Fingertips detection and hand tracking based on curve fitting

摘要

Gesture recognition is important in the human-computer interaction. In this paper, a method for detecting fingertips and gesture recognition using Cubic B-Spline Curve Fitting is proposed. The main steps in the method include: skin segmentation, palm and fingertips extraction, gesture trajectory extraction. According to transform RGB color space to YCrCb, the region of hand can be extracted using OSTU algorithm. Then, the skeleton of the hand is extracted using distance transform method. The center of palm, radius of the palm and fingertips are extracted from the skeleton image. The trajectory of the hand can be extracted and processed using the Cubic B-Spline Curve Fitting method to match a curve. The experiment results show that the fingertips can be extracted accurately, and the hand trajectory can be described efficiently in our method.