Asl Letters Based On Monocular Vision Gesture Detection And Recognition Technology Research

Hand gesture recognition technique is an emerging HCI trend nowadays. The studies are on a structured set of hand gestures in this paper, which is on the ASL alphabet spelling gestures and its detection and recognition under the system based on monocular vision. The two key points on gestures detection and gestures recognition are addressed step by step in this paper. The detection based on skin color and the detection based on hand motion is discussed separately in the hand gesture detection part. Through the skin-color-based detection, a parametric model of skin color GMMs is used to describe the skin color in the different chrominance space and the adaptive threshold method is modified to obtain the detection of hand skin color; as well as the hand-motion-based detection, the hand shape is captured by the background subtraction detection based on background GMMs. Combining with the results of two parallel hand detection, the approach based on connected component analysis is adopted to segment the sign gestures. During the recognition process, the preprocessing for the ROI of hand gestures is involved with forearm removing, gesture angle correction and size regulating. As for the gesture feature extraction, the shape feature of hand silhouette is gained by the MPP toward the gestures with fingers and the sample space dimension of hand gray templates is reduced by the PCA-based scheme for the similarity of hand silhouette between the gestures without fingers. The performance of the ASL alphabet spelling gestures recognition based on MPP feature extraction with RBF classifier and PCA feature extraction with RBF classifier are discussed at the end of this paper.