The Role of Global and Feature Based Information in Gender Classification of Faces: A Comparison of Human Performance and Computational Models

Abstract

Most computational models for gender classification use global information (the full face image) giving equal weight to the whole face area irrespective of the importance of the internal features. Here, we use a global and feature based representation of face images that includes both global and featural information. We use dimensionality reduction techniques and a support vector machine classifier and show that this method performs better than either global or feature based representations alone. We also present results of human subjects performance on gender classification task and evaluate how the different dimensionality reduction techniques compare with human subjects performance. The results support the psychological plausibility of the global and feature based representation.

Investigations into the Robustness of Audio-Visual Gender Classification to Background Noise and Illumination Effects

“…Similar error rates of 6% were achieved using an SVM classifier following CCA and PCA; however, the former used 14 coefficients whilst the latter required 273. They then investigated the importance of local to global features for gender recognition, and discovered that a combination of both gave significantly better classification than either alone [6]. In addition, they showed association between the errors made by the computational models and those made by humans for local features such as eyes and mouths.…”

Face gender classification: A statistical study when neutral and distorted faces are combined for training and testing purposes

“…Since humans use both types of information when analysing faces, some authors have decided to use global (configural information) as well as local (featural information) descriptors assuming that it will ease the problem of classifying gender on automatic systems. Based on this idea, studies combining local and global features [17,18] conclude that using both types of features provides better face characterisations and hence better classification rates than using just one of them. It should be noted that these studies used occlusion-free face images.…”

The use of faces as stimuli in neuroimaging and psychological experiments: A procedure to standardize stimulus features

“…For example, some researchers have studied the developmental changes of face processing in children (e.g., Aylward et al, 2005), whereas others have addressed the question of how emotional expressions are recognized and processed by normal and disturbed individuals (e.g., Surguladze et al, 2006). The difference in processing of familiar and unfamiliar faces was investigated by, among others, Dubois et al (1999), and the effect of features or feature relationships on face perception was reported by, for example, Buchala, Davey, Frank, Loomes, and Gale (2005). A common factor in these studies was the presentation of facial stimuli to the participants.…”