A practical method for detecting and removing anomalous velocity spikes in ultrasonic Doppler velocity-profiling (UDVP) data is presented. UDVP applies the pulsed ultrasonic Doppler effect to measure instantaneous flow velocities, using the rate of movement of suspended particles within the flow as a proxy. This technique has become popular in laboratory studies of sediment erosion, transport, and deposition, because of its ability to acquire high-resolution velocimetric data at a wide range of suspended-sediment concentrations in, for example, hyperconcentrated river flows, transient turbulent mudflows, and turbidity currents. Velocity data recorded by UDVP transducers that are not part of the genuine velocity signal are collectively known as velocity spikes. The new technique to detect velocity spikes proposed here is based on a simple, combined acceleration and magnitude threshold method, with a piecewise Hermite cubic interpolation of four adjacent velocity measurements used to remove the detected velocity spikes. 110 datasets covering different combinations of velocity and suspended-sediment concentration were used to test the MatLab-coded despiking algorithm, and to make recommendations for best practice in future UDVP-based studies.