即時臉部表情辨識與學習回饋系統

本研究設計一套「即時臉部表情辨識學習回饋系統」，此系統能夠即時地分析學生的臉部表情並判斷學生的學習情緒，當學習情緒處於負面時，系統主動地詢問學生的想法與感受，接著通知授課老師以便關心學生的學習狀況，達到即時的補救教學以及在課堂間關心學生，幫助學生解決學習上的問題。本研究設計一個名為「深度可分離之密集擠壓激發卷積神經網路」的臉部表情辨識模型，使用網路公開的FER2013資料集，將28,709張圖片作為訓練資料以及3,589張圖片作為驗證資料，用於訓練並驗證本研究設計的模型，測試資料為即時的學生臉部擷取畫面。本研究使用NVIDIA開發套件實作系統，傳統的桌上型電腦占用教學空間且不易攜帶，本研究開發的辨識系統成本較低，且辨識準確度高於現有的辨識模型，透過即時的臉部表情辨識搭配本研究設計的學生回饋介面，授課老師可以更專心於教學，而學生也能透過系統的輔助即時回饋學習狀況。

關鍵字

臉部表情辨識；線上回饋系統；學習情緒；人工智慧

並列摘要

The paper aims to develop a real-time facial expression recognition and online feedback system. The designed system instantaneously analyzes and recognizes student's learning emotion. When student's emotion is negative, the system actively confirms with the student and simultaneously informs the teacher of learning problem. The proposed system is capable of achieving remedial teaching and caring students so that the learning problem of students can be solved. The paper proposed a Depthwise Separable-Dense-Squeeze-and-Excitation Network, i.e. DS-Dense-SENet. The proposed DS-Dense-SENet is trained by the Facial Expression Recognition on FER2013 dataset, and captures student's real-time facial expression as being test data. In addition, the proposed DS-Dense-SENet is implemented on a cost-effective development board, viz NVIDIA Jetson Nano. Experimental results showed that the proposed system not only achieves real-time facial recognition and feedback but also yields higher testing accuracy. By using the system designed in this paper, teachers are able to focus on teaching and students can notify problems as usual.

並列關鍵字

Facial Expression Recognition ； Online Feedback System ； Learning Emotion ； Artificial Intelligence

國際替代計量

即時臉部表情辨識與學習回饋系統

全文下載

主題瀏覽