电子鼻传感器漂移噪声降噪方法研究
发布时间：2018-12-15 11:50
【摘要】：在现代社会中，人们对食品质量检测的要求越来越高，而传统的化学分析方式检测时间长、成本高、效率低，且需要专业人员进行操作，不利于普及和推广。而检测方便且快速的电子鼻系统的出现正在代替传统的化学分析方式，目前，虽然电子鼻技术取得了极大的进展，但大多数还停留在实验室阶段，还没有广泛实用化，主要原因是电子鼻的漂移问题、背景干扰、个体差异等问题还没有得到有效解决。 通过长期实验数据研究分析发现，漂移主要分为短期漂移与长期漂移两种，其中长期漂移具有变化缓慢且无规律的现象。本文把降低或者消除电子鼻的长期漂移作为主要研究内容。 为了解决电子鼻长期漂移导致鉴别正确率降低这一问题，提出了一种基于快速傅里叶变换的均值偏差率阈值函数来去除电子鼻漂移噪声的方法，通过构造快速傅里叶变换系数的均值偏差率阈值函数，实现了对傅里叶变换系数的动态处理，有效去除了电子鼻漂移噪声。6种白酒检测实例表明，此方法能够有效消除电子鼻长期漂移噪声信号对传感器造成的影响，从而显著的提高了鉴别正确率。 傅里叶变换刻画的是电子鼻检测信号在整个时域的频谱特征，虽然在去除电子鼻漂移信号过程中有较好的应用效果，但是并没有获得漂移信号的特征信息。针对这一问题，本文提出一种在小波变换的基础上结合均值偏差率阈值函数对小波变换系数进行处理的方法，由于小波变换可以实现电子鼻检测信号在时域和频域的局部变换，克服了傅里叶变换的缺点，便于观察分析。实验结果表明，与傅里叶变换相比，小波变换能够进一步去除电子鼻的漂移信号，从而使分类效果更佳。 独立成分分析能从检测的混合信号中分离出分布未知但统计相互独立的源信号，，结合漂移信号未知性等特点，本文尝试从独立成分分析的角度对传感器的检测信号进行分析研究。通过利用独立成分分析，提取与白酒挥发出的气体最相关的独立成分，然后对其进行分析研究，探究漂移规律，从而达到去除漂移噪声信号的目的。实验结果表明，独立成分分析能够有效的提取独立成分，去除了一定的漂移信号。
[Abstract]:In modern society, the demand of food quality detection is more and more high, but the traditional chemical analysis method has long time, high cost, low efficiency, and requires professional operation, which is not conducive to popularization and promotion. The appearance of the electronic nose system, which is convenient and rapid, is replacing the traditional chemical analysis method. Although the electronic nose technology has made great progress at present, most of them are still in the laboratory stage and have not been widely applied. The main reasons are that the drift of electronic nose, background interference and individual differences have not been effectively solved. Through the analysis of long-term experimental data, it is found that the drift is mainly divided into two types: short-term drift and long-term drift, and the medium- and long-term drift has the phenomenon of slow and irregular change. In this paper, reducing or eliminating long-term drift of electronic nose is the main research content. In order to solve the problem that the long term drift of electronic nose leads to the reduction of the discrimination accuracy, a method based on the mean deviation rate threshold function based on fast Fourier transform (FFT) is proposed to remove the drift noise of electronic nose. By constructing the mean deviation rate threshold function of the fast Fourier transform coefficient, the dynamic processing of the Fourier transform coefficient is realized, and the electronic nose drift noise is effectively removed. This method can effectively eliminate the effect of the long-term drift noise signal on the sensor and improve the discrimination accuracy. Fourier transform depicts the spectrum characteristics of the electronic nose detection signal in the whole time domain. Although it has good application effect in the process of removing the electronic nose drift signal, it does not obtain the characteristic information of the drift signal. In order to solve this problem, this paper presents a method to deal with wavelet transform coefficients based on wavelet transform combined with mean deviation rate threshold function. Because wavelet transform can realize the local transformation of electronic nose detection signal in time domain and frequency domain. It overcomes the shortcoming of Fourier transform and is easy to observe and analyze. The experimental results show that the wavelet transform can remove the drift signal of the electronic nose further than the Fourier transform, so that the classification effect is better. Independent component analysis (ICA) can separate source signals with unknown distribution but independent statistics from the detected mixed signals, combining with the unknown characteristics of drift signals. In this paper, the detection signals of sensors are analyzed and studied from the angle of independent component analysis (ICA). Through the use of independent component analysis, extract the most relevant independent components of the gas volatilized from liquor, then analyze and study it, explore the drift law, so as to achieve the purpose of removing drift noise signal. The experimental results show that independent component analysis can effectively extract independent components and remove certain drift signals.
