Gesture Recognition by Capsule Networks and Beyond

There exist communication barriers between the deaf and listeners. Intelligent sign language translation (recognition and generation) systems will break these barriers and help the deaf to access public information easily in public scenarios such as conferences or in-court trials. Sign language recognition is a sub-domain of gesture recognition. This research is about gesture recognition by Capsule Networks (CapsNet), a novel deep learning technology. This research proposed CapsNet based approaches to deal with sign language recognition tasks which are suitable for daytime, nighttime and low-resolution condition. The proposed approaches demonstrated satisfactory performance by achieving state-of-the-art results on relative benchmark datasets.