Temporal Accumulation–Driven Intelligent Spectrum Sensing Method under Low-SNR Conditions

To address the limitation of signal detection performance in low signal-to-noise ratio (SNR) scenarios of cognitive radio systems, this paper proposes a temporal-accumulation–based intelligent spectrum sensing method that enables effective sensing across different observation durations without retraining the model. For different types of intelligent sensing models, three time-accumulation strategies are developed: input-data accumulation, perception-model accumulation, and detection-statistic accumulation. Experimental results demonstrate that the proposed method can markedly improve detection performance under low SNR and exhibits strong robustness across intelligent spectrum sensing models with different input representations and under different channel conditions, further demonstrating the effectiveness and practical engineering potential of the proposed approach.