Background: From the early stages of script writing and reading, there was a big desire for researchers to make an aid device for the blind to read scripts without help of other. This paper introduced a new approach for recognition of Farsi scripts using principal component analysis (PCA). Methods: Materials used for this project were selected from ordinary books, magazines, newspapers, and printed documents. Character samples were selected from four fonts in different positions and three sizes resulted in total number of 20 recognition classes. The five methods used in character recognition were statistical method, Fast Zernike Wavelet Moments (FZWM) method, PCA, PCA with sample averaging, and PCA with eigenvectors averaging. Findings: The accuracy and speed of PCA in recognition of Farsi characters were respectively 1.775% and 7.5 times better than the statistical method. Likewise, it was 2.2% more accurate and 5.12 times faster than FZWM method. Conclusion: Using PCA with combinational averaging in samples and eigenvectors can be a novel method for recognition of Farsi characters.