Comparison of fourier descriptors and Hu moments for hand posture recognition

In this paper, we propose to use Fourier descriptors (FD) for hand posture recognition in a vision-based approach. FD are widely used for shape representation and pattern recognition, they may also be well-adapted for hand posture recognition. The invariance properties of FD are discussed, and we provide a comparison of the performances with Hu moments. First, experiments are performed on the Triesch hand posture database. Then we define our own gesture vocabulary, with 11 gestures, and we perform the acquisition of a large number of images, with 18 persons. Hence tests are performed on a more realistic database, with various hand configurations realized by non-expert users. Results show that FD give very good recognition rates in comparison with Hu moments. This confirms the efficiency of FD and shows their great robustness in real-life conditions.