No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
IMPROVING CNN FEATURES FOR FACIAL EXPRESSION RECOGNITION
Abstract
span class="fontstyle0">Abstract Facial expression recognition is one of the challenging tasks in computer
vision. In this paper, we analyzed and improved the performances both
handcrafted features and deep features extracted by Convolutional Neural
Network (CNN). Eigenfaces, HOG, Dense-SIFT were used as handcrafted features.
Additionally, we developed features based on the distances between facial
landmarks and SIFT descriptors around the centroids of the facial landmarks,
leading to a better performance than Dense-SIFT. We achieved 68.34 % accuracy
with a CNN model trained from scratch. By combining CNN features with
handcrafted features, we achieved 69.54 % test accuracy.
Key Word : Neural network, facial expression recognition, handcrafted features
<br /
Recognition of human emotions by facial expressions is an important research problem that covers many areas and disciplines, such as computer vision, artificial intelligence, medicine, psychology and security. This paper provides an analytical overview of video facial expression databases and approaches to recognition emotions by facial expressions, which include three main stages of image analysis, such as pre-processing, feature extraction and classification. The paper presents both traditional approaches to recognition of human emotions by visual facial features, and approaches
based on deep learning using deep neural networks. We give the current results of some existing algorithms. In the review of scientific and technical literature we empathized mainly the sources containing theoretical and research information of the methods under consideration, as well as comparison of traditional methods and methods based on deep neural networks, which were supported by experimental studies. Analysis of scientific and technical literature describing methods and algorithms for study and recognition of facial expressions, as well as the results of world scientific research,
have shown that traditional methods for classification of facial expressions are second in speed and accuracy to artificial neural networks. The main contribution of this review is providing a common understanding of modern approaches to recognition of facial expressions, which will enable new researchers to understand the main components and trends in the field of recognition of facial expressions. Moreover, comparison of world scientific findings has shown that a combination of traditional approaches and approaches based on deep neural networks achieves better classification accuracy, but artificial neural networks are the best classification methods. The paper may be useful to specialists and
researchers in the field of computer vision.
ResearchGate has not been able to resolve any references for this publication.