“In the realm of wireless communications, a new spectrum sensing method has been introduced to tackle the challenges posed by signal-to-noise ratio (SNR) fluctuations. The multi-scale feature fusion network (MSFFNet) proposed in this paper enhances feature extraction and utilization, particularly in low SNR scenarios. The method includes a multi-scale feature extraction block (MSFEB), an adaptive feature screening strategy (AFSS), and a multi-level feature fusion mechanism (MLFFM), which together significantly boost performance. Expert research has established the MSFFNet system, which provides solutions to improve spectrum sensing in low-SNR environments, laying a foundation for the construction of robust communication systems.”