Due to the increasing demands for the symbiosis between human and robots, humanoid robots (HRs) are expected to offer the perceptual capabilities as analogous as human being. One challenge of HRs is its capability of communication with people. Recently, the author presented a novel system for the visual speech recognition. The visual recognition problem is central to computer vision research. This article views first algorithm for Arabic visual object speech recognition. Visual features are width height, marker distances, dct of lips .The Idea is to find the green markers (objects) around the lips to form a masking polygon that will crop out irrelevant areas of image after converts this image to labeled image. and these objects will track them to extract features. Image filtering techniques introduces many artifacts (extra pixels that are not makers). Missing marker positions due to aliasing in video, we will estimate position of missing markers. Algorithm work for all test and training sets, different lighting, framing conditions
