《電子技術(shù)應(yīng)用》
您所在的位置:首頁 > 模擬設(shè)計(jì) > 設(shè)計(jì)應(yīng)用 > 經(jīng)驗(yàn)?zāi)B(tài)分解及其模態(tài)混疊消除的研究進(jìn)展
經(jīng)驗(yàn)?zāi)B(tài)分解及其模態(tài)混疊消除的研究進(jìn)展
2019年電子技術(shù)應(yīng)用第3期
戴 婷1,,張榆鋒1,,章克信2,何冰冰1,,朱泓萱1,張俊華1
1.云南大學(xué) 信息學(xué)院電子工程系,,云南 昆明650091,;2.昆明醫(yī)科大學(xué)第二附屬醫(yī)院,云南 昆明650031
摘要: 由Huang提出的經(jīng)驗(yàn)?zāi)B(tài)分解(Empirical Mode Decomposition,,EMD)算法是一種數(shù)據(jù)驅(qū)動(dòng)的自適應(yīng)非線性時(shí)變信號(hào)分析方法,,可以把數(shù)據(jù)分解成具有物理意義的少數(shù)幾個(gè)固有模態(tài)函數(shù)(Intrinsic Mode Function,IMF)分量,。然而模態(tài)混疊會(huì)導(dǎo)致錯(cuò)假的時(shí)頻分布,,使IMF失去物理意義,嚴(yán)重影響了EMD分解的準(zhǔn)確性與實(shí)用性,。分別針對(duì)一維和多維EMD抑制模態(tài)混疊,,總結(jié)歸納了相關(guān)研究取得的主要成果,指出了各方法抑制效果的改進(jìn)及仍有的不足,。最后討論了相關(guān)研究及應(yīng)用未來的發(fā)展趨勢(shì),。
中圖分類號(hào): TN911.7
文獻(xiàn)標(biāo)識(shí)碼: A
DOI:10.16157/j.issn.0258-7998.182560
中文引用格式: 戴婷,張榆鋒,,章克信,,等. 經(jīng)驗(yàn)?zāi)B(tài)分解及其模態(tài)混疊消除的研究進(jìn)展[J].電子技術(shù)應(yīng)用,2019,,45(3):7-12.
英文引用格式: Dai Ting,,Zhang Yufeng,Zhang Kexin,,et al. The research progress of empirical mode decomposition and mode mixing elimination[J]. Application of Electronic Technique,,2019,45(3):7-12.
The research progress of empirical mode decomposition and mode mixing elimination
Dai Ting1,,Zhang Yufeng1,,Zhang Kexin2,He Bingbing1,,Zhu Hongxuan1,,Zhang Junhua1
1.Department of Electronic Engineering,Information School,,Yunnan University,,Kunming 650091,China; 2.The Second Affiliated Hospital of Kunming Medical University,,Kunming 650031,,China
Abstract: The Empirical Mode Decomposition(EMD) algorithm proposed by Huang is a data driven adaptive analysis method for nonlinear time-varying signals. The signals can be decomposed into a few Intrinsic Mode Functions(IMFs) components with physical meaning. However, Mode Mixing(MM) can lead to wrong or false components in time frequency distributions, and then cause the decomposed IMFs losing their physical meaning. This seriously affects the EMD accuracies and applications. This study reviews methods of the MM suppression in one-dimensional and multi-dimensional EMD algorithms. The results improvements and limitations in related researches are summarized. Finally, the future development trend of related researches and applications are discussed.
Key words : empirical mode decomposition(EMD);intrinsic mode function(IMF),;mode mixing(MM),;Hilbert transform

0 引言

    傅里葉分析技術(shù)[1]在分析時(shí)變非線性信號(hào)時(shí)存在無法表述信號(hào)的時(shí)頻局部特性的局限性[2]。為了分析處理非平穩(wěn)信號(hào),,人們相繼提出了一系列新的信號(hào)分析方法:短時(shí)傅里葉變換[3],、雙線性時(shí)頻分布[4]、Gabor變換[5],、小波分析[6],、分?jǐn)?shù)階傅里葉變換[7]等。這些算法從不同程度上對(duì)非平穩(wěn)信號(hào)的時(shí)變性給予了恰當(dāng)?shù)拿枋?,改進(jìn)了傅里葉分析的性能[8],。然而,方法仍是全局范疇,,原因在于其信號(hào)分析性能取決于基函數(shù)的選取,,存在局限性。