【学位授予单位】：河南科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP212;TB535
本文编号：2380586
[Abstract]:In modern society, the demand of food quality detection is more and more high, but the traditional chemical analysis method has long time, high cost, low efficiency, and requires professional operation, which is not conducive to popularization and promotion. The appearance of the electronic nose system, which is convenient and rapid, is replacing the traditional chemical analysis method. Although the electronic nose technology has made great progress at present, most of them are still in the laboratory stage and have not been widely applied. The main reasons are that the drift of electronic nose, background interference and individual differences have not been effectively solved. Through the analysis of long-term experimental data, it is found that the drift is mainly divided into two types: short-term drift and long-term drift, and the medium- and long-term drift has the phenomenon of slow and irregular change. In this paper, reducing or eliminating long-term drift of electronic nose is the main research content. In order to solve the problem that the long term drift of electronic nose leads to the reduction of the discrimination accuracy, a method based on the mean deviation rate threshold function based on fast Fourier transform (FFT) is proposed to remove the drift noise of electronic nose. By constructing the mean deviation rate threshold function of the fast Fourier transform coefficient, the dynamic processing of the Fourier transform coefficient is realized, and the electronic nose drift noise is effectively removed. This method can effectively eliminate the effect of the long-term drift noise signal on the sensor and improve the discrimination accuracy. Fourier transform depicts the spectrum characteristics of the electronic nose detection signal in the whole time domain. Although it has good application effect in the process of removing the electronic nose drift signal, it does not obtain the characteristic information of the drift signal. In order to solve this problem, this paper presents a method to deal with wavelet transform coefficients based on wavelet transform combined with mean deviation rate threshold function. Because wavelet transform can realize the local transformation of electronic nose detection signal in time domain and frequency domain. It overcomes the shortcoming of Fourier transform and is easy to observe and analyze. The experimental results show that the wavelet transform can remove the drift signal of the electronic nose further than the Fourier transform, so that the classification effect is better. Independent component analysis (ICA) can separate source signals with unknown distribution but independent statistics from the detected mixed signals, combining with the unknown characteristics of drift signals. In this paper, the detection signals of sensors are analyzed and studied from the angle of independent component analysis (ICA). Through the use of independent component analysis, extract the most relevant independent components of the gas volatilized from liquor, then analyze and study it, explore the drift law, so as to achieve the purpose of removing drift noise signal. The experimental results show that independent component analysis can effectively extract independent components and remove certain drift signals.
【学位授予单位】：河南科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP212;TB535
【参考文献】
中国期刊全文数据库 前10条
1 乔玉坤;王仕成;张金生;张琪;孙渊;;小波域软阈值降噪的地磁匹配导航仿真研究[J];兵工学报;2011年09期
2 王立川;张覃轶;黄伟;;基于气体传感器阵列的白酒特征分析[J];传感技术学报;2010年12期
3 张红梅;王俊;余泳昌;高献坤;花恒明;何玉静;;基于电子鼻技术的信阳毛尖茶咖啡碱检测方法[J];传感技术学报;2011年08期
4 惠国华;吴玉玲;陈裕泉;;基于电子鼻的香蕉储存时间鉴别方法研究[J];传感技术学报;2012年05期
5 高明星;王晓宁;龚煈;廖青;;电子鼻在棉织物异味检测中的应用研究[J];传感器与微系统;2011年06期
6 周海涛;殷勇;;主成分分析在电子鼻鉴别分析中存在的缺陷及其改进方法[J];传感器世界;2008年10期
7 卢松升;刘正之;;基于小波的电力系统信号预定阈值去噪技术[J];电工技术学报;2005年11期
8 张青春;叶小婷;;基于电子鼻技术的混合气体检测方法研究[J];电子技术应用;2011年12期
9 万灵达;杨晓光;;一种基于FFT的高精度频率估计算法[J];电子科技;2010年10期
10 文玉梅,李平;通过信号分离抑制离散变化变量传感器漂移噪声[J];电子学报;1999年11期
本文编号：2380586
本文链接：https://www.wllwen.com/guanlilunwen/gongchengguanli/2380586.html
