In this paper, in order to compensate the blank of ethnic minorities off-line signature verification system in our country, firstly it is conducted a set of preprocessing steps such as, graying, binarization, smoothing, and normalization for the collected Uyghur language signature images having combined with the Uyghur language handwritten signature writing style and characteristics. Then, based on the traditional direction feature extraction method, the signature sample images were divided into a rectangular area of $2\times 4$ after preprocessing, and it is proposed modified 48 dimensional features based on the information in the signatures image pixels within each region and $0^\circ, 3 0^\circ, 6 0^\circ, 90^\circ, 120, 150^\circ$ direction. Finally, three kinds of distance classifiers were used and experimental results were analyzed and evaluated using 900 Uygur signature samples collected from 15 person (20samples/per person), and the total correct rate of signature reached 96.00%. This indicates that, for off line Uyghur handwritten signature verification, the modified 48 dimensional direction feature is an effective feature, which can better reflect the writing style and characteristics of Uyghur language handwriting signature.