Flow velocity is an important parameter in the multi-phase flow detection field. Electrical resistance tomography (ERT) is an advanced and non-invasive measuring technique for velocity estimation that has received considerable interest in the past decade. The process involves using two adjacent ERT sensors to collect measurements in a cross-sectional pipe, and the application of the cross-correlation (CC) principle for computing flow velocity. Currently, however, inaccurate assumptions, uncertain parameters, and noisy data greatly limit ERT efficiency and accuracy in practice. A new flow velocity computation method is presented that uses a single ERT sensor to replace two adjacent ERT sensors. According to the measurements of the single sensor and the corresponding ERT images, two key variables are computed that tightly align with the flow velocity. The proposed method overcomes the limitations of existing flow computation methods to some extents. Moreover, it can compute flow velocities in not only solid-liquid but also gas-liquid two-phase flows. Finally, experiments are conducted to validate the proposed method, and demonstrate its accuracy and reliability over existing methods.