摘要
针对电子鼻的数据特点,提出用一个3维数组保存电子鼻的数据;采用6点平滑方法去除传感器的噪声;在基线校正中,首先通过二阶导数大于零和连续一阶导数大于零的方法找到样本反应起始点,然后减去环境响应值并提取相同长度的数据段,以提高电子鼻的精度和可重复性。对预处理前后的电子鼻数据中提出的特征进行主成分分析发现,预处理后的主成分结果所含的有用信息更多,而且可以很好地区分红富士和姬娜两种不同香味的苹果。
In order to improve the ability of recognition, the influence of data preprocessing in electronic nose was discussed. A 3-D data set was proposed to save electronic nose data. 6 points-smoothing method was used to de-noise the gas sensors data. The baseline was removed and the same respond time was extracted to diminish the influence of environment and improve the precision and repetition of the electronic nose measurements. The first derivative and second derivative of sensors data were used to find the beginning respond points. The maximum of each sensor＇s respond were extracted as features. Principle component analysis was used to analyze the extracted features before and after preprocessing. The results showed that ＂fuji＂ and ＂jina＂ apples were well distinguished after preprocessing. After all, the preprocessing is the base of feature extraction and pattern recognition in electronic nose.
出处
《农业机械学报》
EI
CAS
CSCD
北大核心
2006年第5期83-86,共4页
Transactions of the Chinese Society for Agricultural Machinery
基金
国家"863"高技术研究发展计划资助项目(项目编号:2002AA248051)
国家自然科学基金资助项目(项目编号:30370813)
国家教委博士点基金项目
关键词
电子鼻
数据预处理
主成分分析
Electronic nose, Preprocessing, Principle component analysis
