A DEEP LEARNING APPROACH TO CURRENCY RECOGNITION FOR THE VISUALLY IMPAIRMENT PEOPLE

Abstract [English]

Currency recognition for blind people is an important issue as it allows them to independently manage their finances and enhance their daily lives. In this paper a novel system is proposed for currency recognition using computer vision and machine learning techniques. The system utilizes a camera and image processing algorithms to extract features from banknotes and classify them according to their denomination. Currency recognition technology can greatly benefit visually impaired individuals by providing them with the ability to manage their finances independently. Blind people often face challenges when it comes to managing their money, as they cannot distinguish between different denominations of banknotes. This technology can empower them to confidently identify and manage their money, enhancing their daily lives and improving their financial independence. In addition to helping blind people, currency recognition technology can also benefit other individuals and organizations, such as banks, retailers, and vending machine operators. These entities can use this technology to automate their processes and improve the efficiency of their operations. Convolutional Neural Network (CNN) is a deep learning algorithm that is widely used in image recognition and computer vision applications. CNN has been shown to achieve high accuracy in recognizing complex patterns and features in images, making it an ideal algorithm for currency recognition for blind people. To implement currency recognition using CNN, the system would first need to collect a large dataset of banknote images of different denominations. The CNN would then be trained on the preprocessed images, with the goal of learning the underlying patterns and features that differentiate one banknote denomination from another. The training process would involve forward and backward propagation of the data through the network, with the weights and biases of the filters being updated at each iteration to minimize the error between the predicted and actual denominations .