徐喜荣,潘子琦,李兴华,李金泽.基于小波变换和连续Hopfield神经网络的重叠峰解析策略[J].,2019,59(1):97-105 |
基于小波变换和连续Hopfield神经网络的重叠峰解析策略 |
Analytical strategy of overlapping peaks based on wavelet transform and continuous Hopfield neural network |
|
DOI:10.7511/dllgxb201901013 |
中文关键词:重叠峰解析曲线拟合小波变换Hopfield神经网络 |
英文关键词:overlapping peaks analysiscurve fittingwavelet transformHopfield neural network |
基金项目:国家自然科学基金资助项目(614724656117030361562066);国家级大学生创新训练项目(2016101410168). |
|
摘要点击次数:984 |
全文下载次数:792 |
中文摘要: |
针对化学谱图分析中的重叠峰解析问题,提出了一种基于小波变换和连续Hopfield神经网络的谱图重叠峰解析策略.使用高斯函数作为曲线拟合的数学模型,将重叠峰分解为一系列高斯峰的叠加组合.首先,根据检测信号卷积型小波变换中奇异点特性分析原理,利用卷积型小波变换方法寻找信号的特征点,并根据特征点对信号做出初步解析;然后,构造解析峰参数与解析误差之间的函数关系,并以之作为连续Hopfield神经网络的能量函数;最后,根据该网络自我演化特性经过指定次数的迭代得到误差函数的极小值点,进而得到谱图重叠峰的最终解析参数.谱图重叠峰模拟解析表明,所提算法有效. |
英文摘要: |
Aiming at the problem of chemical spectral overlapping peaks analysis, an analytical strategy for spectral overlapping peaks based on wavelet transform and continuous Hopfield neural network is proposed. Using Gauss function as a mathematical model of curve fitting, the overlapping peaks are decomposed into a series of superimposed combinations of Gauss peaks. Firstly, according to the principle of singularity analysis of convolution wavelet transform and using the convolution wavelet transform, the characteristic points of the signal are found and a preliminary analysis of the signal is made. Secondly, the function relation between analytic peak parameter and analytic error is constructed and used as the energy function of continuous Hopfield neural network. Finally, according to the self-evolution characteristic of network, the minimum point of the error function can be obtained by iterating the specified number, and the final analytic parameter of the spectral overlapping peaks is obtained. The analysis results of the overlapping peaks of the simulated spectrum indicate that the proposed algorithm is effective. |
查看全文查看/发表评论下载PDF阅读器 |
| --> 关闭 |