FPGA-based Neural Network for Arabic and English Handwritten Digit Recognition

In recent years, neural networks have been extensively used for handwritten digit recognition. However, limitations in computational efficiency present challenges for real-time applications. This work introduces a neural network model to classify Arabic and English handwritten digits, achieving an accuracy of up to 96.6%. This FPGA-based implementation demonstrates significant performance improvements, achieving a speedup of 14x to 49x over the proposed software, with a power consumption of 1.67 W and latency of 2.395 µs for processing an image of 2.39 Kbytes.