Yi-na Jeong, Yong-bo Sim, Su-rak Son, Journal of Internet Computing and Services, Vol. 24, No. 3, pp. 43-50, Jun. 2023
10.7472/jksii.2023.24.3.43, Full Text:  HTML
Keywords: facial landmark, deeplearning, facial emotion classification

Abstract

Electronic devices are deeply entrenched in our daily lives and are used so much that they take up most of our time. Using these electronic devices to judge a person's psychological state and give feedback appropriate to the situation can help a lot in life. A method using CNN has been studied as an existing facial emotion classification technique, but this method has the disadvantage that it requires a lot of training data because it receives tens of thousands of pixels as input data. In this paper, to solve this problem, we propose a deep learning model DFLM (Deeplearning facial landmark model) that receives facial landmarks as input data and classifies facial emotions. The proposed DFLM can classify 7 facial expressions by constructing a fully connected neural network with 68 landmarks as inputs. As a result of the simulation, it is possible to learn faster than a CNN model, and its operation time is short, so it will be helpful in areas that need to make predictions in real time.

Statistics
Cite this article