Recognizing Gestures for Humanoid Robot Using Proto-Symbol Space

Abstract

This paper describes the non Verbal communication method for developing a gesture-based system using Mimesis model. The proposed method is applicable to any hand gesture represented by a multi-dimensional signal. The entire work concentrates mainly on hand gestures recognition. It develops a way to communicate between Humans and the Humanoid Robots through gestural medium. The Mimesis is the technique of performing human gestures through imitation, recognition and generation. Different Gestures are being converted into code words through the use of code book. These code words are then converted into Proto-Symbols, these proto symbol then forms basis for training of the Humanoid robot. The recognition part is performed through a “distance vector”, a novel algorithm developed by us which is a combination of Euclidean distance and K-nearest neighbor. The generation part is done through the use of WEBOTS which include use of Humanoid robot HOAP 2 having 25 degrees of freedom. All the process of training, recognition and generation are simulated through MATLAB.