A Comparative Analysis for Measuring Accuracy in Recognizing Hand Gestures using Feedforward Neural Network and CNN Method

Abstract

Aim: The motive of the work is to accurately and precisely recognize the hand gestures and to use them for conversion to text and developing a media playback and snake game. Materials and Methods:ASL Alphabet dataset having 29 hand gestures with 26 alphabets and 3 words are used to train and test the hand gesture recognition framework. The proposed model can perceive hand movements captured through a camera in real-time into text, media playback controller and snake game. The intent is realized by databases with the help of two groups namely Convolutional Neural Networks (CNN) and Feedforward Neural Networks (FNN) to identify the more relevant Movement Hand Images(MHI) and it’s approach to develop text, a media player and snake game. The sample size is estimated as 20 using G Power.. Results and Discussion: Test results exhibited the prediction accuracy of the model with FFN (84.1%) and CNN (77.8%).There exists a statistical significant difference among the study groups with significance value (p<0.05). Conclusion: Hand gesture Recognition framework with FFN delivered better results compared to CNN in terms of performance measures, less data loss and accuracy in recognizing hand gestures, then conversion to Text, development of media playback and a snake game.