EMD Time-Frequency Analysis of Raman Spectrum and NIR
ZHAO Xiao-yu1, FANG Yi-ming2, TAN Feng1, TONG Liang3, ZHAI Zhe4
1. College of Information Technology, Heilongjiang Bayi Agricultural University, Daqing 163319, China 2. College of Electrical Engineering, Yanshan University, Qinhuangdao 066004, China 3. Communication and Electronic Engineering Institute, Qiqihar University, Qiqihar 161006, China 4. Forestry Experiment Center of North China, Chinese Academy of Forestry, Beijing 102300, China
Abstract：This paper analyzes the Raman spectrum and Near Infrared Spectrum (NIR) with time-frequency method. The empirical mode decomposition spectrum becomes intrinsic mode functions, which the proportion calculation reveals the Raman spectral energy is uniform distributed in each component, while the NIR’s low order intrinsic mode functions only undertakes fewer primary spectroscopic effective information. Both the real spectrum and numerical experiments show that the empirical mode decomposition (EMD) regard Raman spectrum as the amplitude-modulated signal, which possessed with high frequency adsorption property; and EMD regards NIR as the frequency-modulated signal, which could be preferably realized high frequency narrow-band demodulation during first-order intrinsic mode functions. The first-order intrinsic mode functions Hilbert transform reveals that during the period of empirical mode decomposes Raman spectrum, modal aliasing happened. Through further analysis of corn leaf’s NIR in time-frequency domain, after EMD, the first and second orders components of low energy are cut off, and reconstruct spectral signal by using the remaining intrinsic mode functions, the root-mean-square error is 1.001 1, and the correlation coefficient is 0.981 3, both of these two indexes indicated higher accuracy in re-construction; the decomposition trend term indicates the absorbency is ascending along with the decreasing to wave length in the near-infrared light wave band; and the Hilbert transform of characteristic modal component displays, 657 cm-1 is the specific frequency by the corn leaf stress spectrum, which could be regarded as characteristic frequency for identification.
[1] Huang N E，Zheng Shen，Long S R，et al. Proc. R Soc. Lond A，1998，454: 903. [2] CAI Jian-hua, WANG Xian-chun, HU Wei-wen(蔡剑华, 王先春, 胡惟文). Transactions of the Chinese Society for Agricultural Machinery(农业机械学报), 2010, 41(9): 182. [3] CAI Jian-hua, WANG Xian-chun(蔡剑华, 王先春). Acta Optica Sinica(光学学报), 2010, 30(1): 267. [4] ZHAO Xiao-yu, FANG Yi-ming, WANG Zhi-gang(赵肖宇, 方一鸣, 王志刚). Spectroscopy and Spectral Analysis(光谱学与光谱分析), 2013, 33(12): 1. [5] Huang N E, Wu M, Long S, et al. Proc. R Soc. Lond A, 2003, 459: 2317. [6] Huang N E, Long S R. Annual Review of Fluid Mechanics, 1999, 31: 417. [7] Wu Z H，Huang N E. Advances in Adaptive Data Analysis，2009, 1(1): 1.
