Finger Gestures Detection Using Convolution Neural Network for Playing Virtual Cricket

Abstract

This is an era of intelligent machines. With the advancement in artificial intelligence, machine learning, and deep learning, machines have started to impersonate humans. Gesture recognition is a hot topic in computer vision and pattern recognition. We wanted to develop a game with less complexity and in an interactive way so that users of any age group can appreciate it. In this paper, the proposed model uses convolutional neural networks (CNN) to recognize the hand gesture with accurate results. This process flow consists of placing your palm over the particular segment over the screen and finger recognition using CNN classifier. The fingers are recognized using the Convolution, Normalization, Activation, Max-pooling, and Dropout layers. In this paper, we have compared different models using various combinations of these layers. Based on the performance and complexity of these models, we have selected a model with higher performance and reasonable complexity which helped us to classify the image correctly. A chatbot is also integrated with the game that will help users to understand the rules of the game. Users need to ask a query to the bot. The Bot will understand the query and return the appropriate answer. A web page will be provided to the user where he/she can play the game and ask queries to the chatbot.