FaceRecGlasses: A Wearable System for Recognizing Self Facial Expressions Using Compact Wearable Cameras

Facial expression images are useful for life-logging because they represent one’s emotions and mental state. However, using a wearable system to capture facial expressions from the front of the user obstructs the field of view. In this paper, we propose FaceRecGlasses, a compact wearable system that constantly records the user’s face and surroundings as images and indexes them with emotions without obstructing the user’s daily life. Our eyeglass-shaped device comprises two compact cameras and three mirrors, which enable the user’s facial expressions and surroundings to be captured constantly. The proposed system outputs a real-time pseudo-face-image by combining the captured facial components with a stored base-face-image. We investigated suitable facial areas to be captured by the proposed system and evaluated the recognition accuracy of the facial expressions captured. The results confirmed accuracies of 87.5%, 66.7%, and 71.4% for neutral emotion, happiness, and surprise, respectively, indicating that this method reproduces these three facial expressions to a high degree and can be applied to indexing of facial images.