Evaluation of Self-taught Learning-based Representations for Facial Emotion Recognition

TL;DR

This paper explores various self-taught learning strategies to generate diverse unsupervised representations for facial emotion recognition, demonstrating improved performance over existing methods.

Contribution

It introduces multiple strategies to create diverse unsupervised representations for FER, enhancing recognition accuracy compared to prior approaches.

Findings

Diverse representations improve FER accuracy

Proposed methods outperform state-of-the-art unsupervised approaches

Ensemble classifiers benefit from diverse feature sets

Abstract

This work describes different strategies to generate unsupervised representations obtained through the concept of self-taught learning for facial emotion recognition (FER). The idea is to create complementary representations promoting diversity by varying the autoencoders' initialization, architecture, and training data. SVM, Bagging, Random Forest, and a dynamic ensemble selection method are evaluated as final classification methods. Experimental results on Jaffe and Cohn-Kanade datasets using a leave-one-subject-out protocol show that FER methods based on the proposed diverse representations compare favorably against state-of-the-art approaches that also explore unsupervised feature learning.