Acoustic emission and electromagnetic radiation play an important role in the damage monitoring of materials such as metals, coal and rock, due to their ability to reflect the fracture process of these materials. However, acoustic emission and electromagnetic radiation signals are highly sensitive signals that are easily affected by the external environment, and the imbalance between effective signal and interference signal samples brings challenges to damage identification and diagnosis. This paper presents a signal expansion method for acoustic emission and electromagnetic radiation utilizing generative adversarial learning. The method constructs an adversarial system with generator and discriminator. The generator is dedicated to deeply mining the distribution of real signals to generate corresponding similar signals, and the discriminator distinguishes the real and generated signals according to the Wasserstein distance measure, and ensures the stability and convergence of the training process through the gradient punishment mechanism. The case study shows that the average normalized cross-correlation and structural similarity index between the expanded samples effective and interference signals and the true samples reaches 93.06% and 91.47% respectively, which provides a solution for solving the shortage of material damage signal samples in deep learning.