WANG Fang, LI Jin-hua
（Engineering Technology Research Center of Shanxi Province for Opto-Electronic Information and Instrument, Taiyuan 030051, China）
Abstract： Monitoring indoor harmful gas can obtain the infrared spectra of mixed harmful gases. Since the absorption bands of mixed gases overlap and their qualitative and quantitative analyses are not easy, feature extraction method based on joint approximative diagonalization of eigenmatrix (JADE) is proposed. By fully mining the hidden information of original data and analyzing higher-order statistics information of the data, each substance spectrum in the mixed gas can be accurately distinguished. In addition, a multi-dimensional data quantitative analysis model of the extracted independent source is established by using support vector machine (SVM) based on regular theory. The experimental results show that the correlation coefficients of the components of mixed gases is above 0.999 1 by quantitative analysis, which verifies the accuracy of this feature extraction method.
Key words： aliasing peak identification; joint approximative diagonalization of eigenmatrix (JADE); quantitative analysis; support vector machine (SVM)
CLD number： TN219 Document code： A
Article ID： 1674-8042（2016）01-0024-06 doi： 10.3969/j.issn.1674-8042.2016.01.005
References
［1］ CHEN Meng. Study on indoor air pollution hazards and solutions. Heilongjiang Science and Technology Information, 2014, (4)： 2.
［2］ CHEN Xi-yao. Discussion of the environmental pollution and preventive measures caused by interior renovation.Resources Economization & Environmental Protection, 2014, (11)： 88.
［3］ WANG Deng-shan. Research of indoor air pollution hazards and purification technology.Contamination Control & Air-Conditioning Technology, 2015, (2)： 33-36.
［4］ SONG Ying-hua. On the application prospect of infrared spectrum technology in the environmental safety field.Energy and Energy Conservation, 2015, (8)： 104-105.
［5］ LI Ji-guang. Investigating the synthetic mechanism of energy compounds by on-line IR spectroscopy combined with independent component analysis. Xi’an： Northwest University, 2014.
［6］ XIAO Da-wei, CHENG Jin-fang, ZHANG Jing-zhuo, et al. Blind signal estimation based on JADE Algorithm for an vector hydrophone array. Journal of Wuhan University of Technology (Transportation Science & Engineering), 2013, (5)： 1012-1016.
［7］ WANG Wen-tao, ZHANG Jian-yun, LIU Xing-hua, et al. Radar anti-mainlobe-jamming based on blind source separation algorithm of JADE. Fire Control & Command Cortol, 2015, (9)： 104-108.
［8］ SONG Qing. The basic and applied research of indepent component analysis in spectral analysis. Shanghai： The Second Military Medical University, 2012.
［9］ BIAN Shuang-wei. Fault diagnosis on Tennessee-Eastman process. Wuhan： Huazhong University of Science and Technology, 2011.
［10］ Cardoso J F. Higher order contrast for independent component analysis. Neural Computation, 1999, 11(1)： 157-193.
［11］ Cardoso J F, Souloumiac A. Blind beamforming for non-Gaussian signals. In: IEE Proceedings of Radar, Sonar and Navigation, 1994, 140(6)： 362-370.
［12］ LIN Ji-peng, LIU Jun-hua. Support vector machine and optimized method for spectral analysis. Spectroscopy and Spectral Analysis, 2006, (12)： 2232-2235.
［13］ LIN Ji-peng, LIU Jun-hua. A new technology study based on seriously overlapped spectrum of quantitative analyzing on multi-hybrid gases. Modern Scientific Instruments, 2006， (1)： 53-57.
基于JADE的室内多组分污染气体混叠峰识别
王芳, 李晋华
（山西省光电信息与仪器工程技术研究中心， 山西 太原 030051）
摘 要： 检测室内有害气体可得到混合有害气体的红外光谱。 由于吸收谱带相互交叠的混合气体不易进行定性定量分析， 提出了基于特征矩阵联合近似对角化(JADE)的特征提取方法。 该方法通过分析数据的高阶统计量信息， 充分挖掘原始数据隐含的信息， 从而准确区分出混合气体中各物质的光谱， 同时应用基于正则理论的支持向量机(SVM)对提取的独立信号源建立多维数据定量分析的模型。 实验结果表明， 混合气体中各组分的定量分析相关系数均保持在0.999 1以上， 验证了该特征提取方法的准确性。
关键词： 混叠峰识别； 特征矩阵联合近似对角化； 定量分析； 支持向量机
引用格式： WANG Fang, LI Jin-hua. Identification of indoor multi-component pollution gas aliasing peak based on JADE. Journal of Measurement Science and Instrumentation, 2016, 7(1)： 24-29. ［doi： 10.3969／j.issn.1674-8042.2016.01.005］
[full text view]
