High counting rates pose a challenging problem for nuclear spectrometric systems, where a physical phenomenon known as the pile-up effect distorts direct measurements, resulting in a significant bias in spectrum estimation. In this paper, we propose an innovative neural network that combines the self-attention mechanism and convolutional neural network (CNN) architectures to address the problem of spectrum estimation. The approach was evaluated on spectral data from a small scintillator (NaI) and simulated signals from a dedicated simulator and compared with conventional methods; furthermore, our method was compared to the state-of-the-art (SOTA) time domain method on the Allpix2 simulator and was evaluated on a real-world dataset. The results demonstrate that the proposed method leads to a more accurate inference of the energy spectrum, even at high count rates. Experiments also show that the proposed method is robust with varying sources, different data scenarios, and noise intensities.

In nuclear spectroscopy, a physical phenomenon known as the pile-up effect distorts direct measurements by causing temporal overlap of detector pulses. Existing deep learning-based pile-up correction methods rely heavily on supervised training with simulated data, which often generalize poorly to real measurements due to simulation–experiment discrepancies. In this work, we propose a contrastive learning framework to learn robust and transferable representations directly from large-scale unlabeled real nuclear pulse signals. The detector output is segmented into physically complete pulse aggregations using a zero-crossing-based strategy, which serve as semantically coherent instances for representation learning. Physics-inspired data augmentations are designed to realistically model detector noise and bandwidth effects while preserving pulse area. A one-dimensional ResNet encoder is employed for efficient representation learning. The learned representations are transferred to pile-up identification and counting-rate estimation tasks. Experimental results on real nuclear radiation detection systems demonstrate that our method achieves strong performance and robustness under high counting-rate conditions, with particularly pronounced advantages in challenging peak pile-up scenarios.