臺灣博碩士論文加值系統

香氣是大多數品咖啡者欣賞和選擇咖啡時主要考慮的因素之一，所以香氣味道一直被認為是咖啡重要的特質指標；但是最常見的評估氣味方式是仰賴於訓練有素的杯測師之主觀評鑑測試。為了擬真杯測師且客觀辨識氣味，本研究旨在利用機器學習及深度學習識別不同產區之新鮮烘焙咖啡香氣；使用5個半導體氣體感測器品測咖啡的氣味，針對4個不同產區的特殊氣味進行辨識。
AI嗅覺感測器陣列是一種模仿人工嗅覺之神經系統，因此本研究要實現人工智慧進行咖啡香氣之分析，採用機器學習演算法：決策樹、k近鄰法、支援向量機、隨機森林；採用深度學習演算法：卷積神經網路，藉由人工智慧演算法訓練模型對氣味數據進行分類準確度比較。近年來，由於類神經網路的興起，深度學習演算法已被使用在氣味分類，數據處理的方式是將氣體感測器偵測波形轉換成為圖像，卻缺少了氣味指紋，本研究以深度學習方法將各產區之咖啡氣味數據視覺化成香氣電子指紋，並經由類神經網路進行識別分類單品咖啡產區。
應用卷積神經網路於咖啡氣味分類問題之可行性，未來可以再透過更精密更廣泛的氣味感測器及資料收集器之整合及測試，以期達到更優良的識別精度，本研究結果可以作為電子鼻系統之參考。

Aroma is one of the major concern for most coffee fans when they enjoy and choose coffee. Therefore, the aroma of fresh coffee is very important. Nevertheless, aroma evaluation is mostly relying on well-trained tasters.
The thesis aims at using intelligent olfactory sensor arrays to identify aromas of the coffee for single varieties via Machine Learning and Deep Learning.
The intelligent olfactory sensor array is a Neural Network that imitates the human sense of smell. Artificial intelligence is required to analyze odors of fresh coffee. Application methods are listed as follows: 1. Machine Learning Algorithms (Decision Tree, k-Nearest Neighbor method, Support Vector Machine, Random Forest) 2. Deep Learning Algorithm (Convolutional Neural Network). Via three artificial intelligence algorithms to classify gas data and to compare accuracy. In nowadays, Deep Learning algorithms have been used for odor classification due to the rise of deep networks. Gas sensors response waveforms have been transformed into gray images but had been short of pattern recognition of aromas. In view of the above problems, to improve the olfaction data visualization of aroma is the key to enhance the quality of recognition accuracy effectively.
This research shows the feasibility of applying CNN in coffee odor classification problem. CNN used as classifier for coffee odor classification, can be tested with more extensive odors sensors and data collectors in order to achieve better accuracy. The research results can be used as a reference for E-Nose system.

一、 緒論
1.1 簡介
1.2 研究背景
1.3 研究動機與目的
1.4 研究步驟
二、 文獻探討
2.1 機器學習與深度學習簡介
2.2 演算法應用於電子鼻的文獻
2.3 CNN演算法應用在氣味辨識
三、 研究方法與系統架構
3.1 研究架構
3.2 實驗環境
3.3 電子鼻系統
3.4 數據採集
3.5 數據處理
四、 深度學習演算法應用於氣味分類實驗
4.1 數據轉電子圖像資料集
4.2 實驗
五、 機器學習演算法應用於氣味分類實驗
5.1決策樹（Decision Trees）
5.2 k近鄰法（k-Nearest Neighbors, KNN）
5.3支援向量機（Support Vector Machines, SVM）
5.4 隨機森林（Random Forests）
六、 研究結果與系統評估
6.1 不同產區氣味辨識結果
6.2 分類演算法系統評估
七、 結論與建議
7.1 結論
7.2 建議

[1] 吳仁彰，2003，“電子鼻技術簡介”，科儀新知，133。
[2] R. Gutierrez-Osuna, H. T. Nagle, 1999, “A method for evaluating data-preprocessing techniques for odor classification with an array of gas sensors ” , IEEE Transactions on Systems, Man, and Cybernetics, Vol. 29, No. 5, pp.626-632, October.
[3] M. Pardo, G. Faglia, G.Sberveglieri, L. Quercia, 2001, “Electronic nose for coffee quality control ”, 18th IEEE Instrumentation and Measurement Technology Conference, Budapest、Hungary, 21-23 May.
