电子鼻传感器阵列优化对猪肉新鲜度法的检测

Pork Freshness Detection Using Optimized Electronic Nose Sensor Array

用电子鼻检测猪肉新鲜度时,传感器阵列的冗余信息会带来负面影响。为了提高识别的准确性,根据猪肉散发的气味选择初始的传感器阵列,利用方差分析方法剔除重复性和区分度不明显的传感器;再通过变异系数分析、相关系数绝对值累加和最小分析、主成分分析(principal component analysis,PCA)第2主成分系数分析,筛选出了适合检测猪肉新鲜度的传感器阵列的优化阵列。本研究采用逐步判别法筛选出合适的特征值,并用贝叶斯判别方法对传感器阵列优化前后的数据进行对比分析。结果表明:通过对传感器阵列的优化,识别率由优化前的86.8%提高到优化后的98.9%。研究表明,本实验的传感器阵列优化方法可以大大提高电子鼻对猪肉新鲜度的识别准确性。

When electronic nose is used for detecting pork freshness, sensor array optimization has a great influence on improving the accuracy by eliminating the negative effects brought about by the redundant information. The initial sensor array was determined by the odor released from pork. Then the sensors with poor repeatability and differentiation were excluded by analysis of variance（ANOVA）. By coefficient of variation, minimum cumulation of absolute correlation coefficient and the second principal component of principal component analysis（PCA）, an optimized sensor array was selected for the detection of pork freshness. This study adopted stepwise discriminant analysis to optimize features and compare the data before and after optimization by using Bayes discriminant method. Results showed that by sensor optimization and feature optimization, the accuracy was increased from 86.8% to 98.9%. This study indicates that sensor array optimization and feature optimization can greatly improve the detection accuracy of pork freshness.