Recognition of Bengali sign language characters is crucial for facilitating communication for the deaf and hard-of-hearing population in Bengali-speaking regions, which encompass approximately 430 million people worldwide. Despite the significant number of individuals requiring this support, research on Bengali sign language character recognition remains underdeveloped. This article presents a novel approach to categorize Bengali sign language characters using the Ishara-Lipi dataset, based on convolutional neural networks (CNNs) and pretrained models. We evaluated our approach using metrics such as accuracy, precision, recall, F1-score, and confusion matrices. Our findings indicate that the CNN model achieved the highest performance with an accuracy of 98%, followed by VGG19 with 94% and ResNet variants achieving around 88%. The proposed model demonstrates robust and efficient classification capabilities, significantly bridging the gap in existing literature. This study holds substantial promise for enhancing assistive technology, thereby improving social inclusion and quality of life for Bengalispeaking deaf and hard-of-hearing individuals.