Spontaneous Facial Expression Analysis Using Optical Flow Technique

Abstract

Investigation of emotions manifested through facial expressions has valuable applications in predictive behavioural studies. A potential application may be to impart intelligence to surveillance systems such as Closed-Circuit Television (CCTV) systems for recognition of emotional facial expressions. A facial recognition program tailored to evaluating facial behaviour for real time application can be met if patterns of emotions can be detected. An exploratory analysis of optical flow data was conducted with an aim to detect patterns and trends to differentiate between the emotional facial expressions: amusement, sadness and fear from the frontal and profile facial orientations. Analysis was in the form of emotion maps constructed from feature vectors obtained by using the Lucas-Kanade implementation of optical flow. Classification of individual emotions showed recognition of amusement was much greater in comparison to the recognition of the negative emotions, sadness and fear. Recognition was not negatively affected using reduced set of feature vectors derived from the emotion maps. Further investigation is necessary to assess the utility of emotion maps to visualise feature representations of emotional expression.