Implementation of interactive system with emotion recognition for videos

[[abstract]]人工智慧快速發展，各種人工智慧應用在人類生活，食、衣、住、行、育、樂各方面都與人工智慧息息相關，其中影像辨識也是人工智慧中重要的技術，常見應用包含人臉身份辨識、情緒辨識、辨識物體等，而近年來人們也在情緒辨識上做更多的運用。在辨識人類情緒的應用中，AI(Artificial Intelligence)在各方面應用也能提供豐富的回應，應用無所不在。本文提出一個可快速辨識人臉情緒的互動系統，透過網路攝影機擷取使用者的臉部影像，擷取49個臉部器官座標，以器官座標的變化量作為表情的特徵參數，並且以深度學習神經網路辨識情緒，辨認的情緒類別包括：憤怒、厭惡、恐懼、快樂、悲傷和驚訝、自然等七種情緒。為了有更好的辨識效果，本研究也使用多重特徵辨識作為輔助，使得情緒辨識率大幅提升，研究成果也整合成圖形化使用者介面，使得情緒辨識互動系統可以應用於實際的場域中。本系統亦能以進行延伸應用不同的場域，例如：打招呼系統、陪伴機器人、商業推薦等等，研究成果極具實用性。[[abstract]]Artificial intelligence applications are progressively developed in human life, including food, clothing, housing, transportation, education, and entertainment. Image recognition is an essential technology in artificial intelligence, including face recognition, emotion recognition, and object recognition. This thesis presents an interactive system that can quickly identify facial emotions. First, a user's facial image is captured through a web camera, and 49 facial organ coordinates are captured. The variation of the organ coordinates is used as the feature of the expression. So a deep-learning neural network can recognize emotions. The recognized categories contain seven emotions: anger, disgust, fear, happiness, sadness, surprise, and nature. This study uses multiple feature recognition as an aid to improve recognition performance. Hence, the research results are also integrated into a graphical user interface to apply the emotion recognition interactive system in practice. In addition, the proposed system can also be extended to various fields, such as greeting systems, companion robots, business recommendations, etc. Accordingly, the research results are practical