A Gesture Recognition Method Based on YCbCr and SURF for Service Robot Interaction

With the increasing requirements of human-computer interaction experience, gesture recognition is used in many applications as a novel interaction method. However, the effects of complex backgrounds, occlusions and illumination pose difficulties for the successful recognition of gestures. For this problem, we propose a gesture recognition method based on the combination of SURF and YcbCr. First, RGB-D was calibrated, and then the collected hand images were processed for noise reduction. Secondly, the coarse recognition of gestures is completed by SURF feature point matching algorithm, and then the fine recognition of gestures is achieved by YCbCr skin tone segmentation algorithm. Finally, the gesture images are further processed by a combined morphological algorithm to improve the integrity of gesture recognition and reduce the influence of interfering factors. The experimental results show that the method can recognize gestures quickly and effectively with certain accuracy and robustness.