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

TL;DR

This paper introduces ShuffleFAC, a highly efficient lightweight neural network designed for real-time ship-radiated sound classification on embedded devices, achieving high accuracy with minimal computational resources.

Contribution

ShuffleFAC integrates frequency-aware convolution with an efficient backbone, enabling accurate ship sound classification on resource-limited embedded systems.

Findings

Achieves 71.45% macro F1-score on DeepShip dataset.

Reduces model size by 9.7x and latency by 2.5x compared to MicroNet0.

Operates with 39K parameters and 3.06M MACs on Raspberry Pi.

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.