    1998年Huang等人提出了一種全新的信號(hào)時(shí)頻分析方法——希爾伯特·黃變換(Hilbert-Huang Transform,,HHT)[9],。該方法首先采用經(jīng)驗(yàn)?zāi)B(tài)分解(Empirical Mode Decom-position,EMD)算法將非平穩(wěn)信號(hào)逐級(jí)分解為若干個(gè)固有模態(tài)函數(shù)(Intrinsic Mode Function,,IMF)和一個(gè)殘余量,,然后再對(duì)各個(gè)IMF分量進(jìn)行希爾伯特變換(Hilbert Transform,HT)得到能夠準(zhǔn)確反映信號(hào)能量在空間(或時(shí)間)各尺度上的分布規(guī)律[9]的Hilbert譜[10],。EMD具有數(shù)據(jù)驅(qū)動(dòng)的自適應(yīng)性,,能分析非線性非平穩(wěn)信號(hào),不受Heisenberg測(cè)不準(zhǔn)原理[11]制約等優(yōu)點(diǎn),。

    然而,Huang提出的基于篩分(Sifting)算法的EMD得到的IMF分量[12]存在模態(tài)混疊(Mode Mixing,,MM)[9],。模態(tài)混疊的出現(xiàn)不僅會(huì)導(dǎo)致錯(cuò)假的時(shí)頻分布,也使IMF失去物理意義,。圍繞模態(tài)混疊的消除或抑制,,國(guó)內(nèi)外開展了一系列的研究,并獲得不同程度的效果,。本文分別針對(duì)一維和多維EMD抑制模態(tài)混疊,,總結(jié)歸納了相關(guān)研究取得的主要成果,指出了各方法抑制效果的改進(jìn)及仍有的不足。最后討論了相關(guān)研究及應(yīng)用未來的發(fā)展趨勢(shì),。

1 經(jīng)驗(yàn)?zāi)B(tài)分解及模態(tài)混疊

    EMD自適應(yīng)的逐級(jí)分解[13]過程中,,IMF必須滿足以下兩個(gè)條件:(1)信號(hào)極值點(diǎn)和零點(diǎn)數(shù)相同或相差一個(gè);(2)由信號(hào)局部極大,、小值點(diǎn)擬合的上,、下包絡(luò)線的局部均值為零,也即上下包絡(luò)線關(guān)于時(shí)間軸局部對(duì)稱[14],。設(shè)待分解信號(hào)為X(t),,EMD算法的計(jì)算步驟如下[9]

zs2-gs1.gif

    式(1)說明EMD分解具有完備性[9],信號(hào)X(t)經(jīng)分解后還能通過所有IMF及剩余分量被精確重構(gòu)出來,。

    EMD在非線性非平穩(wěn)信號(hào)分析中具有顯著優(yōu)勢(shì),。與傳統(tǒng)時(shí)頻分析技術(shù)相比,EMD無需選擇基函數(shù),,其分解基于信號(hào)本身極值點(diǎn)的分布,。而算法本身缺少完整的理論基礎(chǔ),在實(shí)際計(jì)算與應(yīng)用中還存在著許多不足,,包括模態(tài)混疊[15],、端點(diǎn)效應(yīng)[16]、篩分迭代停止標(biāo)準(zhǔn)[12]等,。一般情況下,,每個(gè)固有模態(tài)函數(shù)只包含一種頻率成分,不存在模態(tài)混疊的現(xiàn)象,。但是,,當(dāng)信號(hào)中存在由異常事件(如間斷信號(hào)、脈沖干擾和噪聲等)引起的間歇(Intermittency)現(xiàn)象時(shí),,EMD的分解結(jié)果就會(huì)出現(xiàn)模態(tài)混疊[9],。

