Linguistics

Philosophy

Deep belief network

Support vector machine

Computer science

Facial expression recognition

Supervised learning

Pattern recognition (psychology)

Feature selection

Feature (linguistics)

Artificial intelligence

Deep learning

Artificial neural network

Feature extraction

Convolutional neural network

Facial expression

Facial recognition system

Machine learning

Linear discriminant analysis

Neurocomputing

A semi-supervised emotion recognition algorithm using reduced features as well as a novel feature selection approach is proposed. The proposed algorithm consists of a cascaded structure where first a feature extraction is applied to the facial images, followed by a feature reduction. A semi-supervised training with all the available labeled and unlabeled data is applied to a Deep Belief Network (DBN). Feature selection is performed to eliminate those features that do not provide information, using a reconstruction error-based ranking. Results show that HOG features of mouth provide the best performance. The performance evaluation has been done between the semi-supervised approach using DBN and other supervised strategies such as Support Vector Machine (SVM) and Convolutional Neural Network (CNN). The results show that the semi-supervised approach has improved efficiency using the information contained in both labeled and unlabeled data. Different databases were used to validate the experiments and the application of Linear Discriminant Analysis (LDA) on the HOG features of mouth gave the highest recognition rate.

Semi-supervised facial expression recognition using reduced spatial features and Deep Belief Networks