Sign Language Detection Through PCANet and SVM

The simplest and most process improvement for hearing and deaf persons to communicate with one another is sign language. It is difficult to identify the alphabet in American Sign Language (ASL) using a marker-less vision sensor. Using photos obtained from the color images, this model presents a novel user-independent identification approach for the American Sign Language alphabet. Convolutional neural network designs are utilized for feature learning, as opposed to the conventional methods. Local characteristics retrieved from the segment of hand are effectively trained with a straightforward PCANet deep learning architecture. Building a unique PCANet model for each user and training a single PCANet model utilizing data of all users are the two ways suggested for learning the PCANet model. Following feature extraction, a linear Support Vector Machine (SVM) is used to determine the classes. The efficiency of the planned strategy is estimated using a dataset of actual color images gathered from many users and made available to the public. Trial results show that the suggested strategy achieves best detection performance with maximum accuracy.