Object recognition from human tactile image using artificial neural network

摘要

This paper presents a recognition of images objects that are out of touch by capturing the texture of objects that existed in everyday life which sub-divided according to the different parts of the human hand palm. They are divided into sections based on the physiology of the human hand by dividing the exposure of 15, 20 and 26 proportions and then analyzed. This work is the basis design of human-like hand for a robot to work effectively. These pictures of touch present the genuine sensory system of a human hand. They will be processed to characterize the touched object as the different area for the exposed surface of the object will create different for the palm's pressure. After that, we extract the features of the pressure from pressing the object and use an artificial neural network to distinguish different types of objects. The results of tests on 15 types of objects show the accurate result of the analysis of at the average maximum of 78.38% with the 26 proportions palm .This research will form the basis design of human-like hand for a robot that can recognize objects robot caught.