Central aortic blood pressure is a vital sign to evaluate human's vascular condition, it is more accurate than ordinary brachial artery blood pressure. The tradition measuring method is depended on cardiac catheter invasion. The procedure is dangerous and invasive; therefore noninvasive central aortic blood pressure estimating is a key issue in medical diagnosis. The most important procedures of estimating are based on key points recognition of pulse wave. A strategy of recognition was proposed in the paper, it combines the wavelet transform and differential transformation to detect different key points of pulse wave. The third grade of Coif 1 wavelet transform was used to detect point c, and the rest of key points were searching by differential transformation by zero crossing method. From experiments, the method can detect the key points accurally.