Ahmed K. Jheel & Kadhim M. Hashim
Lip Reading depends on watching the shape of speaker's mouth region ; Especially his lips to understand the letters and words that have been uttered through the movement of the lips, gestures and expressions on the face, the location of the lips and its extracted features. This paper presents a new system include some of stages applied to AV Letters 2 dataset. The first three stages are preprocessing,face\lip detection,lip localization based on a set of new approaches that aims to get region of interest (or lip region) accurately and clearly with the help of new techniques and other operation of image processing. This paper focus on the last two stages of the automatic lip reading system: features extraction and classification. There are three various sets of distinguished features as (SURF, Centroid, HOG) are used in the proposed system. These features give a set of points called key points are exactly located on the lip borders that have been tracked using Euclidean distance calculation. The classification performs using artificial neural network to be a new one of the visual speech recognition tools.