Development of e-nose herb recognition system based on artificial intelligence techniques

Herbs are useful for various applications especially in nutraceutical products and botanical medicine. In normal practice, the herbs identification is done mainly by botanists. However, it is difficult for botanists to recognize herbs based on aroma for the species under the same family. Thereupon, the herbs odors under the same family which is the physical appearance may look almost the same characteristic and also may be having the almost same aromas. Moreover, many factors might influence the accuracy of the human olfactory system as a panel sensory such as physical, mental and fatigue body conditions. Other factors, it requires various experimental exercises, very timeconsuming, less efficient and costly. Electronic nose (E-nose) instruments, derived from numerous types of aroma sensor technologies have been developed for a diverse of applications in a broad field of agriculture including for herbs. The intervention of electronic nose was capable to reproduced human senses using sensor arrays and pattern recognition systems. E-nose in this project was developed as portable type, small size and easy to operate. The ability of the developed E-nose was emphasized to distinctive herbs leaves odor from Lauraceae, Myrtaceae and Zingiberaceae families. Multiple metal oxide semiconductor (MOS) gas sensors were assembled in the E-nose system to detect a broad range of chemical compound that released from the sample. The selected MOS gas sensors were TGS 2610, TGS 2611, TGS 2620, TGS 823 and TGS 832 from Figaro Inc. which was installed in the Enose system as detection array. Meanwhile, the blended herb leaves prepared in sample preparation was found as a preeminent procedure that gives the advantage to secure the long-lasting function of a gas sensor compared to the existing sample preparation in another E-nose system. Finally, data captured by the gas sensors was classified by using two methods which are Artificial Neural Network (ANN) and Adaptive Neuro-Fuzzy Inference System (ANFIS). The percentage of accuracy to classify the herbs species by using ANFIS and ANN was compare to evaluate the effectiveness accordingly. From the result, ANFIS gives as higher as 94.8% percentage of accuracy compare than ANN for 91.7% of accuracy.

Preview Text FK 2016 108 - IR.pdf Download (482kB) | Preview

Actions (login required)

View Item