Research Article
Lower Order Krawtchouk Moment-Based Feature-Set for Hand Gesture Recognition
Table 1
Comparison of recognition accuracies of Hu moments, Zernike moments, and Krawtchouk moment.
| Feature set (up to 3rd order) | -NN using ED () | -NN using MD () | SVM | MLP | ELM | Poly | RBF | PUK | Poly | RBF | Linear |
| Raw feature-set for ISL database | Hu moments (7 features) | 66.9% | 67.8% | 67.9% | 70% | 71.1% | 70.2% | 70.2% | 69.2% | 70% | Zernike moments (6 features) | 77.6% | 74.3% | 73.5% | 76% | 78% | 75.1% | 76.2% | 76% | 75.4% | Krawtchouk moments (9 features) | 84.9% | 85.3% | 82.9% | 83.5% | 84.9% | 84.5% | 66.2% | 65.4% | 66.1% |
| Normalized feature-set for ISL database | Hu moments (7 features) | 68% | 68.3% | 70.2% | 70.4% | 71.9% | 70.1% | 70.8% | 70.2% | 70.2% | Zernike moments (6 features) | 78.2% | 75.7% | 79% | 79.4% | 79.6% | 78% | 79.7% | 79% | 78.4% | Krawtchouk moments (9 features) | 85.5% | 85.9% | 85.9% | 84.8% | 85.9% | 85.2% | 70.7% | 70.2% | 70.1% |
|
|