Recognition of printed symbols is a subject of increasing interest to automatically convert scientific paper documents into electronic form. Several methods have been proposed for recognition of printed and handwritten character or symbols. In many application it is often necessary to copy the contents from original documents like PDF and while copying the data from that document ,when it encounters a special symbols that symbols remains unread in the copied document ,or instead of that symbol dash or underline would replaced in the destination document. So it is very difficult to read the original document. In research work, an attempt is made to recognize the symbols from English text to avoid errors while retrieving the contents. Our goal is to detect, extract and segment the different symbols from text and after extracting the DCT features, recognize it. This work recognizes the various symbols from the text image.