基於臉部情緒辨識之喜好度分析

Preference Analysis Based on Facial Emotion Recognition

方瑋翔(WEI-XIANG FANG)

指導教授 ： 戴紹國

朝陽科技大學/資訊學院/資訊管理系/碩士(2020年)

摘要

當人們看到一件事物會對它產生不同程度的喜好感受，並且常常不自覺的會透過臉部的表情顯露出來。本研究將使用深度學習對人臉所表現出來的各種情緒進行分析，並且產生喜好度的估計值。透過喜好度的估計值，機器可以了解一樣物品，所帶給人們的是正面或負面的感覺，並且透過收集大量的喜好度資訊來提供機器進行更準確的決策分析。雖然目前已經有很多關於人臉情緒辨識的研究，然而這些研究都是針對單一影像的臉部來進行情緒辨識。但是喜好度感受的表現並不能只靠單一的臉部快情得知，而是要觀察一段時間的情緒變化才能確認。所以我們在這邊提出了一個喜好度分析架構，利用卷積神經網路分析一段連續的序列人臉影像，然後預測出使用者的喜好度。最後我們在實驗的部分計算出了容忍值為0.2可得到85%的準確度，這證明我們喜好度分析模組所得到的，喜好度能夠吻合受測者的喜好程度。

關鍵字

喜好度分析架構 ； 臉部情緒分析 ； 卷積神經網路

並列摘要

When people see things, people have different degrees of preference, and they often express it unconsciously through facial expressions. In this study, we will use convolutional neural network to analyze various emotions expressed by facial expressions and generate degrees of preference. Through degrees of preference, the machine can understand that things give people a sense of positive or negative. By collecting a large amount of preference information, the machine can provide a more accurate analysis method. Although there have been many researches on facial emotion recognition, these researches are based on the face of a single image for emotion recognition. However, the degree of preference cannot be known by a single facial expression only, but it can be confirmed by analyzing changes in facial expressions over a period of time. Therefore, here, we propose a preferred analysis framework that uses convolutional neural network to analyze a continuous sequence of facial images and then predict the user's first choice. Finally, we calculated a tolerance of 0.2 in the experimental part to achieve 85% accuracy, which proves that the substitution converted by our substitution analysis framework can match the substitution of the test object.

並列關鍵字

preference analysis framework ； facial expression analysis ； convolutional neural network

參考文獻

[1] A. Dapogny, K. Bailly and S. Dubuisson(2015), "Pairwise Conditional Random Forests for Facial Expression Recognition," 2015 IEEE International Conference on Computer Vision (ICCV), Santiago, pp. 3783-3791, doi: 10.1109/ICCV.2015.431.

[2] A. Dehghan, E. G. Ortiz, G. Shu and S. Z. Masood(2017), "DAGER: Deep Age, Gender and Emotion Recognition Using Convolutional Neural Network," arXiv preprint arXiv:1702.04280.

[3] A. Mollahosseini, D. Chan and M. H. Mahoor(2016), "Going deeper in facial expression recognition using deep neural networks," 2016 IEEE Winter Conference on Applications of Computer Vision (WACV), Lake Placid, NY, pp. 1-10, doi: 10.1109/WACV.2016.7477450.

[4] A. R. Biswas and R. Giaffreda, "IoT and cloud convergence: Opportunities and challenges," 2014 IEEE World Forum on Internet of Things (WF-IoT), Seoul, 2014, pp. 375-376, doi: 10.1109/WF-IoT.2014.6803194.

[5] B. Yang, J. Cao, R. Ni and Y. Zhang(2018), "Facial Expression Recognition Using Weighted Mixture Deep Neural Network Based on Double-Channel Facial Images," In IEEE Access, vol. 6, pp. 4630-4640, doi: 10.1109/ACCESS.2017.2784096.