Abstract: The programmable Josephson voltage standard (PJVS) can generate an AC voltage reference by using stepped waves. When the step wave is generated, it usually contains a transition in each quantum voltage step, resulting in low-accuracy data at ends of voltage steps. The conventional data processing method is to filter the transition process data and use only the data of the stationary region in quantum voltage steps in order to achieve a better reproduction of AC quantum voltage value. Aiming at the commonly used data screening criterion such as 3σ criterion and piecewise sampling method, this paper proposes a robust criterion to distinguish stationary and transition regions based on bi-directionally searching the minimum type-A uncertainty from the center of the quantum voltage step. By automatically identifying the stationary and transition regions of the quantum voltage steps, the accuracy of reproducing AC voltage value based on PJVS voltage staircase waveform is improved. The experimental results show that the proposed criterion distinguishing stationary and transition regions has a good robustness and high accuracy of AC quantum voltage value reproduction. Within 2.5 kHz, the deviation of the amplitude of the AC voltage and the type-A uncertainty of amplitude is less than 2.5×10⁻⁶ for the quantum voltage step wave signal. The research of this paper can effectively improve the measurement accuracy of quantum voltage step wave signal.