2 集合經(jīng)驗(yàn)?zāi)B(tài)分解

    為克服EMD的模態(tài)混疊,2009年Wu和Huang提出一種噪聲輔助信號(hào)分析方法——集合經(jīng)驗(yàn)?zāi)B(tài)分解(Ensemble EMD,,EEMD)[17],。該算法利用EMD濾波器組[18]行為及白噪聲頻譜均勻分布的統(tǒng)計(jì)特性[19],使Sifting過程信號(hào)極值點(diǎn)分布更趨勻稱,,有效抑制由間歇性高頻分量等因素造成的模態(tài)混疊,。設(shè)待分解信號(hào)為X(t),EEMD算法的計(jì)算步驟如下[17]

zs2-2-x1.gif

zs2-gs2.gif

    然而,,在EEMD中,,每個(gè)加噪信號(hào) hi(t)獨(dú)立地被分解,使得每個(gè) hi(t)分解后可能產(chǎn)生不同數(shù)量的IMF,,導(dǎo)致集合平均時(shí)IMF分量對(duì)齊困難,。此外,添加的白噪聲幅值和迭代次數(shù)依靠人為經(jīng)驗(yàn)設(shè)置,當(dāng)數(shù)值設(shè)置不當(dāng)時(shí),,無法克服模態(tài)混疊[20],。雖然增加集合平均次數(shù)可降低重構(gòu)誤差,但這是以增加計(jì)算成本為代價(jià),,且有限次數(shù)的集合平均并不能完全消除白噪聲,,導(dǎo)致算法重構(gòu)誤差大,分解完備性差[21],。

3 互補(bǔ)集合經(jīng)驗(yàn)?zāi)B(tài)分解

    Yeh等于2010年提出了互補(bǔ)集合經(jīng)驗(yàn)?zāi)B(tài)分解(Complementary EEMD,,CEEMD)[22]。該方法向原始信號(hào)中加入正負(fù)成對(duì)的輔助白噪聲,,在集合平均時(shí)相消,,能有效提高分解效率,克服EEMD重構(gòu)誤差大,、分解完備性差的問題,。設(shè)待分解信號(hào)為X(t),CEEMD算法的計(jì)算步驟如下[22]

zs2-3-x1.gif

zs2-3-x2.gif

4 自適應(yīng)噪聲的完整集合經(jīng)驗(yàn)?zāi)B(tài)分解

    為解決集合平均時(shí)IMF分量對(duì)齊問題,,TORRES M E等在2011年從分解過程和添加白噪聲上對(duì)CEEMD進(jìn)行改進(jìn),,提出了自適應(yīng)噪聲的完整集合經(jīng)驗(yàn)?zāi)B(tài)分解(Complete EEMD with Adaptive Noise,CEEMDAN)[24],。設(shè)待分解信號(hào)為X(t),,定義操作算子Ek(·)來表示信號(hào)經(jīng)過EMD分解后得到的第k階固有模態(tài)分量,CEEMDAN算法可描述如下[24]

zs2-4-x1.gif

zs2-4-x2.gif

    Wu和Huang建議[17]使用小振幅值來處理由高頻信號(hào)支配的數(shù)據(jù),,反之則增大噪聲幅值,。在分解過程中添加的是白噪聲經(jīng)EMD分解得到的各階IMF分量,最后重構(gòu)信號(hào)中的噪聲殘余比EEMD的結(jié)果小,,降低了篩選次數(shù),。另一方面,各組信號(hào)經(jīng)CEEMDAN分解出第一階固有模態(tài)分量后立即進(jìn)行集合平均,,避免了CEEMD中各組IMF分解結(jié)果差異造成最后集合平均難以對(duì)齊的問題,,也避免了其中某一階IMF分解效果不好時(shí),將影響傳遞給下一階,,影響后續(xù)分解,。盡管如此,CEEMDAN仍然有一些需要改進(jìn)的方面[23],,如 IMF仍包含殘余噪聲;在分解的早期階段,,信號(hào)會(huì)出現(xiàn)一些“虛假”模式,,導(dǎo)致在前兩階或三階模態(tài)中仍包含了大量的噪聲和信號(hào)的相似尺度[24,26]

