Ship-radiated noise (SRN) recognition is vital for underwater acoustics, with applications in both military and civilian fields. Traditional manual recognition by sonar operators is inefficient and error-prone, motivating the development of automated recognition systems. However, most existing deep learning approaches demand high computational resources, limiting their deployment on resource-constrained edge devices. To overcome this challenge, we propose LW-MS-LFTFNet, a lightweight model informed by time-frequency analysis of SRN that highlights the critical role of low-frequency components. The network integrates a multi-scale depthwise separable convolutional backbone with CBAM attention for efficient spectral representation, along with two LSTM-based modules to capture temporal dependencies in low-frequency bands. Experiments on the DeepShip dataset show that LW-MS-LFTFNet achieves 75.04% accuracy with only 0.85 M parameters, 0.38 GMACs, and 3.27 MB of storage, outperforming representative lightweight architectures. Ablation studies further confirm that low-frequency temporal modules contribute complementary gains, improving accuracy by 2.64% with minimal overhead. Guided by domain-specific priors derived from time-frequency pattern analysis, LW-MS-LFTFNet achieves efficient and accurate SRN recognition with strong potential for edge deployment.

This paper presents a study of position-dependent signal propagation delay in large-pitch pixelated AC-coupled Low-Gain Avalanche Detectors (AC-LGADs). In AC-LGADs, a continuous resistive ${\mathrm{N}}^{+}$ layer and segmented AC-coupled readout electrodes enable charge sharing and simultaneous timing and position measurements. However, lateral signal transport in the resistive layer can introduce a position-dependent delay in the measured signal arrival time. In this work, an IHEP-designed pixel AC-LGAD was characterized using a two-dimensional picosecond laser scan. The measured leading-edge arrival time shows an approximately linear dependence on an effective propagation distance, with a delay slope of about $194.7\pm 1.3\mathrm{ps}/\mathrm{mm}$ for the tested device. After applying a position-dependent delay correction, the sigma of the combined arrival-time distribution over the scanned region is reduced from 88.3 ps to 48.6 ps. To interpret the observed delay, an equivalent two-dimensional lossy transmission-line model is developed for the continuous resistive layer. The model provides a semi-quantitative description of the leading-edge delay and indicates that, within the measured signal bandwidth, the transport is dominated by the resistive term and is therefore dispersive and diffusion-like. A distributed SPICE network including the pad-area response and capacitive charge sharing provides a complementary circuit-level cross-check of the approximately linear distance dependence. These results quantify the propagation-induced timing delay in large-pitch AC-LGADs and provide guidance for timing correction and future optimization of the resistive-layer sheet resistance.