Effect of variation in gesticulation pattern in dynamic hand gesture recognition system

This paper presents a hand gesture recognition system which addresses the effect of variations in gesture pattern during gesticulation. Different gestures can be gesticulated in various patterns which increase the difficulties in recognizing the gestures. We have proposed two new features such as left sector trajectory features and right sector trajectory features which are able to recognize gestures even with the presence of variations in the gesticulation pattern. The effectiveness of the proposed system is illustrated by different experiments with our own gesture database. A comparative study has been made with the proposed features and three state-of-art features such as orientation; combination of location, orientation, velocity; and combination of ellipse and position features. The performance of the system was evaluated using this proposed set of features for different individual classifiers such as ANN, SVM, k-NN, Naïve Bayes and ELM. Finally, the decisions of the individual classifiers were combined using major voting rule to result in classifier fusion model. Based on the experimental results it may be concluded that classifier fusion provides satisfactory results compared to other individual classifiers. An accuracy of 91.07% was achieved using the classifier fusion technique as compared to baseline CRF (79.45%) and HCRF (83.07%) models.