Graph images, such as bar graphs are frequently displayed in technical documents to demonstrate results. Graphs play an essential role in understanding the information, results, and findings of such technical documents. It can be tough for the visually impaired people to understand the graphs as they are also the audience of such documents and if the graph is not described in the text, then there might be chances that the most visually impaired people miss the important information. However, scientific publications do not describe the graphs in textual form. There are many assistive technologies regarding text-to-speech exist but explaining a graph image via speech still lacks behind as screen or text reader software does not describe graph information. In this paper, we propose a fully automated system that extracts features from the bar graph images and then describes them using text-to-speech to the visually impaired audience.