Gesture Recognition System for User Interface Control

The paper presents a gesture recognition system for human-computer communication. The system is based on static gesture detection using a hybrid approach for skin detection, seed markers watershed segmentation and morphological operations, palm centre allocation by a combination of minimal covering circle centre and palm convexity defect points and classification of gestures using decision tree based on geometric skeleton and the palm circle. The gesture recognised is associated to a particular user interface command. The paper presents experimental evaluation of the suggested approach in terms of recognition accuracy under various lighting conditions and usage of different skin detection approaches. Keywords— gesture recognition, human-computer communication, palm detection, skin detection, user interface, watershed segmentation.