Enhanced Convolutional Neural Network for Sign Character Recognition with Mixed Pooling and Mish Activation Function

Communication is an essential part of our daily lives, especially for deaf and dumb people who face difficulty in communicating with others. This paper introduces a method of using sign languages, i.e., to convert various hand gestures to appropriate text. The proposed method will help people with disabilities communicate or express their thoughts and opinions to others. By offering cutting-edge methods for sign character recognition, this neural network model aims to close the communication gap and enhance the quality of life for those it serves. In this work, a dataset of hand gestures is divided into training, validation, and testing sets to apply a convolutional neural network (CNN) that has been used to properly recognise the characters on the image. The CNN is further tested on random test data to obtain a reliable interpretation of the sign language we want. The uniqueness of this model lies in the fact that it makes use of mixed pooling operation and mish activation function which train the data more smoothly and in a strong, robust way, which overall improves the accuracy and precision in recognising sign characters.