To suppress the noise and increase the accuracy of feature extraction for ultrasonic Lamb wave signals, we present a new method based on the Tsallis mode and the fractional-order differential in this paper. Firstly, the fractional-order differentials of the amplitude spectrum of the noisy Lamb signal at different orders are obtained by using the fractional differential theory. Then, the cubic polynomial between the peak amplitude and the derivative order, and that between the peak frequency and the derivative order are proposed based on the Tsallis mode. The characteristic parameters of the amplitude spectrum are extracted with the developed equations. Finally, the Lamb signal without the noise is restored by combining the amplitude spectrum with the phase spectrum. Simulated and experimental results show that the proposed method can improve the performance parameters such as mean square error, r, and signal-to-noise ratio. Consequently, the new method based on the Tsallis mode and the fractional-order differential has the effective noise suppression performance for Lamb wave signals.