5 改進(jìn)的自適應(yīng)噪聲集合經(jīng)驗(yàn)?zāi)B(tài)分解

    針對(duì)CEEMDAN存在的殘余噪聲及“虛假”模式問題,,TORRES M E等試圖估計(jì)每次分解剩余分量rk的“真實(shí)”平均包絡(luò),,進(jìn)一步提出了改進(jìn)算法[23]。定義M(·)為對(duì)信號(hào)進(jìn)行局部包絡(luò)平均運(yùn)算,,即取信號(hào)上下包絡(luò)的平均值,;ni(t)表示方差為1的零均值白噪聲。設(shè)待分解的信號(hào)為X(t),,改進(jìn)的CEEMDAN算法描述如下[23]:

zs2-5-x1.gif

zs2-5-x2.gif

    (6)判斷是否滿足終止條件,,若滿足,則停止分解,。

    與EEMD和CEEMDAN相比,,改進(jìn)的CEEMDAN引入局部包絡(luò)平均減小殘余噪聲;在分解過程中,,依次計(jì)算IMF,,保證了分解的完整性,信號(hào)重構(gòu)誤差更小,。但計(jì)算量過大,,實(shí)時(shí)性有待進(jìn)一步改進(jìn)[23,27],。

6 多維經(jīng)驗(yàn)?zāi)B(tài)分解及其噪聲輔助的模態(tài)混疊抑制

    將EMD直接用于分解多通道信號(hào)時(shí)存在各通道IMF分量在數(shù)量和頻率尺度上難以對(duì)齊問題,,使得重構(gòu)后各通道信號(hào)難以保持信號(hào)原有的相位關(guān)系[28]。Rehman等人在2010年提出了能夠同時(shí)處理多通道信號(hào)的多維經(jīng)驗(yàn)?zāi)B(tài)分解(Multivariate EMD,,MEMD)[28],。在此基礎(chǔ)上,將白噪聲作為信號(hào)其中一維或多維加入進(jìn)行MEMD處理,,提出了噪聲輔助多維經(jīng)驗(yàn)?zāi)B(tài)分解(Noise Assisted MEMD,,NA-MEMD)[29-30]。由于白噪聲具有頻譜均勻分布的統(tǒng)計(jì)特性,,該算法能有效抑制經(jīng)驗(yàn)?zāi)B(tài)分解存在的模態(tài)混疊,。

6.1 多元經(jīng)驗(yàn)?zāi)B(tài)分解

zs2-6.1-x1.gif

zs2-6.1-x2.gif

    MEMD 的提出解決了多通道信號(hào)的模式校準(zhǔn)問題。但MEMD分解也會(huì)得到一些虛假分量,,仍存在模態(tài)混疊問題[33],,影響對(duì)后續(xù)的特征提取。

6.2 噪聲輔助的多元經(jīng)驗(yàn)?zāi)B(tài)分解

zs2-6.2-x1.gif

    NA-MEMD方法是EMD的多變量噪聲擴(kuò)展形式,,算法不但充分利用了MEMD處理白噪聲時(shí)具有的固定通帶的頻率特性,,而且加入額外的獨(dú)立白噪聲確保分解后信號(hào)與噪聲的IMF分量完全可分離。相較于基于EEMD分解的方法無需進(jìn)行IMF的集合平均,,提高了計(jì)算效率,,減小了噪聲干擾,,性能更為優(yōu)越[33,34],。

7 結(jié)論

    EMD將信號(hào)進(jìn)行平穩(wěn)化處理的過程中存在模態(tài)混疊,,影響該方法的性能及應(yīng)用。本文圍繞模態(tài)混疊抑制,,總結(jié)歸納了一維及多維EMD研究方面的主要工作,。EEMD雖然能有效抑制模態(tài)混疊,但在分解過程中添加的輔助白噪聲最終需要增加集合平均次數(shù)來抵消,,計(jì)算耗時(shí)長(zhǎng),,重構(gòu)誤差大。CEEMD在抑制模態(tài)混疊的同時(shí)正負(fù)成對(duì)噪聲相消,,部分降低了殘留噪聲的影響,,減輕了集合平均抑制添加白噪聲的負(fù)擔(dān),提高了計(jì)算效率,。CEEMDAN及其改進(jìn)方法在每次分解時(shí)添加白噪聲的IMF分量,,添加噪聲逐級(jí)減少,固有模態(tài)分量中殘留噪聲更少,,有效減小了重構(gòu)誤差,,且在分解的每個(gè)階段都有一個(gè)全局停止標(biāo)準(zhǔn),分解效率最高,。MEMD對(duì)多維信號(hào)同時(shí)進(jìn)行分解,,確保了各通道IMF分量在數(shù)量和尺度上相匹配,重構(gòu)的各通道信號(hào)間的相位無畸變,。但由于其采用與EMD算法相一致的思想,, MIMF也會(huì)存在模態(tài)混疊。NA-MEMD通過引入輔助噪聲通道,,消除了MEMD中存在的模態(tài)混疊,,同時(shí)還保證了信號(hào)分解的完備性,分解性能最優(yōu),,但由于多維空間極值點(diǎn)包絡(luò)及局部均值的估計(jì)算法過于復(fù)雜,,計(jì)算量最大。特別是對(duì)空間單位球面的采樣顯著增加了采樣,,導(dǎo)致計(jì)算量快速增加,,分解效率最差。因而需要在計(jì)算精度和復(fù)雜度之間折衷考慮,。