[4] Treenet Thepudom, Nuttawoot Sricharoenchai, Teerakiat Kerdcharoen, 2013, “Classification of instant coffee odors by electronic nose toward quality control of production ” , 10th International Conference on Electrical Engineering/Electronics Computer Telecommunications and Information Technolog, Krabi、Thailand, 15-17 May.
[5] Kazimierz Brudzewski, Stanislaw Osowski, Anna Dwulit, 2012, “Recognition of Coffee Using Differential Electronic Nose ” , IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, Vol. 61, No. 6, pp.1803-1810, December.
[6] Dike Bayu Magfira, Riyanarto Sarno, 2018, “Classification of Arabica and Robusta coffee using electronic nose ” , International Conference on Information and Communications Technology (ICOIACT), Yogyakarta、 Indonesia, 6-7 March, pp.645-650.
[7] Whilly Harsono, Riyanarto Sarno, Shoffi lzza Sabilla, 2020, “Recognition of Original Arabica Civet Coffee based on Odor using Electronic Nose and Machine Learning”, International Seminar on Application for Technology of Information and Communication, Semarang、Indonesia, 19-20 September, pp.333-339.
[8] Gwo-Jia Jong, Hendrick, Zhi-Hao Wang, Kai-Sheng Hsieh, Gwo-Jiun Horng, 2006, “Identification of typical wine aromas by means of an electronic nose ” , IEEE Sensors Journal, Vol. 19, No. 22, pp.10796-10803, November.
[9] 王致乂，2014，應用類神經網路於水果氣味分類之研究，實踐大學，碩士論文。
[10] 許婷琬，2008，利用電子鼻及電子舌辨別蜂蜜真假之研究，中臺科技大學，碩士論文。
[11] 莊雅婷，2008，台灣特色茶感官特性與電子舌及電子鼻分析之相關性，中臺科技大學，碩士論文。
[12] Boonyawee Grodniyomchai, Khattiya Chalapat, Kulsawasd Jitajornwanich, Saichon Jaiyen, 2019, “A Deep Learning Model for odor Classification Using Deep Neural Network”, 5th International Conference on Engineering, Applied Sciences and Technology (ICEAST), Luang Prabang、Laos, 2-5 July, pp.333-339.
[13] 薛丁仁，蕭文澤，2019，“半導體式晶片型氣體感測器研發”，科儀新知，218。
[14] 古明地正俊，長谷佳明，2018，AI人工智慧的現在．未來進行式，林仁惠譯，遠流，臺北市。
[15] Giuseppe Bonaccorso, 2017，初探機器學習演算法，賴屹民譯，碁峰資訊，臺北市。
[16] Chollet, Francois, 2018, Deep learning with Python, Manning, New York.
[17] 黃彥臻，2009，“電子鼻應用趨勢與廠商分析”，工研院IEK，8月21日。
[18] 韓懷宗，2008，咖啡學：祕史、精品豆與烘焙入門，一版，時周文化，臺北市。
[19] Choi Nak Eon, 2017，咖啡香味的科學，謝雅玉譯，方言文化，臺北
市。
[20] Oded Maimon, Lior Rokach, 2005, Data Mining and Knowledge Discovery Handbook, Tel-Aviv University, Israel.
[21] Hossein Rezaei Estakhroueiyeh, Esmat Rashedi, 2015, “Detecting moldy Bread using an E-nose and the KNN classifier ”, 5th International Conference on Computer and Knowledge Engineering (ICCKE), Mashhad、Iran, 29-29 October.
[22] Liang Xie, Xiaodong Wang, 2009, “Gas quantitative analysis with support vector machine ” , Chinese Control and Decision Conference, Guilin、China, 17-19 June.
[23] Syuan-He Wang, Ting-I Chou, Shih-Wen Chiu, Kea-Tiong Tang, 2021, “Using a Hybrid Deep Neural Network for Gas Classification ” , IEEE Sensors Journal, Vol. 21, No. 5, pp.6401-6407, March.
[24] Matteo Pardo, Giorgio Sberveglieri, 2002, “Coffee Analysis With an Electronic Nose ” , IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, Vol. 51, No. 6, pp.1334-1339, December.
[25] Manojkumar Kukade, Tanay Karve, Damayanti Gharpure, 2019, “Identification and Classification of Spices by Machine Learning”, IEEE International Conference on Intelligent Systems and Green Technology (ICISGT), Visakhapatnam、India, 29-30 June.
[26] Long Li, Hua-Yao Li, Zhixiang Hu, Jingyao Liu, Licheng Zhou, Zhilai Tian, Huan Liu, 2019, “Construction and Simple Application of a Primary Electronic Nose System ”, IEEE International Symposium on Olfaction and Electronic Nose (ISOEN), Fukuoka、Japan, 26-29 May.
