A new vision based finger writing character recognition system (FWCRS) is proposed in this paper. The FWCRS allows people to write characters virtually just using his finger-tip (we call this "fingerwriting"). The trajectories of the finger-tip are tracked and reconstructed as a kind of inkless character pattern and finally recognized by a classifier. In this paper, a simple but effective background model is built for the FWCRS to segment human finger from cluttered background. A robust fingertip detection algorithm based on feature matching is presented. The fingerwriting character is finally recognized by a DTW classifier. Experiments show that the FWCRS can recognize finger-writing uppercase & lowercase English characters with the accuracy of 95.6%, 98.5% respectively.