Research Area: Machine Learning
Facial expression recognition (FER) provides an effective way to recognize human emotions. Since, the facial expression shows the internal feelings of a person; therefore, it can be used to understand human behavior from image sequences. Wavelets and deep learning have been effectively and extensively used separately in many computer vision applications. Being motivated from wavelets and deep learning, we introduce an integration of convolutional neural network (CNN) and discrete wavelet transform (DWT) for FER. The proposed FER framework consists of three key modules, namely, face processing, feature representation using DWT and expression recognition using CNN. We have used benchmark CK+ dataset for experiments and evaluated the performance of the proposed method in terms of recognition accuracy. The developed framework has been compared with existing FER methods and shows its effectiveness with an accuracy of 98.73%.
Keywords:
Deep Learning
Facial Expression Recognition
Wavelet Domain
discrete wavelet transform (DWT)
Author(s) Name: Pooja Sharma & Rajiv Singh
Journal name:
Conferrence name: Soft Computing and Signal Processing
Publisher name: Springer
DOI: 10.1007/978-981-16-1249-7_10
Volume Information: pp 91–100
