Hand Gesture Recognition System for Touch-less Car Interface Using Multiclass Support Vector Machine

Touch-less car interfaces are getting the attention in automobile industries. By using the hand gestures it is possible to control some activities of cars, for this effective recognition system for hand gesture is required. This paper presents the recognition system of hand gesture for touch-less car interface. This system accepts video frames sequence and segmentation is applies on theses frames. Here the new segmentation method is applied by detecting skin portion using HSV, YCgCr and YCbCr color space. From the segmented images their color and edge characteristics are extracted and then stored in the database with their respective labels. Edge histogram descriptor is used for retrieving the shape characteristics from images where a colour structure descriptor (CSD) is used to capture the spatial distribution of color in an image. Then for identification of gesture multiclass SVM classifier is used. For the experiment, hand gesture database from Cambridge is used where video based frames of images are present for 9 gestures.