透過您的圖書館登入
IP:18.118.200.86

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。讓企業則能透過喜好度的估計值得知當顧客看到商品時是正向或者是負向的感受,並且透過收集大量的喜好度資訊來提供企業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單靠某一瞬間的臉部分析得知,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,利用深度學習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與準確率來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自覺的喜好程度。

並列摘要


Facial expression will reveal the fancy degree when watching a thing. For example, when a customer views and admires an exhibition, his facial expressions are affected by the pleasant or disgusting feelings brought by the merchandise. This study uses deep learning to analyze the various emotions from facial expressed and generates estimates of the fancy-degree. Although there have been many studies on face emotion recognition, which can recognition the emotion by facial expression from a single image. However, the fancy-degree cannot learn from the facial analysis of a particular moment, but to observe the emotional changes for a period of time. Therefore, we propose a fancy-degree analysis framework to analyze a continuous sequence of face images by deep learning. The experiments demonstrate that the results generated by our module match the subject's conscious fancy-degree.

並列關鍵字

無資料

延伸閱讀