Robust Handwritten Digit Recognition System using Hybrid Artificial Neural Network on FPGA

Handwritten digit recognition (HDR) deals with variety of sources which includes papers, photographs, touch screens to recognize them into different categories. Deep Learning and Machine Learning algorithms are usually difficult to implement on FPGA because of their requirements in terms of power. In this work, handwritten number recognition for FPGAs method is employed using ANN networks. The proposed ANN’s correctness was determined by verification and comparison with numerous ANN networks, and the recognition rate is recorded to be 99.38%. The Xilinx Zybo is used to replicate the network. The utilization of FPGA resources was computed for various activation functions like ReLU, sigmoid size-5, sigmoid size-10 and the accuracy is retained. The parameters like power consumption and timing comparison are observed. The developed IP block occupies less of the total area and results in resource-saving and effective and making it appropriate for additional embedded applications.