City2Scene: Improving Acoustic Scene Classification with City Features

2025-03-21

AI Summary

To address the limited generalization of Acoustic Scene Classification (ASC) models caused by inter-city acoustic environmental disparities, this paper proposes a scene classification method that integrates city-specific acoustic priors. The core method introduces city classification as an auxiliary supervision task and transfers city-level acoustic discriminative knowledge to the primary ASC model via knowledge distillation. Additionally, it establishes a city-scene joint labeling framework, compatible with both CNN- and Transformer-based backbones, to jointly learn cross-city invariant features and city-specific acoustic cues. Evaluated on the DCASE 2023 Task 1 benchmark, the approach consistently improves the accuracy of multiple state-of-the-art ASC models—achieving an average gain of +1.8%. Empirical results demonstrate that explicitly modeling city-level acoustic variations significantly enhances the robustness and generalization capability of ASC systems.

Technology Category

Machine Learning: Multi-class/Multi-label Learning & Extreme ClassificationComputer Vision: Scene Analysis & UnderstandingSearch and Optimization: Learning to Search

Application Category

Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMsSearch and Retrieval-Augmented AI: Multilingual and cross-lingual Web searchSystems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Acoustic scene recordings are often collected from a diverse range of cities. Most existing acoustic scene classification (ASC) approaches focus on identifying common acoustic scene patterns across cities to enhance generalization. In contrast, we hypothesize that city-specific environmental and cultural differences in acoustic features are beneficial for the ASC task. In this paper, we introduce City2Scene, a novel framework that leverages city features to improve ASC. City2Scene transfers the city-specific knowledge from city classification models to a scene classification model using knowledge distillation. We evaluated City2Scene on the DCASE Challenge Task 1 datasets, where each audio clip is annotated with both scene and city labels. Experimental results demonstrate that city features provide valuable information for classifying scenes. By distilling the city-specific knowledge, City2Scene effectively improves accuracy for various state-of-the-art ASC backbone models, including both CNNs and Transformers.

Problem

• Leveraging city-specific features for acoustic scene classification
• Improving ASC accuracy using city knowledge distillation
• Addressing environmental and cultural differences in acoustic scenes

Innovation

• Leverages city features for scene classification
• Uses knowledge distillation for city-specific knowledge transfer
• Improves accuracy for CNNs and Transformers models