Ultra-Lightweight Network for Ship-Radiated Sound Classification on Embedded Deployment

2026-01-20

AI Summary

This work proposes ShuffleFAC, a lightweight acoustic model tailored for efficient and real-time classification of ship-radiated noise in resource-constrained maritime monitoring scenarios. By integrating frequency-aware convolution with channel shuffling and leveraging depthwise separable and pointwise grouped convolutions, the model substantially reduces computational overhead. Evaluated on the DeepShip dataset, ShuffleFAC achieves a macro F1-score of 71.45% with only 39K parameters and 3.06M MACs. On a Raspberry Pi, it attains an inference latency as low as 6.05 ms—2.5× faster than MicroNet0—while being 9.7× smaller in model size, demonstrating exceptional deployment efficiency without compromising accuracy.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Planning, Routing, and Scheduling: Planning with Language Models
• Natural Language Processing: Speech

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

This letter presents ShuffleFAC, a lightweight acoustic model for ship-radiated sound classification in resource-constrained maritime monitoring systems. ShuffleFAC integrates Frequency-Aware convolution into an efficiency-oriented backbone using separable convolution, point-wise group convolution, and channel shuffle, enabling frequency-sensitive feature extraction with low computational cost. Experiments on the DeepShip dataset show that ShuffleFAC achieves competitive performance with substantially reduced complexity. In particular, ShuffleFAC ($\gamma=16$) attains a macro F1-score of 71.45 $\pm$ 1.18% using 39K parameters and 3.06M MACs, and achieves an inference latency of 6.05 $\pm$ 0.95ms on a Raspberry Pi. Compared with MicroNet0, it improves macro F1-score by 1.82 % while reducing model size by 9.7x and latency by 2.5x. These results indicate that ShuffleFAC is suitable for real-time embedded UATR.

Problem

• ship-radiated sound classification
• embedded deployment
• ultra-lightweight network
• resource-constrained systems
• real-time acoustic recognition

Innovation

• ShuffleFAC
• Frequency-Aware Convolution
• Lightweight Neural Network
• Embedded Acoustic Classification
• Channel Shuffle