Tactile Pattern Recognition Based on PCA and BP Neural Network

To improve the efficiency of communication in human-robot cooperation through tactile information, this paper proposes a method to recognize human intedend direction in 2-D using an equipment with tactile arrays. The PCA method is employed in this study to extract essential information thus reduse computation complexit and increase robustness. BP neural network is implemeted for calssifying the intedend direction of human operators. Three members of the project team were involved in the study. The efficicency of proposed algorithm is investigated. Experimental results shows that the proposed methed chold achieve 93.1% recognition accuracy if both the training data and validation data contain tactile images from all the users.