Vision-based gesturerecognition is a key technique to achieve a new generation of human-computerinteraction. As few text input search system by gesture recognition isdeveloped, based on the existing gesture recognition techniques, we use thegestures which are corresponding to the Chinese letters and numbers as inputgesture and use Microsoft kinect to obtain depth image to conduct hand gesturesegmentation. First, the edge of the gesture is extracted by Canny algorithm,and then the feature is extracted based on wavelet moment. Finally the gestureletters are obtained. Achieved the text input system based on gesturerecognition. Experiments show that the system is able to achieve Chinesecharacters accurately and effectively.