    針對(duì)模態(tài)混疊抑制,,未來還可以從添加的輔助信號(hào)形態(tài)、發(fā)生模態(tài)混疊的IMF再處理及對(duì)信號(hào)濾波后再分解三個(gè)方面展開探索,。此外,,從理論上深入研究EMD處理過程中模態(tài)混疊發(fā)生的機(jī)理也有助于探索新的抑制方法,,提高EMD算法的精度和效率,提升其應(yīng)用水平和適應(yīng)范圍,。

參考文獻(xiàn)

[1] STEIN E M,WEISS G L.Introduction to Fourier analysis on Euclidean spaces[M].Princeton University Press,,1971,,212(2):484-503.

[2] BIRKHOFF G.A limitation of fourier analysis[J].Journal of Mathematics and Mechanics,1967,,17(5):443-447.

[3] GRIFFIN D,,LIM J S.Signal estimation from modified short-time Fourier transform[J].IEEE Transactions on Acoustics,Speech,,and Signal Processing,,1984,32(2):236-243.

[4] LOUGHLIN P J,,PITTON J W,,ATLAS L E.Bilinear time-frequency representations: new insights and properties[J].IEEE Transactions on Signal Processing,1993,,41(2):750-767.

[5] QIAN S E,,CHEN D P.Discrete Gabor transform[J].IEEE Transactions on Signal Processing,1993,,41(7):2429-2438.

[6] MALLAT S G.A theory for multiresolution signal decomposition:the wavelet representation[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,,1989,11(7): 674-693.

[7] ALMEIDA L B.The fractional Fourier transform and time-frequency representations[J].IEEE Transactions on Signal Processing,,1994,,42(11):3084-3091.

[8] ROEPSTORFF G.Fourier decomposition[M].Path Integral Approach to Quantum Physics.Springer Berlin Heidelberg,1994.

[9] HUANG N E,,SHEN Z,,LONG S R,et al.The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis[J].Mathematical,,Physical and Engineering Sciences,,1998,454(1971):903-995.

[10] HUANG N E,,WU Z H,,LONG S R,et al.On instantaneous frequency[J].Advances in Adaptive Data Analysis,,2009,,1(2):177-229.

[11] HEISENBERG W.Uber den anschaulichen Inhalt der quantentheoretischen Kinematik und Mechanik[J].Zeitschrift Für Physik,1927,,43(3-4):172-198.

[12] CHENG J S,,YU D J,,YANG Y.Research on the intrinsic mode function(IMF)criterion in EMD method[J].Mechanical Systems and Signal Processing,2006,,20(4):817-824.

[13] HUANG N E,,WU M L C,LONG S R,,et al.A confidence limit for the empirical mode decomposition and hilbert spectral analysis[J].Mathematical,,Physical and Engineering Sciences,2003,,459(2037):2317-2345.

[14] WANG G,,CHEN X Y,QIAO F L,,et al.On intrinsic mode function[J].Advances in Adaptive Data Analysis,,2010,2(3):277-293.

[15] HU X Y,,PENG S L,,HWANG W L.EMD revisited:a new understanding of the envelope and resolving the mode-mixing problem in AM-FM signals[J].IEEE Transactions on Signal Processing,2012,,60(3):1075-1086.

[16] SU Y X,,LIU Z G,LI K L,,et al.A new method for end effect of EMD and its application to harmonic analysis[J].Advanced Technology of Electrical Engineering and Energy,,2008,27(2):33.

[17] WU Z H,,HUANG N E.Ensemble empirical mode decomposition:a noise-assisted data analysis method[J].Ad vances in Adaptive Data Analysis,,2009,1(1):1793-5369.

