摘要
详细讨论了网络优化参数、模拟的测量过程中噪声及杂质对网络收敛性能及预测误差的影响。为加速网络收敛,提高紫外光谱检索的正确率,采用了导数光谱对反向传播的人工神经网络(BP-ANN)进行训练和检索,该方法对检索光谱中噪声、杂质,尤其是斜坡背景的允许程度明显提高。文章还将ANN方法与普通的相关系数法的识别结果进行了比较。结果表明,优化参数下的人工神经网络的库检索法在抗噪、容杂等方面都明显地优于普通的相关系数法,是一种很有效的紫外库检索方法。
The effects of optimization of network parameters, noise, and impurity on the network were investigated detailedly. To speed up the convergence of the network and enhance the resolution of the library search of UV spectra, the derivative spectra for BP-ANN library search was proposed. The method has a higher tolerance to noise and impurity levels than using ordinary UV spectra, especially to slop background levels. Finally, the resolutions of library search of UV spectra with ANN with optimized parameters were compared with conventional correlation coefficient method. Results showed that the ANN is superior to conventional correlation coefficient method and is an effective method for library search of UV spectra.
出处
《光谱学与光谱分析》
SCIE
EI
CAS
CSCD
北大核心
2006年第5期908-912,共5页
Spectroscopy and Spectral Analysis
基金
首都师范大学资源环境与地理信息系统北京市重点实验室开放基金项目(2004211-03)资助
关键词
人工神经网络
有机环境污染物
紫外光谱
库检索
Artificial neural network
Environmental pollutant
Ultraviolet spectra
Library search