Heterogeneous sound classification with the Broad Sound Taxonomy and Dataset

TL;DR

This work targets broad, heterogeneous sound classification using the Broad Sound Taxonomy (BST), a two-level scheme with $5$ top-level and $23$ second-level classes (total $28$), evaluated on the BSD10k dataset of over $10{,}309$ manually annotated sounds. It compares acoustic-only representations with embeddings that fuse acoustic and semantic information, notably CLAP and LAION-CLAP, using a $k$-NN baseline and hierarchical evaluation. The results show that semantic-rich embeddings, especially CLAP, yield higher accuracy across second-level ($0.761$) and top-level ($0.873$) tasks, with error analyses revealing substantial gains from integrating hierarchical information and semantic context. The dataset and findings support more robust, context-aware sound processing in real-world environments and motivate future hierarchical modeling and data curation efforts.

Abstract

Automatic sound classification has a wide range of applications in machine listening, enabling context-aware sound processing and understanding. This paper explores methodologies for automatically classifying heterogeneous sounds characterized by high intra-class variability. Our study evaluates the classification task using the Broad Sound Taxonomy, a two-level taxonomy comprising 28 classes designed to cover a heterogeneous range of sounds with semantic distinctions tailored for practical user applications. We construct a dataset through manual annotation to ensure accuracy, diverse representation within each class and relevance in real-world scenarios. We compare a variety of both traditional and modern machine learning approaches to establish a baseline for the task of heterogeneous sound classification. We investigate the role of input features, specifically examining how acoustically derived sound representations compare to embeddings extracted with pre-trained deep neural networks that capture both acoustic and semantic information about sounds. Experimental results illustrate that audio embeddings encoding acoustic and semantic information achieve higher accuracy in the classification task. After careful analysis of classification errors, we identify some underlying reasons for failure and propose actions to mitigate them. The paper highlights the need for deeper exploration of all stages of classification, understanding the data and adopting methodologies capable of effectively handling data complexity and generalizing in real-world sound environments.

Figures (2)

• Figure 1: Class hierarchy for the Broad Sound Taxonomy (BST).
• Figure 2: Confusion matrix for the best-performing k-NN model trained with CLAP.