Hybrid method for hand gesture recognition based on combination of Haar-like and HOG features

In this paper a new method is proposed for hand gesture recognition. The proposed method increases hand gesture recognition rate and decreases false positive error rate by using combination of Haar-like and Histogram of Oriented Gradients (HOG) features. Also some new Haar-like features are proposed proportional to hand posture to solve major Haar-like problem that is high false positive error rate in hand posture recognition. These features improve recognition rate to 83%. The experiments showed that hybrid method can recognize hand gesture by 93.5% accuracy which is 25% higher than previous method, and decrease the false positive error from 92% to 8%.