Fuzzy Logic Classifier and Conditional Responses Algorithm for Gestural Input Game

Abstract

This paper presents hand gesture classification using fuzzy logic, with an application to the Rock-Paper-Scissors game. It includes a comparison between a simple threshold classifier and the Fuzzy Logic classifier. The real-time gesture recognition is implemented for the Leap Motion input device. Although the human’s game gesture pattern may seem random, some recent work suggests how it is possible to predict an opponent’s gesture, from a previous gesture, with better-than-random success. Based on this, we designed a conditional-response algorithm making the behavior of the virtual player more similar to a real human. Using networking tools, we enable game play remotely against other players. The game was experimentally evaluated with eleven players, reaching an average online classification accuracy of 97.35% for hand pattern recognition. The evaluation metrics are represented in confusion matrices.