Abstarct: OCR is an optical text recognition system that has been widely used in different applications. The main focus of existing OCR systems has been on Latin languages. In recent studies, OCR systems, used for language with cursive style have been introduced, followed by some challenges. In this thesis, we proposed an LSTM-baxsed OCR system for the Persian language. We investigated some involved parameters in the proposed system. The proposed OCR system addressed the challenge false recognition of sub-word “LA” and “Zero-width non-breaking space”. In addition, we presented a preprocessing algorithm to remove “justification” using an image-processing-baxsed technique. Moreover, in this thesis, a new dataset was generated, comprising of five million images with 8 popular Persian fonts and in 10 various font sizes. The experimental results with our generated dataset shows that the accuracy of the proposed OCR is increased by 2%, compared to the Persian OCR currently available in Tesseract. Moreover, the proposed system has an accuracy of 99.688 at the letter level.