Innovative Currency Identifier for the Blind through Audio Output using Deep Learning
Rajasri Maganti, Jessica Gutla, Yaswanth Mallimpalli, Sandeep Yelisetti · 2024
With an estimated 4.95 million blind people and 70 million visually impaired individuals in India are facing challenges in their day-to-day monetary transactions, worsened by the complexities in identifying tactile features of notes post-demonetization, this research presents an approach to enhance independence and accessibility. The introduction of the Currency Identification System (CIS), which is specifically designed for Indian cash and is accomplished via an android application with a camera interface for voice output and recording. We examine the shortcomings of current systems while highlighting the particular difficulties experienced by the blind people in our society. As a result, we have developed a novel CIS that combines two deep learning algorithms: Convolutional Neural Networks (CNN) and Residual Networks (ResNet). We trained a model using these two algorithms and compared the algorithms to use the best model for our application. The outcomes demonstrate the ResNet’s exceptional performance, with an 90% accuracy rate. By demonstrating the effectiveness of deep learning models in these kinds of applications, our research adds to the growing field of assistive technology. With the help of our research, we shed light on the ways in which technology can promote independence and inclusion in the lives of people, which is a critical step in the direction of increased financial accessibility and control.