[27] Huixiang Liu, Dongbing Yu, Yu Gu, 2019, “Classification and Evaluation of Quality Grades of Organic Green Teas Using an Electronic Nose Based on Machine Learning Algorithms ” , IEEE Access, Vol. 7, pp.172965-172973, lDecember.
[28] Yu Thazin, Theerapat Pobkrut, Teerakiat Kerdcharoen, 2018, “Predication of Acidity Levels of Fresh Roasted Coffees Using E-nose and Artificial Neural Network”, 10th International Conference on Knowledge and Smart Technology (KST), Chiang Mai、Thailand, 31 Jan-3 February, pp.210-125.
[29] Pei-Feng Qi, Qing-Hao Meng, Ming Zeng, 2017, “A CNN-based simplified data processing method for electronic noses ” , ISOCS/IEEE International Symposium on Olfaction and Electronic Nose (ISOEN), Montreal QC、Canada, 28-31 May.
[30] Gwo-Jia Jong, Hendrick, Zhi-Hao Wang, Kai-Sheng Hsieh, Gwo-Jiun Horng, 2019, “A Novel Feature Extraction Method an Electronic Nose for Aroma Classification ” , IEEE Sensors Journal, Vol. 6, No. 1, pp.173-178, January.
[31] Alin Tisan, Marcian Cirstea, Stefan Oniga, Attila Buchman, 2010, “Artificial olfaction system with hardware on-chip learning neural networks ” , 12th International Conference on Optimization of Electrical and Electronic Equipment, pp.884-889.
[32] Sigeru Omatu, Yoshinori Ikeda, Mitsuaki Yano, 2015, “Odor measurement and intelligent classification ”, 10th Asian Control Conference (ASCC), Kota Kinabalu、Malaysia, 31 May-3 June.
[33] Dongbing Yu, Xiaoran Wang, Huixiang Liu, Yu Gu, 2019, “A Multitask Learning Framework for Multi-Property Detection of Wine ” , IEEE Access, Vol. 7, pp.123151-123157, August.
[34] Danli Wu, Dehan Luo, Kin-Yeung Wong, Kevin Hung, 2019, “POP-CNN: Predicting Odor Pleasantness With Convolutional Neural Network ” , IEEE Sensors Journal, Vol. 19, No.23, pp.11337-11345, December.
[35] Y. Lecun, L. Bottou, Y. Bengio, P. Haffner, 1998, “Gradient-based Learning applied to document recognition”, Proceedings of the IEEE, Vol. 86, No. 11, pp.2278-2324, November.
[36] Ekachai Phaisangittisagul, H. Troy Nagle, 2008, “Sensor Selection for Machine Olfaction Based on Transient Feature Extraction ” , IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, Vol. 57, No. 2, pp.369-378, January.
[37] M. Pardo, G. Sberveglieri, B. Sisk, N. Lewis, 2003, “Classifier comparison and sensor selection for e-noses ”, SENSORS IEEE, Toronto ON、Canada, 22-24 October, pp.606-610.
[38] 謝孟弦，2018，“氣體感測器-打造電子鼻未來應用情境”，工研院IEK， 7月26日。
[39] Sebastian Raschka，2016，Python機器學習，劉立民，吳建華譯，博碩文化，新北市。
[40] 鄒慶士，2019，大數據分析與應用實戰，一版，鄒慶士，臺北市。
[41] Giancarlo Zaccone，2017，深度學習快速入門：使用TensorFlow，傅運文譯，博碩文化，新北市。
[42] 劉凡平，2017，大數據時代的演算法，松崗，臺北市。
[43] 雷祖強，周天穎，萬絢，楊龍士，許晉嘉，2007，“空間特徵分類器支援向量機之研究”，航測及遙測學刊，12卷，2期，頁145-163， 6月。
[44] Ian Goodfellow, Yoshua Bengio, Aaron Courville, 2016, Deep Learning (Adaptive Computation and Machine Learning Series). The MIT Press, United States.
[45] Vinod Nair, Geoffrey E. Hinton, 2010, “Rectified Linear Units Improve Restricted Boltzmann Machines” , Proceedings of the 27th International Conference on Machine Learning Haifa, Israel, pp. 807– 814.
[46] Benjamin Sanchez-Lengeling, Jennifer N Wei, Brian K Lee, Richard C Gerkin, Alán Aspuru-Guzik, Alexander B Wiltschko, 2019,” Machine Learning for Scent: Learning Generalizable Perceptual Representations of Small Molecules”, arXiv preprint arXiv:1910.10685, October.