The framework has been applied to recognize the twenty eight different signs as illustrated in the Fig. 1. The image sequences were obtained while subjects were performing hand signs in front of a video camera. Since each subject stands roughly the same position related to the camera, the variation of hand size in images is limited. Two different lighting conditions were used. In the current implementation, each hand sign was represented by five images sampled from the video. Figure 11 shows several examples of these sequences.
Figure 11:
Eight sample sequences.
From top to bottom, they represent the signs
``happy'', ``hot'', ``nothing'', ``parent'', ``pepper'',
``smart'', ``welcome'', and ``yes''.