An automated teller machine includes a camera, a memory, and a hardware processor. The camera captures video data from a user, including at least one hand movement. The processor receives the video data from the camera and splits it into a sequence of images. The processor then splits each image into a set of features and forms vectors from the features. The processor uses the vectors to determine if each image belongs to a subset of images corresponding to a motionless sign-language pattern, or to a subset of images corresponding to a moving sign-language pattern. The processor stores the words and/or phrases assigned to each identified sign-language pattern in an input phrase. The processor determines that the input phrase represents a user request and then processes the request.