Machine Learning for Face, Emotion, and Pain Recognition PDF

This Spotlight explains how to build an automated system for face, emotion, and pain recognition. The steps involved include pre-processing, face detection and segmentation, feature extraction, and finally recognition to classify features and show the accuracy of the system. State-of-the-art algorithms are used to describe all possible solutions of each step. For face detection and segmentation, several approaches are described to detect a face in images: Viola-Jones, color-based approaches, histogram-based approaches, and morphological operation. Local binary patterns, edge detectors, wavelets, discrete cosine transformation, Gabor filters, and fuzzified features are used for feature extraction. The last step includes three approaches for recognition: classification techniques (with a special focus on deep learning), statistical modeling, and distance/similarity measures.

Authors: Anbarjafari, Gholamreza; Rasti, Pejman; Noroozi, Fatemeh