Character recognition from handwritten images is of great interest in the pattern recognition research community for their good application in many areas. To implement the system, it requires two steps, viz., feature extraction followed by character recognition based on any classification algorithm. A convolutional neural network (CNN) is an excellent feature of extractor and classifier. It is having multiple applications fields such as robotics, medicine, and security and surveillance. In this article, CNN is implemented for the NIST dataset with appropriate parameters for training and testing the system. The system provides accuracy of up to 94%, which is better with respect to others. It also takes a very low amount of time for training the system.