With the development of Internet entertainment, more and more people have become video bloggers. The viewer’s evaluation is important to them. However, due to people’s preferences and emotional factors, it’s difficult to generally tell their true feeling about the videos. So people’s facial expressions when watching a video are the most intuitive and true evaluation of a video. Also, for applying this function, we only need to ask for the agreement of viewers. Compared with text comments, that will save their time and make people willing to help. Also, text comments are normally filled with bias. It doesn't give good feedback on the video quality to the creators. In this project, I have done a sequence of research considering emotion recognition technology. We will discuss some research problems such as how emotion recognition is different from other recognition technologies and how could we improve it and use it to benefit our life. Mainly go through the implementation and modification of the machine learning model step by step. Showing some convenient approaches to the readers. Helping to build up basic cognition of neural networks. In the end, I extended this project to simulate an application that makes live predictions of emotions and collected all the results in a period. Hope that can raise some interesting ideas for you.