Gas-liquid two-phase flow is widely encountered in industrial processes, and accurate flow pattern recognition is of great significance to process monitoring and control. The traditional methods of flow pattern recognition are usually based on extracting a single sensor signal feature, and it is difficult to fully describe the flow patterns' characteristics. In this paper, a method to fuse differential pressure signals and pulsed wave ultrasonic echoes for flow pattern recognition is presented. Empirical modal decomposition of differential pressure signals to extract signal energies in different frequency ranges to characterize the gas phase structure, and gas-liquid interface information is obtained based on the energy distribution of pulsed-wave ultrasonic echoes. The results show that the fusion of these two sensors can identify flow pattern with an accuracy of 94.6%.