A face-house paradigm for architectural scene analysis

Abstract

This interdisciplinary study proposes a method for architectural design analysis of house facades which is based on face detection and facial expression classification. The hypothesis is that abstract face expression features can occur in the architectural design of house facades and will potentially trigger emotional responses of observers. The approach used statistical learning with support vector machines for classification. In the computer experiments the system was trained using a specifically composed image data base consisting of human faces and smileys. Afterwards it was applied to a series of test images of human facial expressions and house facades. The experiments show how facial expression pattern associated with emotional states such as surprise, fear, happiness, sadness, anger, disgust, contempt or neutral could be recognised in both image data sets.