Lightweight and Generalizable Acoustic Scene Representations via Contrastive Fine-Tuning and Distillation

ABSTRACT Acoustic scene classification (ASC) models on edge devices typically operate under fixed class assumptions, lacking the transferability needed for real-world applications that require adaptation to new or refined acoustic categories. We propose ContrastASC, which learns generalizable acoustic scene representations by structuring the embedding space to preserve semantic relationships between scenes, enabling adaptation to unseen categories without retraining. Our approach combines supervised contrastive fine-tuning of pre-trained models with contrastive representation distillation to transfer this structured knowledge to compact student models. Our evaluation shows that ContrastASC demonstrates improved few-shot adaptation to unseen categories while maintaining strong closed-set performance. Index T erms-- Acoustic Scene Classification, Contrastive Learning, Knowledge Distillation, Model Fine-tuning 1. INTRODUCTION Acoustic scene classification (ASC) has attracted significant research attention as a crucial capability for context-aware AI systems on edge devices [1, 2].