Open-Set Signal Recognition Based on Transformer and Wasserstein Distance

Open-set signal recognition provides a new approach for verifying the robustness of models by introducing novel unknown classes into model testing and breaking conventional closed-set assumption, which has become very popular in real-world scenarios. In present work, we propose an efficient open-set algorithm, contains three key sub-modules: representation sub-module based on vision transformer (ViT) structure, set distance metric Wasserstein distance, class space compression reciprocal point separation central loss. this representing features signals are established transformer-based neural networks, i.e., ViT, order to extract global information about time series-related data. The employed is used modeling potential without using corresponding samples, while between different spaces mathematically modeled terms instead classical Euclidean distance. Numerical experiments tasks show that proposed algorithm can significantly improve efficiency both known categories.