[18] FLANDRIN P,,RILLING G,,GONCALVES P.Empirical mode decomposition as a filter bank[J].IEEE Signal Processing Letters,2004,,11(2):112-114.

[19] WU Z H,,HUANG N E.A study of the characteristics of white noise using the empirical mode decomposition method[J].Mathematical,Physical and Engineering Sciences,,2004,,460(2046):1597-1611.

[20] HUANG N E,SHEN S S P.Hilbert-Huang transform and its applications[M].World Scientific,,2005.

[21] HELSKE J,,LUUKKO P.Ensemble empirical mode decomposition(EEMD) and its completevariant (CEEMDAN)[J].International Journal of Public Health,2015,,60(5):1-9.

[22] YEH J R,,SHIEH J S,,HUANG N E.Complementary ensemble empirical mode decomposition: a novel noise enhanced data analysis method[J].Advances in Adaptive Data Analysis,2010,,2(2):135-156.

[23] COLOMINAS M A,,SCHLOTTHAUER G,TORRES M E.Improved complete ensemble EMD:a suitable tool for biomedical signal processing[J].Biomedical Signal Processing and Control,,2014,,14(1):19-29.

[24] TORRES M E,COLOMINAS M A,,SCHLOTTHAUER G,et al.A complete ensemble empirical mode decomposition with adaptive noise[C].2011 IEEE International Conference on Acoustics,,Speech and Signal Processing(ICASSP),,2011:4144-4147.

[25] COLOMINAS M A,SCHLOTTHAUER G,,TORRES M E,,et al.Noise-assisted emd methods in action[J].Advances in Adaptive Data Analysis,2012,,4(4):1793-5369.

[26] COLOMINAS M A,,SCHLOTTHAUER G,F(xiàn)LANDRIN P,,et al.Descomposición empírica en modos por conjuntos completa con ruido adaptativo y aplicaciones biomédicas[C].XVIII Congreso Argentino de Bioingeniería SABI 2011-VII Jornadas de Ingeniería Clínica.2011.

[27] HUMEAU-HEURTIER A,,ABRAHAM P,MAHE G.Analysis of laser speckle contrast images variability using a novel empirical mode decomposition: comparison of results with laser Doppler flowmetry signals variability[J].IEEE Transactions on Medical Imaging,,2015,,34(2):618-627.

[28] REHMAN N U,MANDIC D P.Multivariate empirical mode decomposition[J].Mathematical,,Physical and Engineering Sciences,,2010,466(2117):1291-1302.

[29] REHMAN N U,,MANDIC D P.Filter bank property of multivariate empirical mode decomposition[J].IEEE Transactions on Signal Processing,,2011,59(5):2421-2426.

[30] REHMAN N U,,PARK C,,HUANG N E,et al.EMD via MEMD:multivariate noise-aided computation of standard EMD[J].Advances in Adaptive Data Analysis,,2013,,5(2):1793-5369.

[31] MANDIC D P,REHMAN N U,,WU Z H,,et al.Empirical mode decomposition- based time-frequency analysis of multivariate signals:the power of adaptive data analysis[J].IEEE Signal Processing Magazine,,2013,30(6):74-86.

[32] CUI J J,,F(xiàn)REEDEN W.Equidistribution on the Sphere[M].Society for Industrial and Applied Mathematics,,1997,18(2):595-609.

[33] PARK C,,LOONEY D,,REHMAN N U,et al.Classification of motor imagery BCI using multivariate empirical mode decomposition[J].IEEE Transactions on Neural Systems and Rehabilitation Engineering,,2013,,21(1):10-22.

[34] LOONEY D,MANDIC D P.Multiscale image fusion using complex extensions of EMD[J].IEEE Transactions on Signal Processing,,2009,,57(4):1626-1630.




作者信息:

戴  婷1,張榆鋒1,,章克信2,,何冰冰1,朱泓萱1,,張俊華1

(1.云南大學(xué) 信息學(xué)院電子工程系,,云南 昆明650091;2.昆明醫(yī)科大學(xué)第二附屬醫(yī)院,,云南 昆明650031)

此內(nèi)容為AET網(wǎng)站原創(chuàng),,未經(jīng)授權(quán)禁止轉(zhuǎn)載。