Abstract: The written characteristic of Uyghur language has brought difficult broblems for handwritten character recognition in technology. This paper proposes a method of handwriting recognition of Urghur letters using BP neural network. According to the tracks of the script, the method extracts features of the letters, forms a coding with fourteen dimensions. Secondly, trains the BP neural network and studies using the coding, builds good system of recognition. Simulations show that this method enhanced the veracity of recognition.