Blood pressure (BP) is one of the indispensable elements of physiological health characteristics and a significant indicator for predicting and diagnosing hypertension and cardiovascular diseases. This paper proposes a two-domain fusion model to estimate BP continuously from pulse wave acquired with a pressure sensor. Method. The optimal external pressure applied on the pressure sensor is first determined in order to capture pulse wave in the radial artery. The captured pulse wave is then processed in both the time and frequency domains via filtering and fast Fourier transform. Finally, a set of features are extracted from these two domains and input into a neural network along with blood pressure values measured by a commercial sphygmomanometer for training. The model is then tested on new data for accuracy evaluation. Results. The proposed two-domain fusion method achieved a high degree of accuracy in measuring blood pressure.