[an error occurred while processing this directive]

高等学校化学研究 ›› 2011, Vol. 27 ›› Issue (3): 385-391.

• Articles • 上一篇    下一篇

Determination of Quality Properties of Soy Sauce by Support Vector Regression Coupled with SW-NIR Spectroscopy

LIU Tong1,2, BAO Chun-fang1 and REN Yu-lin1*   

  1. 1. College of Chemistry, Jilin University, Changchun 130021, P. R. China;
    2. Harbin Centre for Disease Control and Prevention, Harbin 150056, P. R. China
  • 收稿日期:2010-06-21 修回日期:2010-08-25 出版日期:2011-05-25 发布日期:2011-04-29
  • 通讯作者: REN Yu-lin E-mail:ryl@jlu.edu.cn
  • 基金资助:

    Supported by the Harbin Technological Innovation Special Fund Research Projects, China(No.RC2006QN020015).

Determination of Quality Properties of Soy Sauce by Support Vector Regression Coupled with SW-NIR Spectroscopy

LIU Tong1,2, BAO Chun-fang1 and REN Yu-lin1*   

  1. 1. College of Chemistry, Jilin University, Changchun 130021, P. R. China;
    2. Harbin Centre for Disease Control and Prevention, Harbin 150056, P. R. China
  • Received:2010-06-21 Revised:2010-08-25 Online:2011-05-25 Published:2011-04-29
  • Contact: REN Yu-lin E-mail:ryl@jlu.edu.cn
  • Supported by:

    Supported by the Harbin Technological Innovation Special Fund Research Projects, China(No.RC2006QN020015).

PDF (PC)

被引次数

11

摘要： The modern near-infrared(NIR) spectroscopy analysis is a simple, efficient and nondestructive technique, which has been used in chemical analysis in diverse fields. Shortwave NIR spectroscopy is also a rapid, flexible, and cost-effective method to control product quality in food industry. The method of support vector regression coupled with shortwave NIR spectroscopy was explored for the nondestructive quantitative analysis of the important quality parameters of soy sauce, including amino nitrogen content, total acid content, salt content and color ratio. In this study, the support vector regression(SVR) models based on subtractive spectra and positive spectra were found and compared, the results show that the subtractive spectrum was more excellent than the positive spectrum. Meanwhile, R and RSE were determined, respectively, by means of original spectra and pretreated spectra[standard normal variate (SNV), first-derivative and second-derivative], and the corresponding models were successfully established. The best prediction was achieved by a support vector regression model of the first derivative transformed dataset. In addition, the result obtained by the proposed method was compared with that of Partial Least Squares(PLS), which showed that the generalization performance of the classifier based on SVR was much better than that of PLS. The results demonstrate that shortwave NIR spectroscopy combined with SVR is promising for the quality control of soy sauce.

关键词: Shortwave near-infrared spectroscopy, Support vector regression, Processing, Soy sauce

Abstract: The modern near-infrared(NIR) spectroscopy analysis is a simple, efficient and nondestructive technique, which has been used in chemical analysis in diverse fields. Shortwave NIR spectroscopy is also a rapid, flexible, and cost-effective method to control product quality in food industry. The method of support vector regression coupled with shortwave NIR spectroscopy was explored for the nondestructive quantitative analysis of the important quality parameters of soy sauce, including amino nitrogen content, total acid content, salt content and color ratio. In this study, the support vector regression(SVR) models based on subtractive spectra and positive spectra were found and compared, the results show that the subtractive spectrum was more excellent than the positive spectrum. Meanwhile, R and RSE were determined, respectively, by means of original spectra and pretreated spectra[standard normal variate (SNV), first-derivative and second-derivative], and the corresponding models were successfully established. The best prediction was achieved by a support vector regression model of the first derivative transformed dataset. In addition, the result obtained by the proposed method was compared with that of Partial Least Squares(PLS), which showed that the generalization performance of the classifier based on SVR was much better than that of PLS. The results demonstrate that shortwave NIR spectroscopy combined with SVR is promising for the quality control of soy sauce.

Key words: Shortwave near-infrared spectroscopy, Support vector regression, Processing, Soy sauce