In the era of informatization, handwritten digit recognition has wide applications in computer vision and pattern recognition. In order to improve the accuracy and efficiency of handwritten digit recognition, four convolutional neural network (CNN)-based algorithm models were designed and tested. An empirical study was conducted on the MNIST dataset to compare the training performance of different models. The experimental results show that the multilayer CNN model performs best, with an accuracy rate of 98.9%, and each training cycle takes about 20 seconds. This indicates that increasing the number of convolutional layers and using high-level APIs help improve recognition accuracy, further enhancing the application performance of CNN in handwritten digit recognition. New insights and methods are provided for designing efficient and accurate handwritten digit recognition models.

convolutional neural network  /  handwritten digit recognition  /  MNIST dataset  /  deep learning