Hybrid evolutionary neuro-fuzzy approach based on mutual adaptation for human gesture recognition

One of the most important techniques in human-robot communication is gesture recognition. If robots can read intentions from human gestures, the communication process will be smoother and more natural. Processing for gesture recognition typically consists of two parts: Feature extraction and gesture classification. In most works, these are independently designed and evaluated by their own criteria. This paper proposes a hybrid approach based on mutual adaptation for human gesture recognition. We use a neuro-fuzzy system (NFS) for the classification of human gesture and apply an evolution strategy for parameter tuning and pruning of membership functions. Experimental results indicate the effectiveness of mutual adaptation in terms of the generalization.