Real Time Classification of American Sign Language for Finger Spelling Purpose

Two viewpoints based real‐time recognition for hand gestures

“…2. This setup is an extension of the monocular vision setup [7], used for fingerspelling. A plain white reflective surface is selected as background to reduce efforts in image processing and to aid in self‐occlusion.…”

“…Foreground segmentation is used to detect the presence of gesture. Skin tone detection is realised by analysing conditional colour space formulae [7], followed by background segmentation via intensity enhancement. Fig.…”

“…The binary image pair has some edges in common which are used as a reference point in 3D view. Illumination correction of the image from previous work [7] is revisited to visualise the regions mostly affected due to lighting. This is achieved by applying the transformation matrix in (1) to image as it slides along a path traced by a Mobius strip (see Fig.…”

“…Highly elevated regions in gesture are highly reflective in Bob's viewpoint. Illumination distribution is achieved by Gaussian distribution of intensity [7] and value of multiplier C in (2), for the image pairs.…”

“…Fluctuating illumination is a challenging task, which affects the chromatic aspect of image [6]. Illumination distribution method can be used to minimise the disturbance caused by lighting [7]. Adequate lighting is also important in differentiating colours correctly.…”