(R2_c)为0.986,RMSEC为0.409,预测集的相关系数(R2_p)为0.969,RMSEP为0.458,模型稳健性和精度都要优于全光谱建模以及遗传算法波长优选方法。显示了DFA在近红外光谱数据筛选方面的优越性。     

Relationship between Continuum of Hurst Exponents of Noise-like Time Series and the Cantor Set

In this paper, we have modified the Detrended Fluctuation Analysis (DFA) using the ternary Cantor set. We propose a modification of the DFA algorithm, Cantor DFA (CDFA), which uses the Cantor set theory of base 3 as a scale for segment sizes in the DFA algorithm. An investigation of the phenomena generated from the proof using real-world time series based on the theory of the Cantor set is also conducted. This new approach helps reduce the overestimation problem of the Hurst exponent of DFA by comparing it with its inverse relationship with α of the Truncated Lévy Flight (TLF). CDFA is also able to correctly predict the memory behavior of time series.     

Detrending fluctuation analysis based on high-pass filtering

Detrended fluctuation analysis (DFA) is a scaling method commonly used for detecting long-range correlations in non-stationary time series. The DFA method uses a trend based on polynomial fitting to extract and quantify fluctuations at different time scales. Basically, such procedure acts as a (non-dynamical) high-pass filter that removes time series components below a given time scale. As an alternative to the polynomial fitting approach, this paper proposes a DFA method based on well-known high-pass filters (e.g., Butterworth, elliptic, etc.). Numerical results show that the proposed DFA approach yields results similar to traditional DFA method. Maybe, the main advantage of the proposed DFA method is that efficient implementations of high-pass filters are available commercially.     

Sparse Long Blocks and the Micro-structure of the Longuest Common Subsequences

Consider two random strings having the same length and generated by an iid sequence taking its values uniformly in a fixed finite alphabet. Artificially place a long constant block into one of the strings, where a constant block is a contiguous substring consisting only of one type of symbol. The long block replaces a segment of equal size and its length is smaller than the length of the strings, but larger than its square-root. We show that for sufficiently long strings the optimal alignment (OA) corresponding to a longest common subsequence (LCS) treats the inserted block very differently depending on the size of the alphabet. For two-letter alphabets, the long constant block gets mainly aligned with the same symbol from the other string, while for three or more letters the opposite is true and the block gets mainly aligned with gaps. We further provide simulation results on the proportion of gaps in blocks of various lengths. In our simulations, the blocks are “regular blocks” in an iid sequence, and are not artificially inserted. Nonetheless, we observe for these natural blocks a phenomenon similar to the one shown in case of artificially-inserted blocks: with two letters, the long blocks get aligned with a smaller proportion of gaps; for three or more letters, the opposite is true. It thus appears that the microscopic nature of two-letter OAs and three-letter OAs are entirely different from each other.     

Comparison of detrending methods for fluctuation analysis

We examine several recently suggested methods for the detection of long-range correlations in data series based on similar ideas as the well-established Detrended Fluctuation Analysis (DFA). In particular, we present a detailed comparison between the regular DFA and two recently suggested methods: the Centered Moving Average (CMA) Method and a Modified Detrended Fluctuation Analysis (MDFA). We find that CMA performs the same as DFA in long data with weak trends and is slightly superior to DFA in short data with weak trends. When comparing standard DFA to MDFA we observe that DFA performs slightly better in almost all examples we studied. We also discuss how several types of trends affect different types of DFA. For weak trends in the data, the new methods are comparable with DFA in these respects. However, if the functional form of the trend in data is not a-priori known, DFA remains the method of choice. Only a comparison of DFA results, using different detrending polynomials, yields full recognition of the trends. A comparison with independent methods is recommended for proving long-range correlations.     

Non-linear characteristics and long-range correlations in Asian stock markets

We test several non-linear characteristics of Asian stock markets, which indicates the failure of efficient market hypothesis and shows the essence of fractal of the financial markets. In addition, by using the method of detrended fluctuation analysis (DFA) to investigate the long range correlation of the volatility in the stock markets, we find that the crossover phenomena exist in the results of DFA. Further, in the region of small volatility, the scaling behavior is more complicated; in the region of large volatility, the scaling exponent is close to 0.5, which suggests the market is more efficient. All these results may indicate the possibility of characteristic multifractal scaling behaviors of the financial markets.     

Scaling characteristics of ocean wave height time series

Fluctuations in the significant wave height can be quantified by using scaling statistics. In this paper, the scaling properties of the significant wave height were explored by using a large data set of hourly series from 25 monitoring stations located off the west coast of the US. Detrended fluctuation analysis (DFA) was used to investigate the scaling properties of the series. DFA is a robust technique that can be used to detect long-range correlations in nonstationary time series. The significant wave height data was analyzed by using scales from hourly to monthly. It was found that a common scaling behavior can be observed for all stations. A breakpoint in the scaling region around 4-5 days was apparent. Spectral analysis confirms this result. This breakpoint divided the scaling region into two distinct parts. The first part was for finer scales (up to 4 days) which exhibited Brown noise characteristics, while the second one showed 1/f noise behavior at coarser scales (5 days to 1 month). The first order and the second order DFA (DFA1 and DFA2) were used to check the effect of seasonality. It was found that there were no differences between DFA1 and DFA2 results, indicating that there is no effect of trends in the wave height time series. The resulting scaling coefficients range from 0.696 to 0.890 indicating that the wave height exhibits long-term persistence. There were no coherent spatial variations in the scaling coefficients.